text
stringlengths 65
113k
|
|---|
In recent years there has been extensive research on malformations of cortical development (MCDs) that result in clinical features like developmental delay, intellectual disability, and drug-resistant epilepsy (DRE). Various studies highlighted the contribution of microtubule-associated genes (including tubulin and kinesin encoding genes) in MCD development. It has been reported that <i>de novo</i> mutations in <i>KIF2A</i>, a member of the <i>kinesin-13</i> family, are linked to brain malformations and DRE. Although it is known that KIF2A functions by regulating microtubule depolymerization via an ATP-driven process, <i>in vivo</i> implications of <i>KIF2A</i> loss of function remain partly unclear. Here, we present a novel <i>kif2a</i> knock-out zebrafish model, showing hypoactivity, habituation deficits, pentylenetetrazole-induced seizure susceptibility and microcephaly, as well as neuronal cell proliferation defects and increased apoptosis. Interestingly, <i>kif2a</i><sup>-/-</sup> larvae survived until adulthood and were fertile. Notably, our <i>kif2a</i> zebrafish knock-out model demonstrated many phenotypic similarities to <i>KIF2A</i> mouse models. This study provides valuable insights into the functional importance of <i>kif2a</i> in zebrafish and phenotypical hallmarks related to <i>KIF2A</i> mutations. Ultimately, this model could be used in a future search for more effective therapies that alleviate the clinical symptoms typically associated with MCDs. |
The organization of neural circuits that form the locomotor central pattern generator (CPG) and provide flexor-extensor and left-right coordination of neuronal activity remains largely unknown. However, significant progress has been made in the molecular/genetic identification of several types of spinal interneurons, including V0 (V0D and V0V subtypes), V1, V2a, V2b, V3, and Shox2, among others. The possible functional roles of these interneurons can be suggested from changes in the locomotor pattern generated in mutant mice lacking particular neuron types. Computational modeling of spinal circuits may complement these studies by bringing together data from different experimental studies and proposing the possible connectivity of these interneurons that may define rhythm generation, flexor-extensor interactions on each side of the cord, and commissural interactions between left and right circuits. This review focuses on the analysis of potential architectures of spinal circuits that can reproduce recent results and suggest common explanations for a series of experimental data on genetically identified spinal interneurons, including the consequences of their genetic ablation, and provides important insights into the organization of the spinal CPG and neural control of locomotion. |
Meta-analysis of voxel-based morphometry dyslexia studies and direct analysis of 293 reading disability and control cases from six different research sites were performed to characterize defining gray matter features of reading disability. These analyses demonstrated consistently lower gray matter volume in left posterior superior temporal sulcus/middle temporal gyrus regions and left orbitofrontal gyrus/pars orbitalis regions. Gray matter volume within both of these regions significantly predicted individual variation in reading comprehension after correcting for multiple comparisons. These regional gray matter differences were observed across published studies and in the multisite dataset after controlling for potential age and gender effects, and despite increased anatomical variance in the reading disability group, but were not significant after controlling for total gray matter volume. Thus, the orbitofrontal and posterior superior temporal sulcus gray matter findings are relatively reliable effects that appear to be dependent on cases with low total gray matter volume. The results are considered in the context of genetics studies linking orbitofrontal and superior temporal sulcus regions to alleles that confer risk for reading disability. |
The brainstem pre-Bötzinger complex (preBötC) generates inspiratory breathing rhythms, but which neurons comprise its rhythmogenic core? Dbx1-derived neurons may play the preeminent role in rhythm generation, an idea well founded at perinatal stages of development but incompletely evaluated in adulthood. We expressed archaerhodopsin or channelrhodopsin in Dbx1 preBötC neurons in intact adult mice to interrogate their function. Prolonged photoinhibition slowed down or stopped breathing, whereas prolonged photostimulation sped up breathing. Brief inspiratory-phase photoinhibition evoked the next breath earlier than expected, whereas brief expiratory-phase photoinhibition delayed the subsequent breath. Conversely, brief inspiratory-phase photostimulation increased inspiratory duration and delayed the subsequent breath, whereas brief expiratory-phase photostimulation evoked the next breath earlier than expected. Because they govern the frequency and precise timing of breaths in awake adult mice with sensorimotor feedback intact, Dbx1 preBötC neurons constitute an essential core component of the inspiratory oscillator, knowledge directly relevant to human health and physiology. |
<b>Highlighted Research Paper:</b> Moment-to-Moment Fluctuations in Neuronal Excitability Bias Subjective Perception Rather than Strategic Decision-Making, by Luca Iemi and Niko A. Busch. |
Grasping is an action engraved in the human genome, enabling newborn infants to hang from a monkey-bar immediately after birth. The grasp force provides rich information about the brain’s control of arm movements. In this study, we tested the hypothesis that the grasp force increases to improve the hand’s movement precision during reaching. In two reaching experiments, subjects increased grasp force to suppress movement imprecision that arose from both self-generated motor noise and from an unpredictable environment. Furthermore, the grasp force did not increase constantly, but increased specifically along the movement where the hand’s deviation was greatest. The increased grasp was premeditated and was not a reaction to environmental forces, suggesting that the central nervous system has a predictive, state-dependent model of movement precision during reaching. The grasp force provides a high temporal resolution and calibration-less estimate of movement precision adaptation.
## Significance Statement
Humans use their hands on a daily basis to interact with the environment. Many tasks require the hand’s movement to be precise. Standard measures of movement precision resort to measuring the stiffness of the arm, which is notoriously difficult to measure during motion. We show that the power grasp force is correlated with movement precision, and that it provides a real-time measure of movement precision adaptation. Furthermore, the grasp force measure reveals that the brain has a state-dependent adaptation of movement precision, such that it increased grasped force in locations where the hand’s deviation was greatest.
## Introduction
Grasping with the hand is a fundamental motor action in humans that can be evoked in infants ( ), alongside the traction response, where the passive stretching of the shoulder abductors and the arm’s flexors cause the fingers, elbow, and shoulder flexors to flex in synergistic response ( ). As infants mature, their arm movements become smoother and more precise ( ). As adults, humans rarely make mistakes when moving the arm, like when reaching to grab a mug. However, some skilled movements that require precision are difficult even for adults.
Taking hammering as an example, the hammer must strike the nail head precisely, which is challenging due to self-generated motor noise ( ). The head of the hammer must remain flat against the nail during contact, which is difficult as unpredictable contact forces can destabilize the hammer ( ). Failure at such tasks occurs when the hand’s movement is perturbed unpredictably because the central nervous system (CNS) uses delayed sensory feedback to correct its movement ( ). Thus, both unpredictable self-generated motor noise and environmental interactions result in reduced movement precision that cannot be corrected immediately by the CNS. It should be noted that precision is different from accuracy, as precision relates to variance whereas accuracy refers to bias.
A vast literature exists on how humans adapt to a force field that perturbs the accuracy of the hand’s motion when reaching toward a point target ( ). The pioneering study of revealed the ability of the CNS to learn to move in a velocity-dependent force field. Before the introduction of the force field, the hand moves from one point to another in a straight but somewhat curved trajectory. When the force field is introduced, the hand’s trajectory curves outward and causes the subject to miss the intended target. Movement accuracy is regained by learning the dynamics of the force field and countering the force field’s effects on the hand via appropriate forces produced by the hand ( ).
To recover movement accuracy in the force field, a forward model of the force field’s dynamics is learned by the CNS. However, if the external forces are unpredictable and cause movement imprecision, the CNS uses a different strategy of coactivating the muscle pairs in the arm to increase its stiffness, which reduces the impact of unpredictable forces on the movement of the arm ( ; ; ). Thus, the CNS’ adaptation of movement precision in the presence of unpredictable external forces has been estimated by measuring the stiffness or the cocontraction of the arm ( ).
A recent study reported another method of measuring the CNS’ adaptation to unpredictable forces. The authors of this study measured increases in the pinch grip force when subjects reached in an unpredictable force field ( ). They found a positive correlation between the pinch grip force and the variability of the external forces. One issue with the pinch grip methodology is the strong coupling between the pinch grip force and the load force, which may confound the interpretation of the data. Furthermore, did not test whether subjects increase pinch grip force when increased movement precision is desired in the absence of external forces.
In this study, we hypothesize that the power grasp force positively correlates with the desire to increase movement precision. We test two predictions based on this hypothesis in two experiments. First, we test the hypothesis that the grasp force increases when subjects want to improve movement precision. This hypothesis is tested by asking subjects to keep their hand within a wide or narrow visual track while reaching toward a target, and measuring the changes in the grasp force. Second, we hypothesize that changes in grasp force reflect a desire to improve movement precision, and do not reflect the actual movement precision per se. This second hypothesis is tested by asking subjects to reach in a diverging force field that pushed their hand laterally from the center line. As subjects were unaware of when the force field would activate, we predicted that the grasp force would not increase in response to the force field on the first trial. We also predicted that in catch trials, where the force field was unexpectedly turned off, the grasp force would not decrease although the actual movement precision was high.
## Materials and Methods
Ten male subjects, who all gave their informed consent, participated in the study. The experiment was conducted in accordance with the Declaration of Helsinki, and the study was approved by the Ethical Review Board for epidemiological Studies at the Tokyo Institute of Technology.
The subjects were seated facing the KINARM planar robotic manipulandum from BKIN Technologies ( ). Subjects held onto the KINARM interface via a handle that was affixed with a three-axis force sensor (Tec Gihan) to measure the grasp force from the palm of the hand during reaching movements. An Edero Armon arm support was used under the elbow to support the arm’s weight when using the robotic interface. Visual feedback was provided on a monitor that was placed upside-down such that subjects viewed a reflection of the monitor on a thin film mirror placed above the hand, which obscured it from view. The position of the hand was visible as a white circular cursor during both the visual track and the divergent force field experiments.
Experimental setup, protocol, and results from the first experiment where subjects reached along a visual track A , Subjects held the handle of a robotic manipulandum, where the position of their arm was hidden behind a film mirror. Subjects received visual feedback of the position of their hand as a cursor on the monitor such that it could be seen above the hand. The elbow was supported such that the hand, elbow, and shoulder were level along a horizontal plane. A force sensor was placed between the palm of the hand and the handle to measure the grasp force. B , In the first experiment of reaching along a visual track, subjects were presented with two visual feedback conditions showing either a wide or a narrow track. Subjects were instructed to keep their cursor inside the track and reach the circular target at the end of the track. Subjects first made reaching movements without a visual track in a training block, after which they experienced wide and narrow blocks in consecutive sequence for three repetitions per condition. C , The group mean grasp force (solid traces) and load force (dotted traces) are plotted as a function of normalized time, where the shaded area is 1 SE. The data were separated into the wide (blue) and the narrow (red) conditions. The Pearson correlation coefficient between the grasp and load forces was not significantly different from zero. Thus, no significant correlation between grasp and load force was observed when reaching along a visual track. D , The lateral deviation and the grasp force of the population mean and SE are plotted as a function of trials with the plot color indicating the training trials (black), wide track (blue) and narrow track (red) conditions. E , The mean lateral deviation and the mean grasp force from the population is shown for the wide and narrow conditions. The lateral deviation was lower and the grasp force was higher when the visual track was narrow. F , The lateral deviation is plotted as a function of the grasp for every subject in the wide and narrow conditions, with a black line connecting the data from the same subject. The thin blue and red lines show standard deviation from the linear mixed effects model fits. An increase in grasp force was observed to subsequently reduce the lateral deviation.
### Visual track reaching experiment
Subjects moved their hand to reach a target of radius 2 cm placed 25 cm away from their initial starting position. Subjects were instructed to prevent the cursor from deviating off a red visual track that was displayed between the initial and target positions ( ).
The experimental structure consisted of seven blocks where each block contained 15 trials. The first block consisted of training trials, where the red region was absent. Blocks 2, 4, and 6 were the “wide” condition that demanded low movement accuracy blocks as the visual track was ±4 cm wide. On the other hand, the “narrow” condition tested in blocks 3, 5, and 7 required high movement accuracy due to a narrow visual track with a width of ±0.4 cm. The cursor in this experiment had a diameter of 0.4 cm. Feedback of the movement duration was provided to the subject on a trial-by-trial basis. Movements that were faster than 900 ms and longer than 1100 ms triggered a feedback message of “fast” or “slow,” respectively, to ensure that subjects reached with comparable movement speeds in both the wide and narrow conditions. No feedback was given of the lateral deviation after the trial.
As we observed a linear relationship between both the lateral deviation and the grasp force as a function of trials, we fitted these data from the visual track experiment using the linear mixed-effects model of the form where the response Y is either the vector of data from grasp force F or from the lateral deviation x , T is the trial number, C is the visual track condition (narrow or wide), is the intercept, to are the parameters for each predictor, and is the unexplained variance of the response Y for each subject s .
A likelihood ratio test was employed to examine the significance of the condition parameter C on explaining the data. If deemed significant, this implied that the width of the visual track had a significant impact on the grasp force and the lateral deviation.
If the condition was deemed to significantly influence the grasp force or the lateral deviation, a one-sample t test was conducted on the data, which was grouped separately for the wide and narrow conditions. This enabled us to test our hypothesis of whether the grasp force increased when reaching along a narrow visual track and whether a reduction in lateral deviation was observed in the narrow track. However, these tests alone were not sufficient to establish a relationship between the grasp force and the lateral deviation as they only examine the effect on a group level. We examined how each subject’s lateral deviation changed as a function of the change in their grasp force. Here, a non-parametric sign test was employed as these data were observed to violate normality. The normality of all datasets was tested using an Anderson–Darling test before post hoc testing.
### Divergent force field experiment
The same 10 subjects that took part in the first experiment participated in the divergent force field task ( ). Subjects were instructed to reach a target 20 cm away from the starting position. The diameter of the cursor was 1 cm in this experiment. Feedback was provided on a trial-by-trial basis about the duration of the movement, which had to be between 500 and 600 ms. Each participant practiced a standard reaching task for 15 training trials, after which they experienced 25 divergent field blocks. The divergent force field was designed to amplify lateral reaching errors by applying the following force to the hand, where the stiffness and x is the lateral position of the hand such that the initial and target positions are at . Each divergent field block was composed of four trials where the last trial was a catch trial where the force field was switched off, i.e., . The catch trials tested whether subjects were simply grasping the handle as a reaction to the forces from the robot, or were increasing their grasp force to improve their movement precision. If the hand’s lateral deviation was greater than ±4 cm in a force field trial, the force field was switched off and the subjects were shown a “failed” trial message.
The divergent force field reaching experiment supports the hypothesis that the grasp force increases when better movement precision is desired, and not when the actual movement precision changes. A , A schematic of the experiment and its protocol. The force field pushed the hand away laterally if it deviated from the line that connected the start and the target positions. To succeed, a subject must reach as straight as possible with minimum lateral deviation. Subjects experienced 15 training trials in the null field conditions, after which they experienced 100 force field trials. Of these 100 trials, every fourth trial was a catch trial where the force field was switched off. B , The first 15 trajectories in the force field trials from a sample subject are shown, where the gray trajectories show failed trials where their hand hit the safety margins placed 4 cm to the left and right of the center line. A trial was successful when subjects stopped inside the red target. C , The mean grasp force from all subjects, averaged over each trial, is plotted as a function of trials. In training trials (black), the grasp force continually declines. The grasp force in the first force field trial is similar to the level observed in training trials, but begins to increase until it peaks at approximately the fourth force field trial. Although the grasp force declines during the force field trials, it never reaches the same level as the training trials. Furthermore, the grasp force in catch trials is indistinguishable from the force field trials, revealing that the grasp is not a reaction to the forces from the force field, but is premeditated. D , The group mean grasp force from the last training trial (dashed blue trace) and the first force field trial (solid blue trace) are plotted as a function of normalized time. On the same figure, the group mean perpendicular distance from the last training trial (dashed red trace) and the first force field trial (solid red trace) are plotted as a function of normalized time. Although the perpendicular distance increased dramatically due to the force field, the grasp force remained constant. E , The group mean grasp force (blue traces) and the group mean perpendicular distance (red traces) from all catch trials (dashed blue trace) and all force field trials (solid blue trace) are plotted as a function of normalized time. The grasp force was similar between the catch trials and the force field trials, although the perpendicular distance was smaller in catch trials where the force field was switched off.
Post hoc one-sample t tests were conducted to examine the difference between the last training trial and the first force field trial, and the difference between the catch trials and the force field trials. If the grasp force is different in either of these cases, our hypothesis must be rejected.
## Results
### Experiment 1: reaching along a wide or narrow visual track
In the first experiment, subjects had to make point-to-point reaching movements toward a circular target of radius 2 cm that was placed 25 cm away from the initial starting position. Subjects were instructed to prevent the cursor from deviating off a red visual track that was displayed between the initial and target positions ( ).
The literature reports that, with a pinch grip, significant correlation is observed between the pinch grip force and the load force ( ). Such a correlation could undermine our study as the grasp force may simply reflect the load experienced by the hand during reaching. The grasp force and the load force are plotted as a function of normalized time in . The data from every trial’s whole movement were selected for this analysis, where the start of the trial was when the target appeared, and the end was when the hand stopped inside the target. We calculated the Pearson correlation coefficient in all trials between the grasp force and the load force supplied by the subject against the robotic interface. If the load force has a significant impact on the grasp force measurement, this must be taken into account in subsequent analysis as it may influence the results. However, we found that the group mean correlation between the grasp and load force was (mean and SE), which was not significantly different from zero (one-sample t test, t = 2.09, p > 0.07). Thus, no significant coupling was observed between the grasp force and load force in this experiment.
Next, we examined whether the variance in the reaching movement was different between the wide and narrow conditions. We normalized all trajectories in time and calculated the mean trajectory for each wide and narrow condition using the trials from all three blocks. We then calculated the lateral deviation, defined as the absolute distance halfway into the movement between the cursor’s x -axis position and the mean trajectory, for each trial, which is plotted as a function of trials for the population mean in . The lateral deviation appeared to be functionally dependent on the width of the visual track. We employed a fit with a linear mixed-effects model (Eq. 1) on the data from all trials, which was labeled by subject, trial number and track condition. A likelihood ratio test revealed a significant effect of the visual track condition on the lateral deviation (χ (2) = 21.95, p < 10 ). Using the regressed linear model, we calculated the difference in the lateral deviation between the wide and narrow track conditions for the trial range 16–115, i.e., blocks 2–7. A one-sample t test on the lateral deviation showed that the subjects’ mean lateral deviation, shown in , was significantly smaller in the narrow (0.17 ± 0.02 cm) than the wide (0.23 ± 0.03 cm) condition ( t = –4.32, p < 0.0019), indicating that the subjects’ trajectories were more precise in the narrow condition. What facilitated the subjects’ ability to improve their lateral movement precision in the narrow track?
If the grasp force is correlated with movement precision during reaching, a selective increase in grasp force should be observed in the narrow condition where smaller lateral deviation was observed. The population mean grasp force from each trial is plotted as a function of trials in , where the blue and red points are from the wide and narrow conditions, respectively. The average grasp force from a single trial was calculated using data from the entire movement, where the start of the trial was when the target appeared, and the end of the trial was when the hand stopped inside the target. On visual inspection, there appeared to be a functional dependence of the grasp force on the visual track condition and the trial number. The grasp force data from all trials and all subjects were fit with a linear mixed-effects model with the trial number and track condition as predictors (refer to Eq 1). A likelihood ratio test showed that the grasp force was significantly affected by the visual track condition (χ (2) = 118, p < 10 ). We calculated the mean grasp force in the wide and narrow conditions from the linear model fits, and a one-sample t test showed that the grasp force was significantly higher in the narrow than in the wide condition ( t = 4.41, p < 0.0012).
To directly assess the effect of the grasp force on the movement, we plotted the lateral deviation as a function of the grasp force for each subject in the wide (blue) and narrow (red) conditions in , with a black line connecting the data from the same subject. Data from each subject were averaged across all three blocks in the wide and narrow conditions to yield one data point per subject per condition. An increase in grasp force was observed to reduce the hand’s lateral deviation, and a non-parametric sign test, which was employed since the data violated normality according to an Anderson–Darling test, found this relationship to be significant ( p < 0.022). In summary, these results suggest that the grasp force is related to the hand’s movement precision during reaching.
### Experiment 2: divergent force field
In the second experiment, we tested subjects reaching in a divergent force field. The force field applied a force that pushed the hand laterally away from the center line if it deviated laterally from the straight line between the initial and target positions ( ). The same 10 subjects that took part in the first experiment participated in the divergent force field task ( ). The trajectories from the first 15 trials inside the divergent force field from a sample subject are plotted in . The grasp force, averaged over each trial, is plotted as a function of trials in , where the points are the group mean and the shaded area is the SE from all 10 subjects. In the 15 training trials, where subjects reached toward the target without the force field, the grasp force was generally low ( , black dots) and continually declined with practice. On trial 16, when the force field was first experienced by subjects, the grasp force was effectively the same as the last training trial. The grasp force increased rapidly in the second and third force field trials, and peaked at approximately the fourth force field trial, after which the grasp force declined exponentially but not to the original level observed in the training trials.
Recall that the second prediction from our hypothesis dictates that increases in the grasp force should only be related to a desired increase in movement precision. Thus, the grasp force should not change even if the actual movement precision increases or decreases. The movement precision in this force field experiment is denoted by the perpendicular distance, defined as the absolute distance from the line at x = 0.
First, we examined how the grasp force and the perpendicular distance changed from the last training trial ( , dashed trace) to the first force field trial (solid trace). The group mean grasp force and the group mean perpendicular distance are plotted, in , as a function of normalized time, where time 0 was the time of target onset and the end was where the subject reached the target. The perpendicular distance was observed to increase dramatically on the first force field trial. The group mean grasp force in the last training trial was 3.1 ± 0.7 N (mean and SE) and for the first force field trial it was 2.9 ± 0.7 N. A paired sample t test found that the difference in the grasp force between the last training trial and the first force field trial was not statistically significant ( t = 2.20, p > 0.055). Although the perpendicular distance increased dramatically, the grasp force did not change.
Next, we examined whether the grasp force was different on catch trials in comparison to the force field trials. The group mean grasp force (blue traces) and the perpendicular distance (dashed traces) from all catch trials (dashed traces) and all force field trials (solid traces) is plotted as a function of normalized time. Notably, the grasp force profile is different from the training trial in , where it was constant throughout the movement. The grasp force appears to increase in tandem with the perpendicular distance in the force field trials. We found the mean grasp force, taken over the whole trial, and calculated the difference between the mean grasp force in catch trials with the neighboring force field trials. This difference was 0.08 ± 0.12 N, which was statistically not different from zero (one-sample t test, t = 0.65, p > 0.53). Hence, the grasp force did not change in catch trials, although the perpendicular distance clearly decreased.
## Discussion
In this study, we measured the power grasp force while subjects completed two reaching tasks. The first task asked subjects to stay within a visual track during reaching, and the second task had subjects reach to a target while their hand was perturbed by a diverging force field that amplified lateral reaching errors. The results from both experiments support our hypothesis that the grasp force is positively correlated with the desire to increase movement precision. Namely, the grasp force increased when the visual track was narrower and required higher movement precision. Furthermore, the grasp force in the force field did not change in tandem with changes in the actual perpendicular distance, but with the desire to change it.
The latter observation, that the grasp force was not significantly correlated with the load force from the force field, is of importance. Several studies have reported the high correlation between the pinch grip force and the load force from the environment ( ; ). This coupling had to be taken into account when subjects adapted their pinch grip force when reaching in a variable force field ( ) to avoid the confound where a change in the grip force may be mistaken for an adaptation to the variability of the force field. Instead, the change may have been due to the increased load force from the variable force field. In our experiment, we employed a divergent force field, which has a similar effect to the variable force field used in , namely that unpredictable forces are imposed on the subject’s hand that cause movement variability. Unlike the pinch grip force, the power grasp force during reaching was not correlated with the load force from the force field. The power grasp force thus avoids confounds when interpreting changes in the grasp force. However, this may be valid only when the forces from the force field are approximately orthogonal to the placement of the grasp force sensor, and so caution is still required when interpreting the changes in the grasp force.
In both experiments, the grasp force increased from the initial exposure to a visual or force field condition, but gradually decayed as a function of trials. As reported in , two opposing adaptation phenomena are likely at play. The first is a fast and sensitive adaptation that increases grasp force to rapidly improve movement precision due to a task demand. The second is a slower adaptation process that optimizes grasp force to conserve effort ( ; ). The summed contributions of both fast and slow adaptation processes explain why the grasp force increases rapidly from initial exposure, but continually decays throughout our experiment. This decay can be observed in the training trials in both experiments, implying that the gradual decay in the grasp force must be taken into account when interpreting the data. In our first experiment, this was accounted for by the linear mixed-effects model that included a trial gradient. The decay in the grasp force does not contradict our hypothesis, but it shows that an effort conservation process is continually working to reduce excessive grasp force production.
In addition to these adaptation processes, we observed that the increase in grasp force, when switching from a wide to a narrow visual track, was less pronounced in the later trials. Subjects may have learned to update their motion plan to move straighter without having to rely on increasing grasp force to remain inside the visual track ( ). Such a strategy was likely possible when staying inside the narrow visual track but was infeasible when reaching in the diverging force field that punished even minor lateral deviations. This may explain why the grasp force plateaus to a value above the nominal level observed in the training trials in the divergent force field experiment.
The changes in the grasp force observed when subjects learned to reach in the diverging force field are similar to the results of another study that examined the adaptation of arm cocontraction during the learning of a divergent force field ( ). In their study, they also found an initial, rapid increase in cocontraction, followed by a slow and gradual decay, which plateaued above the baseline observed in the training trials. The similarity between the grasp force observed in our study, and the arm cocontraction observed in the study of , raises the possibility of using the grasp force as a tool to probe the CNS’ desire to improve movement precision in rapid movements, such as during a golf swing ( ), where the delay introduced by the processing of the electromyography data may be detrimental to the analysis. As such, the grasp force methodology could complement existing methods to measure the cocontraction of muscles to further our understanding of the CNS’ desire to improve movement precision.
|
Striatal activity is necessary to initiate and execute sequences of actions. The main excitatory input to the striatum comes from the cortex. While it is hypothesized that motor and premotor cortico-striatal projections are important to guide striatal activity during the execution of sequences of actions, technical limitations have made this challenging to address. Here, we implemented a task in mice that allows for the study of different moments to execute a serial order sequence consisting of two subsequences of actions. Using this task, we performed electrophysiological recordings in the premotor (M2) and primary motor (M1) cortices, and state-dependent optogenetic inhibitions of their cortico-striatal projections. We show that while both M2 and M1 contain activity modulations related to the execution of self-paced sequences, mainly, the premotor cortico-striatal projections contribute to the proper execution/structuring of these sequences.
## Significance Statement
It is currently hypothesized that synapses from the primary motor (M1) and premotor (M2) cortices that innervate the striatum may guide the proper execution of sequences. Here, we evaluated this hypothesis by training animals to execute self-paced sequences: performing recordings in M2-M1 or manipulating their cortico-striatal projections during the execution of these sequences. We show that both, M2-M1 cortico-striatal projections contribute to sequence initiation, however sequence execution is predominantly influenced by M2. Remarkably the contribution of the cortico-striatal projections from M2 is mainly before the initiation of the sequence working to sustain the structure of the sequences, mainly during the beginning. These findings may have implications for pathologic conditions where the self-paced generation of sequences of actions is impaired.
## Introduction
In everyday life, we continuously move between sequences of motor actions. One of the main proposed drivers involved in the learning and execution of motor sequences are cortico-striatal projections. The study of action sequences in relation to cortico-striatal function has become increasingly important since the discovery that symptoms in patients with Parkinson’s disease and obsessive-compulsive disorder maybe be caused by disruptions to cortico-striatal projections ( ; ; ).
The striatum, the primary input to basal ganglia (BG), is a subcortical structure whose activity is necessary to initiate and execute a sequence of actions ( ; ; ; ; ). Recent evidence suggests that a specific subcircuit within the BG, the indirect pathway, is essential for the transition between subsequences ( ; ).
The striatum’s main glutamatergic inputs come from the cortex and the thalamus ( ). Several studies suggest that the cortical inputs are essential to execute motor sequences ( ; ; ; ; ; ; ; ; ; ; ; ). However, the specific contribution of cortico-striatal projections to the execution of self-paced action sequences remains unclear.
To date, it is known that the supplementary motor area in primates (SMA), which corresponds to the secondary motor cortex (M2) in rodents, is active before starting a sequence of actions ( ; ; ; ; ). SMA activity is important to adapt the behavior in response to contingency changes (block changes; ; ). These findings have led to the hypothesis that the premotor cortex guides the striatal activity to initiate and execute action sequences. In rodents, a decrease in the activity of neurons in M2 decreases the probability of alternating between two actions ( ). Strikingly however, studies using cortical lesions suggest that the striatum (BG) can control the execution of a sequence of actions independently from the cortex ( ; ). Therefore, we implemented a self-paced serial order sequence task that allows for probing the contribution of M2 cortico-striatal projections during the initiation, execution, and transition between subsequences of actions. By recording neuronal activity in premotor (M2) and primary motor (M1) cortices and performing time-dependent optogenetic inhibitions of the cortico-striatal projections, we identified specific contributions of the premotor cortico-striatal projections to the execution of self-paced serial order action sequences. Our results support a model in which the cortico-striatal terminals from M2 guide the appropriate execution of self-paced sequences of actions.
## Materials and Methods
### Animals
The institutional committee of the Cell Physiology Institute, at the National Autonomous University of Mexico, approved all procedures for the care and use of laboratory animals (protocol number FT121-17). This protocol follows the National Norm for Animals’ use (NOM-062-ZOO-1999). Male and female mice from two to three months of age at the start of experiments were used for this study. Two genotypes were used: C57BL/6J (The Jackson Laboratory, RRID: ) or Emx1-Cre mice (targeting the Cre recombinase expression in pyramidal cortical neurons), which had been backcrossed into C57BL/6J for at least six generations ( ). Emx1-Cre parental line was donated by Professor Rui M. Costa from the Champalimaud Center for the Unknown (RRID: ). All animals were obtained from our breeding colony in our institutional bioterium (the Emx1-Cre line is maintained in heterozygosis). Animals are housed under a 12/12 h light/dark cycle (lights on at 6 A.M.) with ad libitum access to food and water before beginning behavioral experiments.
### Training
We used operant conditioning boxes equipped with two retractable levers to implement forced and self-paced serial order sequences in mice (21.6 cm long × 17.8 cm wide × 12.7 cm high; Med-Associates, catalog #MED-307W-D1). One lever was positioned on the left panel (subsequence 1; S1), and the other was located on the front panel (subsequence 2; S2) on the left side of the magazine ( ). A small sugar pellet, 14 mg (Bio-Serv, catalog #F05684), was delivered as a reward in the magazine. Entries to the magazine were registered with an infrared beam. A second infrared beam was positioned between the magazine and left lever press ( ) to calculate the latency to start the sequences of actions. Mice were subjected to food restriction throughout training and given enough food after daily training sessions to keep them at 80–85% of their original weight, depending on performance.
Mice learn to execute serial order sequences of lever press. A , Timeline of training days. After 7 d of pretraining (see Materials and Methods), animals enter a phase of forced sequences (AAAA→BBBB→pellet). After 3 d, animals started training on blocks of forced and self-paced sequences, switching blocks once five correct trials are achieved per block. B , left, Diagrams of the sequential pressing on lever 1 (AAAA = 4 presses) → lever 2 (BBBB = 4 presses) delivering a reward in the magazine (pellet) in forced sequences, where levers are presented individually. Right panels, Correct forced sequence depicted with a green symbol and the two different categories of errors that mice executed in forced sequences. C , left, Diagrams of the sequential pressing on lever 1 (AAAA… >= 4 presses) → lever 2 (BBBB… >= 4 presses) delivering a reward in the magazine (pellet) in self-paced sequences, where levers are presented together. Right panels, Example of a correct sequence depicted with a green symbol and the four categories of errors that mice executed while performing self-paced sequences. D , Example of the sequences of one animal’s presses during the first day (early in training) and eleventh day (late in training) during sessions of blocks of forced–self-paced sequences. Note how the self-paced sequences become very stereotyped late in training [the blocks are indicated by light red (forced) or light blue (self-paced) at the bottom of each panel]. E , Percentage of correct sequences [correct/(errors + correct)]. Note that day 1 in the x -axis means the first day in which animals were required to perform blocks of forced and self-paced sequences as explained in A–C . However, the requirement of forced sequences (AAAA→BBBB→pellets) was present 3 d before. All curves show mean ± SEM WT ( n = 8). F , The correct forced sequences were of one type (S1 = 4→S2 = 4) but of several types for self-paced sequences, dominated by Long-S1 sequences (S > 4→S2 >= 4). G , The different categories of errors that mice executed while performing forced or self-paced sequences late in training. H , Latency to start, measured as the time between breaking the infrared beam placed outside the magazine and the first press in the sequence. I – L , Mean duration, transition, interpress intervals, number of presses from WT animals ( n = 8) measured late in training; * p < 0.05, specific p values and statistical test specified in the text and . Extended Data includes the latency, transition time, IPI along with training, and a comparison of the proportion of the different errors early and late in training. n.s. = p > 0.05.
Descriptive statistics and data extended from the main figures
For the first exposure to the operant box, animals were placed in the box without levers for 30 min. A total of 30 pellets were delivered individually at random intervals (on average every 60 s). Over the next 3 d, the animals were presented with lever 1 or lever 2 and received a reward each time they pressed the lever. After eight rewards, lever 1 was retracted, and lever 2 was presented. The session finished once the animal got 16 pellets, or 30 min had passed. Afterward, the training schedule changed to 3 d at a fixed ratio eight (FR8) schedule on each lever, whereby animals had to accumulate eight lever presses to receive a reward and retract the lever, followed by a second lever presentation, which also required 8 presses to receive the reward and retract the second lever (pretraining). The session finished when mice reached 30 rewards (15 on lever 1, 15 on lever 2) or 30 min had passed. If the animals checked the magazine before reaching eight continuous lever presses, a time out (10 s) was presented. We used 3 d in FR8 since in previous studies have seen this induces animals to press in bouts of around four presses ( ). Next, animals were trained to do forced serial order sequences (3 d). Here, animals were presented with a lever 1; they had to do four continuous lever presses for the lever to be retracted, followed by extension of a second lever (lever 2). Animals had to do four continuous lever presses on lever 2 for it to be retracted, and a reward was delivered into the magazine ( ). If the animals executed fewer than four lever presses on lever 1 or lever 2 by visiting the magazine, a time out (10 s) was presented. The session finished when mice got 30 rewards, or 30 min had passed. After these 3 d of forced sequences, the animals entered the last stage of training: blocks of forced–self-paced sequence sessions. These sessions began with a block of forced trials, switching to a block of self-paced trials in the same session. The switch between blocks was conditioned by achieving five consecutive correct sequences. For the self-paced sequences, the animals were presented with the two levers at the same time. They had to do at least four presses on lever 1 followed by at least four presses on lever 2, to obtain a reward in the magazine. Unlike in forced blocks, in self-paced blocks the animals decided when to execute the transition between the subsequences of presses from lever 1 to lever 2 ( ). Upon finishing a self-paced block, both levers are retracted for 3 s for an intertrial interval before starting a new forced block. If the animals executed fewer than four presses on lever 1 or lever 2 by visiting the magazine, a time out (10 s) was presented. The session finished once the animals received 70 rewards (35 forced and 35 self-paced rewards) or 30 min had passed. The block sessions continued for 11–13 d until a stable performance was reached; see ). All timestamps were recorded with a resolution of 10 ms with the Med-PC IV software suite (Med-Associates).
### Single-unit recording and antidromic photo-identification
To record the cortical activity, either a fixed or movable 16-electrode array [tungsten (35 μm; Innovative Neurophysiology) was implanted]. The neurons’ spikes were sorted online (Central software, Blackrock Microsystems), and clear waveforms with a signal-to-noise ratio >3:1 were used for further analysis using offline-sorting (Offline Sorter, Plexon Inc.). To define whether a recorded unit projected to the striatum, we used in vivo antidromic photo-identification ( ). In short, we injected ChR2 (UPENN-vector core catalog #AV-1-20298P) in the cortex of interest of Emx1-Cre animals. A movable electrode array was implanted 200 μm above the ChR2 expression site during the same surgery, and an optic fiber (Thorlabs catalog #CFLC230-10, FT200EMT) was implanted into the striatum (ipsilateral to injection). Using the electrode array, we could record activity of cortical neurons. At the end of the behavior session, we used a light stimulation (2 mW, 473 nm, 10 Hz, 1 s, 10-ms pulses; Laserglow) delivered by the optic fiber while recording neuronal activity. This allowed us to verify whether the recorded neuron was antidromically photo-activated (Extended Data ). The movable array allowed us to search for responding cells in at least five sessions per animal, advancing the array 100 μm 24 h before the recording. Only units that responded to the light (i.e., presented a correlation of >0.9 between the behavioral spike and the antidromic spike and presented a latency to light <10 ms) were considered.
### Per-trial rescaling of neural activity
In both sequence types, forced and self-paced, each sequence had a slightly different durations and could not be directly averaged. To mitigate this, we employed a time rescaling procedure ( ), to evaluate each recorded unit’s overall response pattern. Thus, we performed a rescaling of trials with at least four presses in each sequence, defining the following alignment events: first, second, penultimate, and last press of subsequence 1 and subsequence 2. Spike trains were transformed to instantaneous firing rate by applying a Gaussian convolution (σ 25 ms) at 100 Hz. The activity was rescaled through linear interpolation to the average interpress interval and inter sequence interval of all recorded animals’ sessions.
### Analysis of the electrophysiological recordings in vivo
Once a putative unit was isolated (through online and off-line sorting), the spikes’ timestamps were aligned to the first lever press in the sequences of actions using custom-developed scripts in MATLAB (MathWorks).
Z score test: to determine differences between the z score activity from M2 versus M1, we averaged the z score activity in the same time window from the two indicated cells and calculated the difference in terms of z score with the following equation:
where Z1 is the average normalized activity of the M2 units and Z2 for the M1; N1 is the sample of units from M2, and N2 the sample of units from M1 ( ; Extended Data ). The results were compared in a normal distribution table to determine the corresponding p value ( ).
Activity modulation in premotor and motor cortical neurons during the execution of self-paced sequences of actions. A , Image of a mouse while pressing one of the levers and photomicrographs of coronal sections of M2 (middle) or M1 cortex (right), illustrating the cannula/electrode array’s tracks in the recorded sites. B , C , Raster plots and perievent histogram from an M2 or an M1 unit, respectively, aligned to the first (P1), second (P2), penultimate (PLast-1), and last lever press (PLast) of S1 and S2 for self-paced sequences. Bottom panels, Mean firing rate from the upper panels. D , E , Z score of individual units. F , Representative firing of different cells presented as z score, illustrating the different categories of modulation during the execution of the sequences Init & S1 execu = Initiation and S1 execution; Init. S1 & S2 = Initiation of S1 and S2; Trans.(init. S2) = Transition or initiation S2; S2 execut. = S2 execution; S1 & S2 execut. = S1 and S2 execution. G , % of cells related to each category in F . H , I , Mean z score from the units recorded that presented significant modulation 1 s before the initiation of the sequence from M2 or M1. J , Comparison of the mean z score from H , I , M2 (dark blue) versus M1 (light blue) for positively modulated units; * p < 0.05, z score test. Init. = initiation, S1 = subsequence 1, Trans = transition, and S2 = subsequence 2. Extended Data includes the activity modulation in premotor and motor cortical neurons during the execution of forced sequences of actions. n.s. = p > 0.05.
### Regression analysis of the neuronal activity and behavior
We made linear regression analysis to ask whether the firing rate in each bin of time was correlated with parameters of the task. For this purpose, we first aligned the spikes to each epoch (first, second, penultimate, and last lever press for each sequence) and took 5 s before and 5 s after each epoch. Spike trains were transformed to instantaneous firing rate as described above. We calculated the spike frequency using a 200-ms time window with 10-ms steps. We separated the trials and sorted them by the number of lever presses in the sequences (2,4,6, up to 16, grouped by every increment), for the sequence and transition duration (sorted by duration in descending manner, seven categories), for latency (grouped every one second, starting with 0.5 s.). Then we made a regression analysis with permutation test in two manners. For the first, we used a bin of time of 200 ms with a sliding window of 10 ms and ask whether the variable of interest was correlated with the firing rate ( ; Extended Data ). For the second regressions analysis, we used specific windows of time: (1) mean firing rate 1 s before the start of the sequence; (2) mean of firing rate during the sequence; (3, 4, and 5) average of firing rate during S1, firing rate during transition or firing rate during S2 ( ). To resolve statistically whether the regression’s p value was significant, we ran 1000 permutations and divided the sum of times that the p value > p value initial between the number of permutations. Only regressions with β coefficient different from zero ( p < 0.05 and R > 0.6) were accepted.
Both M2 and M1 contain units encoding the temporality of the self-paced sequences of actions. A , top panels, Examples of two units recorded in M2. In each example, each line is a self-paced sequence depicting the neuronal activity (in z score, black to yellow) and lever press (white points subsequence 1, gray points subsequence 2). The color bar to the right of each plot shows the grouped categories plotted in the middle and bottom panels. Middle row panels, Mean firing rate from each category presented in the upper row. Bottom row, Regression fits from the time bin depicted in gray in the middle row panels. B , top row panels, Significant regression analysis per unit; columns i, between the firing rate (FR) 1 s before the start of the sequences (Seq) versus the duration of the sequences; columns ii, FR during the sequences versus duration of sequences; columns iii, FR in the subsequence 1 (S1) versus the duration of S1; columns vi, FR during the transition versus the transition time; and columns v, FR during the subsequence 2 (S2) versus the S2 duration. Bottom panels, The proportion of units that presented significant regression ( R > 0.6 and p < 0.05). The dashed lines depict comparisons with χ test. Corrections for multiple comparison was considered (see Materials and Methods). Extended Data shows the recorded units in M2 or M1 encoding the temporality of the forced sequences of actions. Extended Data shows the confirmation of the M2-M1 projections into the LS and linear regressions between the activity and the temporal parameters of the execution of sequences from the photo-identified M2 or M1 cortico-striatal neurons. n.s. = p > 0.05.
### ROC curve analyses and permutations
To determine the percentage of modulated units along time in each epoch, we performed a ROC curve analysis to ask whether the spike frequency in each time bin was different from baseline time. We first aligned the spikes to each epoch (e.g., first press) using 4 s before and 5 s after each epoch. We calculated the spike frequency using a 200-ms time window. We used a baseline from –4.5 to –4 s before the first press in the sequences. We compared the spike frequency in each bin of time against the baseline, using a sliding window of 10 ms. To resolve statistically whether the area under the curve (AUROC) was significant, we ran 1000 permutations and divided the sum of AUROC values that fall in either >0.5 or <0.5, by the number of permutations. The AUROC value was significant if the outcome was p < 0.05. Furthermore, in each epoch, we obtained a binary matrix comparing each bin to baseline. This binary matrix was used to find the number of units modulated in each time bin. For the linear regression, to resolve statistically whether the regression’s p value was significant, we ran 1000 permutations and divided the sum of times that the p value > p value initial between the number of permutations.
### Stereotaxic opsin injection and fiber implantation
For surgeries, animals were anesthetized using a mix of oxygen (1 l/min) and 1% isofluorane (1–2% for interventional procedures). For the optogenetic experiments: after anesthesia, each animal was bilaterally injected using glass pipettes with 500 nl of viral stock solution [either rAAV5-EF1a-DIO-eArch3.0-EYFP (Vector core, University of North Carolina), or AAV1.EF1a.DIO.eYFP.WPRE; AAV1.EF1a.DIO.hChR2(H134R)-eYFP.WPRE titer > 1 × 1012 (Vector core UPENN University Pennsylvania catalog #AV-127056)] by pressure into either the M2 or M1 or lateral striatum (LS), coordinates from bregma, M2: AP 2.34 mm, ML 1.25 mm, DV 0.60–0.70 mm. M1: AP 0.5 mm, ML 1.60 mm, DV 0.60–0.70 mm. LS: AP 0.50 mm, ML 2.50 mm, DV 2.40 mm below the brain’s surface. After the injections were done (23 nl every 5 s; Nanoject II, Drummond Scientific), we waited 15 min to allow time for virus to spread, and a fiber-optic (300 μm; NA 0.37) was implanted into each hemisphere of the striatum. The optical fibers were fixed to the skull using acrylic cement (Lang Dental Manufacturing Co, Inc).
### Retrobead injections
For the retrograde labeling experiments, 300 nl of retrobeads (Lumafluor) were injected into the LS. Coronal sections (50 μm) were obtained to determine the total number of cells labeled in the M2 or M1. The quantification was done in one slice every 300 μm covering these regions.
### Retro-Cre injections
Similarly, to the retrobead injections, 300 nl of mCherry-Retro-Cre (Addgene catalog #55632-AAVrg, RRID: ) AAV were injected into the LS, and 300 nl of DIO-eYFP were injected into the M2 cortex of Emx1-Cre. Coronal striatal sections of 50 μm were obtained to determine the axons crossing by the dorsomedial or the dorsolateral striatum.
### Cortico-striatal fibers quantification
The axonal quantification protocol was as previously reported ( ). In short, we extracted the brains and sectioned the striatum (50-μm coronal sections). Z stacks at 63× magnification were acquired (192 × 192 × 10 μm; 1-μm interslice) from a random quadrant; using a randomly positioned grid covering either the dorsolateral or the dorsomedial striatum (ZEN lite software, Zeiss, LSM 710). These Z stacks were imported into ImageJ; then, a maximum projection image was used to apply a filter (Hessian filter), allowing the quantification of fibers as defined by the number of fibers crossing a randomly generated line spaced ∼20 μm.
### Cortical microstimulation of forelimb region in the M1
Micro-stimulation experiments were performed to identify the M1’s coordinates corresponding to the contralateral forelimb region. The animal was placed in the stereotaxic apparatus under anesthesia (ketamine 0.15 mg/g mouse + xylazine 0.01 mg/g mouse). Access to motor cortex was achieved by trepanning a window of at least 1 mm in diameter around the center point (AP +1, ML 1). Electrical stimulation was performed using a 300-μm concentric bipolar electrode on the dura’s surface with 15 square pulses of 200 μs at a frequency of 200 Hz using a stimulator device (DS2, Digimiter). With the administration of voltage pulses, we looked for contralateral forepaw movement and corroborated this using a camera. This stimulation was performed on a grid every 100 μm in the AP and ML direction. The coordinates where the stimulation resulted in movement of the contralateral brachial biceps were taken to implant the recording electrode (mouse1: AP 0.5, ML 1.5; mouse 2, AP 0.0, ML 1.5).
### Temporally defined optogenetic striatal inhibition in vivo
Light was delivered via 300 micrometer-diameter implantable fibers (Doric lenses) coupled to a single longitudinal mode laser (MSL-FN-556, CNI lasers; 556 nm). For the optogenetic inhibition experiments, a free launching system controlled by an AOM (AAoptoelectronics) and fast speed shutter (Thorlabs catalog #SH5, SC10) triggered by TTL output from the MED-PC behavioral box was used to deliver the light. Power at the fiber tip was verified for every experiment using a power meter (Thorlabs catalog #PM130D). The power was adjusted to be 20 mW at the tip of the fiber for the green light. To define the time point for the optogenetic inhibition of the cortico-striatal projections before the initiation of sequences, we took advantage of the fact that animals developed stereotypical sequences. Thus, when the animal moved from the magazine to the first lever (left lever), the infrared beam was broken, sending the timestamp to trigger the light on and allowing the quantification of the latency to initiate the sequences of actions ( , ; Extended Data ). To define the time point for the light manipulations during the sequences’ execution, we used the timestamp of the first lever press in the sequences ( ). To define the time point for the light manipulations during the transition between forced subsequences, we used the timestamp of the penultimate lever press of subsequence 1. In self-paced sequences, we quantified the mean of lever presses to define a penultimate lever press (although we confirmed that it was two presses before the last of S1 in this case; see ). During the session of optogenetic inhibition, there were control (light off) trials and stimulation trials. The stimulation trials were randomly presented throughout the session (50% of total trials).
The inhibition of the premotor and primary motor cortico-striatal projections before sequence initiation impairs the initiation, but only the premotor disrupts the execution. A , top left, Diagram illustrates the injection site of Arch3.0-eYFP and the optrode (electrode array+ fiber optic) implantation into the LS. Top right, Plot depicting the activity of several units recorded in the striatum ( z score) aligned to the inhibition of the M2 cortico-striatal projections into the LS (green line above). Bottom left, representative perievent time histogram and raster plot of a striatal unit’s activity aligned to the inhibition of the M2 cortico-striatal projections (green shadow). Bottom right, Mean z score for the units that decreased their activity (green) or for all units recorded (black) during the inhibition. Inset, Pie chart showing the proportion of modulated and non-modulated units (comparing baseline time vs light). B , top panels, Sagittal diagrams and cortical photomicrographs of the injection of Arch3.0-eYFP into either M2 ( n = 9) or M1 ( n = 8) cortices and optical fibers implantation into the LS (Str). Scale bar: 500 μm. Bottom panels, Photomicrograph showing the average of Arch3.0-eYFP expression of the corresponding groups. Coronal diagrams representing the position of the optical fiber’s tips into the LS (green dots: Arch3.0-eYFP; gray dots: eYFP). C , Scheme of the inhibition protocol before the initiation of the sequences. The light inhibition (2 s) is triggered by breaking the infrared beam positioned outside of the magazine, when coming to the lever (dashed red lines on the schemes). D , Percentage of correct sequences [correct/(errors + correct)] throughout training. E , upper panels 1–4, Quantification per animal in off versus on trials of the proportion of each category of error. F , Effect of the inhibition on the number of lever press per sequence. G , Effect of the inhibition of the cortico-striatal projections on the latency to initiate self-paced sequences. H , Effect of the inhibition of the cortico-striatal projections on the proportions of times per block that animals returned to the magazine after having crossed the infrared sensor outside of the magazine (which set the timestamp to trigger the light inhibition; see ). On the paired plots, each line depicts the mean effect per animal during trials of optogenetic inhibition (on; green shadow) versus trials without optogenetic inhibition (off; no shadow) from the same session. In panels E–H , Δ on-off panels are obtained from the mean difference per animal in the on-off trials, adding the control group in gray. The mean difference in change score panels (mean d.) are obtained between the experimental groups and the control group (see Materials and Methods). Paired plots, two-sided permutation t test; panels Δ on-off, Arch animals versus control eYFP animals, unpaired two-sided permutation t test; * p < 0.05. The exact p values are described in the text and . Extended Data shows the effects of the inhibition of the lateral striatal neurons during the initiation and execution of the action sequences. Extended Data shows the M2/M1 retrogradely labeled cells from the LS, the thalamus, and the pons. Extended Data shows that the premotor cortico-striatal projections innervate the direct pathway with a stronger synaptic weight than the indirect pathway. n.s. = p > 0.05.
The inhibition of premotor cortico-striatal projections at the start of the execution decreased Long-S1 self-paced sequences. A , Scheme of the inhibition execution protocol. Light inhibition was triggered by the first press in the sequences. The green shadow depicts 2 s of continuous light inhibition. B , upper panels 1–4, Quantification per animal in off versus on trials of the proportion of each category of error. C , Effect of the inhibition on the number of lever presses per sequence. D , As in B , C evaluating the proportion of animals performing Long-S1 sequences S1 > 4→S2 >= 4). On the paired plots, each line depicts the mean effect per animal during trials of optogenetic inhibition (on; green shadow) versus trials without optogenetic inhibition (off; no shadow) from the same session. In panels B–D , Δ on-off panels are obtained from the mean difference per animal in the on-off trials, adding the control group in gray. The mean difference in change score panels (mean d.) are obtained between the experimental groups and the control group (see Materials and Methods). Paired plots, two-sided permutation t test; panels Δ on-off, Arch animals versus control eYFP animals, unpaired two-sided permutation t test; * p < 0.05. The exact p values are described in the text and .
Inhibition of premotor but not primary motor cortico-striatal projections at the moment of the transition increases the transition time inside self-paced sequences. A , Scheme of the inhibition protocol transition in self-paced sequences. The light inhibition was triggered by the press before the penultimate press of the subsequence 1. The dashed and black arrows denote the period of the transition. B , Estimation of the press that delivered the timestamp for light inhibition for M2→LS and M1→LS, light and dark blue, respectively. C , upper panels, Effect of the inhibition of the premotor or primary motor cortico-striatal projection on the transition time for self-paced sequences. D , Δ on-off for the inhibition before and execution protocols from the inhibition of M2→LS presented in , , respectively. E , No effect was detected by the inhibition of either M2→LS or M1→LS projection in the different categories of errors. Upper panels 1–4, Quantification per animal in off versus on trials of the proportion of each category of error. In C , E , each line in the paired plots depicts the mean effect per animal during trials of optogenetic inhibition (on; green shadow) versus trials without optogenetic inhibition (off; no shadow) from the same session. The panels Δ on-off are obtained from the mean difference per animal in the on-off trials, adding the control group in gray. The mean difference in change score panels (mean d.) are obtained between the experimental groups and the control group (see Materials and Methods). Paired plots, two-sided permutation t test; panels Δ on-off, Arch animals versus control eYFP animals, unpaired two-sided permutation t test; * p < 0.05. The exact p values are described in the text and .
### Behavioral quantification during optogenetic inhibitions
The percentage of correct sequences of actions was quantified. A correct sequence was defined as the sequence with at least four presses on the first lever followed by at least four lever presses on the second lever. We also calculated the proportion of incorrect sequences (errors; ), by dividing the number of incorrect sequences by the total number of stimulation trials (on trials) or control light trials (off trials). Latency to initiate was calculated as the time between crossing the infrared beam out of the magazine and the first lever press in the sequences. The duration of a sequence was the time from the first press of subsequence 1 to the last press of subsequence 2. The transition between sequences was the time from the last press in subsequence 1 to the first press in subsequence 2. All animals were video recorded during the optogenetic manipulations. This allowed us to verify that all animals used both forepaws to execute the presses.
### Ex vivo whole-cell recordings
To express ChR2 in the M2 cortico-striatal projections, we injected 300 nanoliters of ChR2 into M2 ( n = 4 mice). After 8–16 d, the animals were deeply anesthetized and transcardially perfused to obtain striatal slices as described in . Postsynaptic currents in whole-cell configuration were evoked by pulses of blue light (1 ms), with a pair of cells (one dMSN and one iMSN) recorded per slice. The recordings were acquired as described in .
### Experimental design and statistical analysis
Significance was determined by p < 0.05. For the proportions, paired or between groups, the χ test, the Wilcoxon test, or the Mann–Whitney U test was used, as appropriate. For comparisons along sessions, a non-parametric Friedman statistical test was used. When multiple comparisons were employed, a Benjamini–Hochberg correction was used to adjust the p value ( ; Extended Data , ). All statistical analyses were performed using GraphPad, R [R core Team (2013)] and MATLAB. Additionally, for the optogenetic manipulations experiments, we use estimation statistics based on confidence intervals (CIs) as described in and . The effect sizes and CIs are reported as: effect size [CI width lower bound; upper bound]. Five thousand bootstrap samples were taken, and the p value(s) reported are the likelihood(s) of observing the effect size(s), if the null hypothesis of zero difference is true. To account for multiple comparisons at each inhibition protocol, we considered a false discovery rate (FDR)-adjusted p value ( q value) < 0.10, two-sided, as significant ( ).
### Data and software availability
All data and MATLAB scripts will be available on request to the lead or corresponding authors.
## Results
To identify the premotor cortico-striatal contribution to serial order sequence execution, we developed a behavioral task in mice allowing us to perform two types of experiments. Mice were trained to execute two subsequences of lever presses on two levers in serial order ( ). For the first experiment, we aimed to identify whether neuronal activity in M2 or M1 is modulated during the task’s serial order sequences. We performed in vivo electrophysiological recordings in M2 or M1 to measure neuronal activity while animals executed these lever press sequences ( , ). For the second experiment, we aimed to identify whether the projections from these cortical regions contribute to the sequences’ execution, by optogenetically inhibiting the cortico-striatal projections either before or during the sequences’ execution, or during the transition between the two subsequences ( - ).
### Mice can learn to execute serial order sequences of lever presses
To investigate the contribution of cortico-striatal synapses to the structuring and execution of self-paced serial order sequences, we trained mice in a serial order task to execute two subsequences of lever presses ( ). The execution of a correct serial order sequence was achieved when animals performed at least four presses on a lever 1 (subsequence 1; S1) followed by four presses on lever 2 (subsequence 2; S2), which lead to the delivery of a reward in the magazine (pellet). To execute self-paced sequences, we first trained animals to execute forced sequences (forced: only one of the two levers was exposed at any point, signaling the animals when to press; ). Later in training, blocks of forced and self-paced serial order sequences were intercalated (self-paced: both levers remained exposed so that animals decided when to press and transition between subsequences; ). The requirement of intercalated blocks was necessary as animals’ performance dropped when they were required to perform only self-paced sequences (tested in a group of eight animals: data not shown). The main difference between a forced and a self-paced serial order sequence was that in the former, completing four presses on lever 1 retracted it and exposed lever 2; if four presses were executed on lever 2, it was also retracted, followed by the delivery of a pellet in the magazine ( ). On the other hand, the execution of a correct self-paced sequence also required at least four presses on each lever, but in this case, both levers remained exposed throughout the self-paced trials ( ).
A session containing both different kinds of sequences is identified in the figures as blocks of “forced–self-paced” sessions. During one of these sessions, the animals were required to achieve correct blocks of five forced and five self-paced sequences consecutively for 30 min. shows an example animal during the first (early) and the eleventh (late) day (rewards are labeled with filled red triangles for forced and in blue for self-paced). After 21 d of training, and 9 d performing blocks forced–self-paced sessions, animals showed a stable performance of correct sequences, demonstrated by the fact that performance no longer changed from sessions 9 to 11 [performance forced, χ (2) = 2.51, p = 0.30, performance self-paced, χ (2) = 1.0, p = 0.65, Friedman test, wild-type (WT) animals, n = 8, ]. No differences were found in comparisons of a number of task-relevant parameters between forced versus self-paced sequences: latency to initiate, the transitions intrasubsequences, the intervals between presses intra subsequences ( p > 0.05, Mann–Whitney U test, WT animals, n = 8; Extended Data ).
While animals executing forced sequences reached an 83 ± 1.5% in correct performance (correct sequences/errors + correct sequences), the execution of self-paced sequences only reached 61 ± 1.7% (WT animals, n = 8; , red and blue tick curves, respectively). To further investigate this performance, we quantified the types of errors and modes in which the animals executed correct sequences. An error was defined as a sequence of presses in which the animal did not achieve four presses on lever 1 followed by four presses on lever 2. The errors were grouped into four categories: (1) incorrect start (starting on the opposite lever); (2) breaks in S1; (3) premature switches from S1 to S2; and (4) breaks in S2 ( , right panels 1–4, respectively; note that 1 and three were only possible during self-paced). These four categories were quantified late in training ( ). The decrease in Breaks in S1 and S2 was the primary determinant in performance improvement of both forced (ForcBreaks) and self-paced (S-PBreaks) sequences compared with pretraining ( p < 0.05, Mann–Whitney U test, WT animals, n = 8; Extended Data , panels 2, 4). A decrease in incorrect-start errors accounted for the major improvement in the self-paced performance ( p = 0.007, Mann–Whitney U test, WT animals, n = 8; Extended Data , panel 1). No difference in performance between forced and self-paced sequences was detected when accounting for only the available errors in both forced and self-paced sequences: BreaksS1+BreaksS2 ( p = 0.5, Mann–Whitney U test, WT animals, n = 8; , blue dashed curve vs red curve).
The correct execution in forced sequences was S1 = 4→S2 = 4; in self-paced the execution of correct sequences grouped in several options (as anything with S1 = >4→S2 >= 4 was correct). The majority of correct sequences were Long-S1 sequences (S1 > 4→S2 >= 4; WT animals, n = 8; ).
Besides the execution structure of the sequences, we quantified the length parameters. The latency to start, the transition between subsequences, the interpress intervals, the duration, all became faster as training progressed (Friedman test, p < 0.05; WT animals, n = 8; Extended Data ). Later in training, the latency to initiate the sequences was not different between forced and self-paced (forced: 1.1 ± 0.1 s, self-paced: 1.5 ± 0.3 s, p = 0.535, Mann–Whitney U test, WT animals, n = 8; ). Neither the transition time between subsequences (S1 →S2 ) or the interpress intervals intersubsequences between the two modes ( p > 0.05, Mann–Whitney U test; ). However, the duration and the number of presses were longer in the S1 of self-paced than in forced [forced : 3.2 ± 0.2 s vs self-paced : 4.8 ± 0.4 s, forced = 4 ± 0 vs 6.1 ± 0.6 in self-paced , p = 0.0002, Mann–Whitney U test; Extended Data ).
This task parameterization showed that mice could learn and execute forced and self-paced serial order sequences in blocks. The main difference in the execution of these two modes of sequences (besides that self-paced have more possibilities for errors) is that animals execute longer chunks of presses in the subsequence 1 of self-paced sequences.
### Activity modulation in premotor and motor cortical neurons during the execution of serial order sequences
Once we established that mice are capable of executing serial order sequences, we questioned if, as predicted, the activity of M2 and M1 cortical neurons was modulated during the execution of these trained sequences ( ; ; ; ). For this purpose, we trained eight animals as in . After 3 d in the blocks of forced–self-paced sessions, we performed brain surgeries to implant a mobile electrode array either in M2 ( n = 6; , middle panel) or M1 ( n = 2; here, we verified the region to evoke forepaw movements by microstimulation; see Materials and Methods; , right panel). After 4 d of recovery, the training restarted until animals reached a stable performance in the execution of forced and self-paced sequences (10–15 blocks sessions postsurgery). From these animals, 34 well-isolated units in M2 and 26 units in M1 were analyzed. Representative examples of these neurons’ activity during the executions of sequences are presented in . The z score heatmaps of show all recorded units during the execution of self-paced sequences (the same units during the execution of forced sequences are presented in Extended Data ).
To answer whether the same units were active during the different phases of the sequences, we measured the mean activity of neurons above two z scores during the execution of the sequences. We classified neuronal activity into categories based on different parameters of sequence execution ( ). We found no statistical difference in the comparison of M2 versus M1 activity, only a tendency for more modulated/engaged M2 neurons related to the initiation of a sequence ( ). To explore this difference during sequence initiation, we compared the mean z score between M2 and M1 activity from the units that showed a significant modulation 1 s before starting the sequences ( ). This comparison showed a bigger z score positive modulation in M2 than in M1 during the execution of the sequences (M2 = 14 of 34 units, M1 = 8 of 26 units; mean z score in self-paced : M2 = 1.32, M2 = 0.85 vs M1 = 0.55, M1 = 0.22; p < 0.05, z score difference test; , bars labeled “S1 and S2”). Remarkably M2 units showed a stronger modulation in self-paced sequences, even before the initiation of the sequence (M2 = 1.4 vs M1 = 0.8, p = 0.04 z score difference test; , bars labeled “Init” in blue) and a similar tendency during the initiation of forced sequences (Extended Data , bars labeled “Init” in red). No difference in the number of recruited units throughout the execution of the sequences was detected ( p > 0.05, χ test; Extended Data ).
### Both M2 and M1 contain units encoding the length of the serial order sequences of actions
After the evaluation of the mean z score related to the execution of the serial order sequences, we tested the hypothesis that M2/M1 encode the length parameters of the execution of these sequences ( ; ; ; ; ; ; ; ; ; ; ). To evaluate this possibility, we performed linear regressions between the duration of the segments of the sequences (total sequence, S1, transition, or S2) and the neural activity recorded from individual neurons in either M2 or M1. We reasoned that if these cortices contained activity encoding the sequence length, their cortico-striatal projections could convey this information to the striatum.
shows two examples of units that presented significant regressions ( R > 0.6 and p < 0.05) in specific time bins during the execution of serial order self-paced sequences ( , middle row panels, gray shadows). shows the proportion of units that presented with significant regressions based on five questions: column i, does the activity before the first press (1 s) encode the sequences’ length? ii, does the mean activity during the sequence encode the length of the sequence? iii, iv, and v, Is there a relationship between the duration of S1, the transition or the duration of S2, and the mean firing rate during these epochs? The answer to these five questions is presented per recorded unit in either M2 or M1 ( , upper panels) and as the % of units with significant regressions per region ( , bottom panels). Notably, the biggest proportion of units with significative regressions was between the sequence duration and the firing rate before the initiation of the sequence in both M2 and M1 ( , columns i) or the sequences duration and the mean firing rate of M2 during the sequences ( , columns ii).
The percentage of recruited units over time (units presenting significant regressions, 200-ms bin, sliding 10 ms; R > 0.6 and p < 0.05) is presented in Extended Data . Despite small tendencies, no difference was detected in the proportion of units recruited overtime either when comparing M2 versus M1 nor when comparing within each region between forced and self-paced sequences (1 s before or one after the start of the sequences). Similarly, no differences were found between neuronal activity and the duration of the sequence, duration of S1, duration of the transition, or the number of presses (comparison of bars p > 0.05, χ test; Extended Data ).
To address whether specific cortico-striatal M2 or M1 neurons encode for sequence parameters, we recorded from antidromic photo-identified cortico-striatal neurons in vivo ( ; ). To achieve this, we first confirmed the lateral region of the striatum (LS) that is innervated by both M2 and M1 projections (Extended Data ; ; ; ; ). Then we trained mice in which the expression of Channelrhodopsin-2 (ChR2) was targeted to the excitatory cortical neurons in M2 ( n = 4) or M1 ( n = 2) and their projections to the striatum (using Emx1-Cre mice; ; ). We then implanted an electrode array above the ChR2 expression and an optical fiber into LS to antidromically activate the cortico-striatal units (Extended Data ; ). Following this procedure and after a stable performance in the execution of blocks of forced–self-paced sequences, the activity of M2 or M1 units was recorded. During these recordings, we recorded cortical and cortico-striatal photo-identified units (PID units). A PID unit was that, at the end of the session, responded to antidromic light stimulation with short latency (<10 ms; ). The mean latency of M2+M1→LS was 5.1 ± 0.4 ms; n = 21; Extended Data , bottom right panel). Following these criteria, we identified 10 M2 and 11 M1 cortico-striatal PID units (Extended Data ). As in the non-PID units ( ), the major proportion of PID units presented significant regressions between the sequences duration and the firing rate before the initiation, the duration, and the mean firing rate during the execution of the sequences (Extended Data , columns i, ii).
In summary, in , and Extended Data , we show that the activity of M2 and M1 is related to the execution of the trained sequences. The mean z score analysis showed that M2 had a bigger modulation than M1 during the execution and even before the initiation of the serial order sequences. Furthermore, the regression analysis showed that both M2 and M1 contain units that encode the sequences’ execution length.
### Inhibition of lateral striatal neurons before initiation of a serial order sequences impairs its execution
This study’s main goal was to establish whether cortico-striatal projections of the premotor and motor cortex contribute to the execution of sequences. To address this point, we first verified that neuronal activity in the lateral striatal (LS) region that receives inputs from M2 and M1 (Extended Data ) contributes to the execution of serial order sequences. For this purpose, we trained a group of animals in which the inhibitory opsin archaerhodopsin (Arch3.0) was injected bilaterally into LS followed by fiber-optic implantation above the injection sites (Extended Data ). Then, once the animals reached around 80% success (correct sequences) in the performance of forced and 60% in self-paced sequences (Extended Data , panel 3), we performed sessions of optogenetic inhibition, delivering the light to inhibit Arch3.0-expressing neurons in a state-dependent manner during the execution of the sequences. We randomized three protocols (each protocol on a different day): (1) before the initiation; (2) during the execution; or (3) during the transition between subsequences (Extended Data , respectively). We observed that optogenetic inhibition before starting the sequence delayed the initiation (Extended Data ). Inhibition during the execution of the sequence increased premature switches (Extended Data , panel 3) and decreased the correct Long-S1 self-paced sequences (Extended Data ). Inhibition during the intersubsequence transitions increased the transition time between forced subsequences and showed a tendency to increase the time between self-paced subsequences (Extended Data ). Note that the data for the forced sequences are only presented in the figures when significant effects were detected; otherwise, they are presented only in .
### Presequence initiation inhibition of either motor or premotor cortico-striatal projections impairs initiation, but only the latter disrupts the execution
Once we verified that direct inhibition of LS neurons impaired the execution of sequences, we asked whether the premotor (M2→LS) or the motor (M1→LS) cortico-striatal projections contributed to the execution of these sequences. We first verified the inhibition of striatal neurons by the optogenetic inhibition of cortico-striatal projections using Arch3.0 with 2-s pulses of light ( ). This pulse length is in a proper range far from the biophysical constraints of using Arch for optogenetic inhibition ( ; ). Furthermore, we evaluated the proportion of labeled cells in M2/M1 that contain corticofugal axons crossing by the area of the LS manipulations (Extended Data ).
Next, we tested whether the contribution of cortico-striatal projections is time dependent ( ; ; ; ). We performed optogenetic inhibitions of the M2→LS ( n = 9) or M1→LS ( n = 8) projections in a state-dependent manner, using three protocols: (1) before initiation ( ); (2) during execution ( ); and (3) during the transition between subsequences ( ).
Most of the effects of optogenetic inhibition were on self-paced sequences. Therefore, for - , forced sequences are only presented when effects were detected; otherwise, only self-paced sequences are presented. The full data for forced and self-paced inhibitions is in the .
To carry out these experiments, a group of Emx1-Cre mice was subject to bilateral Arch3.0-eYFP expression into M2 or M1 (or eYFP for control animals) and bilateral fiber optic implantation into the LS ( ; see - ). After surgery, animals were allowed to recover for 3 d before training started and continued until the performance was stable ( ). During the optogenetic inhibition sessions, and depending on the protocol, a continuous pulse of green light was applied randomly in 50% of the trials (2 s; triggered by the behavior), allowing us to compare the effects of light inhibition on each animal within the same session.
For the inhibition “before” initiation protocol, we aimed to assess whether the cortico-striatal projections contribute to sequence execution by interfering with the initiation/preparation of the sequences for which we took advantage of the stereotyped behavior of the trained animals. An infrared beam was placed between the magazine and the levers (red arrow-dashed line coming out of the magazine; ), it was possible to trigger a pulse of light when the animal crossed the infrared beam before starting the sequences ( ).
The inhibition of the M2→LS projections before initiation increased premature switches (M2→LS on = 20 ± 5 vs M2→LS off = 8 ± 1; the paired mean difference was 12.2 [95.0%CI 6.8, 27] and p = 0.0001, q = 0.01 two-sided permutation t test; , panel 3, light blue data; ). This effect was not observed in control animals (Δ on-off comparison to the control group: the unpaired mean difference between control and M2→LS was 10.4 [95.0%CI 3.7, 2.4]. The p value of the two-sided permutation t test was 0.0224, q = 0.06; , panel 3, bottom part). The increased premature switches were accompanied by decreased presses within the self-paced sequence (M2→LS on = 9 ± 0.6 vs M2→LS off = 10 ± 0.4; the paired mean difference was −1.16 [95.0%CI −2.42, −0.495], p = 0.016, q = 0.05 two-sided permutation t test; , light blue data).
This inhibition protocol also increased the latency to start the sequences (M2→LS on = 2.9 ± 0.4 vs M2→LS off = 1.8 ± 0.3; the paired mean difference was 1.1 [95.0%CI 0.556, 1.96], p = 0.0026, q = 0.01 two-sided permutation t test; , light blue left panel; forced sequences were also delayed, see , data related to ), an effect not observed in control animals ( , Δ on-off panel, gray data). This increased latency to start was accompanied by an increase in the animals’ return to the magazine in the trials that received light inhibition, suggesting that inhibition interrupted the proper serial order sequence initiation (see ; M2→LS on = 9 ± 2% vs M2→LS off = 1 ± 0.7%, the paired mean difference between control and M2→LS was 8.29 [95.0%CI 5.14, 11.7], p = 0.006, q = 0.03 two-sided permutation t test. Unpaired mean difference between control and M2→LS was 9.86 [95.0%CI 6.0, 13.8], p = 0.0002, q = 0.002 two-sided permutation t test; , Δ on-off panel).
Conversely to the inhibition of M2→LS projections, the increase in premature switches and the decrease in the number of presses within the sequence were not observed when we inhibited M1→LS projections ( , dark blue data). Instead the inhibition of the M1→LS projections increased the latency to start the self-paced sequences (M1→LS on = 3.2 ± 0.4 vs M1→LS off = 2.6 ± 0.3; the paired mean difference was 0.614 [95.0%CI 0.40, 1.1] and p = 0.009, q = 0.04 of the two-sided permutation t test; , comparison to control: panel Δ on-off same figure, dark blue vs gray data, unpaired mean difference between control and M1→LS was 0.88 [95.0%CI 0.4, 1.7] and p = 0.003, q = 0.01 of the two-sided permutation t test), and the returns to start (comparison to control: panel Δ on-off in the same figure, dark blue vs gray data; unpaired mean difference between control and M1→LS was 11.8 [95.0%CI 3.3, 26.1] and p = 0.02, q = 0.06 of the two-sided permutation t test; ).
Together these results suggest that the M2→LS but not the M1→LS projections contribute to the execution/structuring of self-paced sequences while both M2→LS and M1→LS contribute to the initiation.
### Inhibition of premotor cortico-striatal projections at the start of the execution decreased Long-S1 sequences
One prediction for the cortico-striatal projections’ contribution to serial order sequences execution is that their requirement would be time dependent (see , ; ; ). To further prove this idea, we performed a second protocol: inhibition during “execution,” which consisted of performing light inhibition of the cortico-striatal projections once the execution of the sequences started (triggered by the first press in S1; ). Using this protocol, we observed that the inhibition of the M2→LS projections did not modify the proportion of the different categories of errors ( ) nor the number of presses in the sequences ( ), but it did decrease the proportion of Long-S1 sequences (M2→LS on = 35 ± 5 vs M2→LS off = 50 ± 6, the paired mean difference was −14.6 [95.0%CI −18.9, −9.8] and p = 0.002, q = 0.04 of the two-sided permutation t test; , upper panels; comparison to control eYFP-group: Δ on-off panel, unpaired mean difference between control and M2→LS was −18.4 [95.0%CI −28.3, −8.8] p = 0.003, q = 0.04 of the two-sided permutation t test , bottom panel), with no overall effect on the duration or the number of presses in the sequences (see ; data related to ). The inhibition of the M1→LS projections did not yield any significant effects with the inhibition during execution protocol ( , dark blue data).
The results from the inhibition of M2→LS or M1→LS projections during the beginning of the sequence execution highlight that the activity of the M2→LS projections is important for the length of the sequences (particularly for S1). Furthermore, these results also suggest that, at least later in training, the M1→LS projections are not required for sequence execution.
### Inhibition of premotor cortico-striatal projections during the transition between serial order subsequences increases the transition time
So far, we have shown that the cortico-striatal projections from M2→LS contribute to the appropriate initiation and execution of self-paced serial order sequences ( , ). However, in the previous experiments, the light inhibition never occurred during the transition between subsequences. Given that a small proportion of units from either M2 or M1 showed significant modulation during the transition-moment ( ), we asked whether inhibiting directly during the transition could reveal whether the M2→LS or M1→LS contribute to the transition. For this purpose, we set up an inhibition during “transition” protocol, in which the press before the penultimate press of S1 triggered light inhibition for 2 s ( , for self-paced; in forced sequences, it was the third press; see , data related to ).
shows the mean press in which the animals of the M2→LS or the M1→LS groups received the inhibitory stimulation (or the corresponding timestamp in the “off” trials); on average, it was the press −2 counting from the final press in the S1 (M2→LS on = −2.0 ± 0.1 vs M2→LS off = −1.9 ± 0.1, p = 0.25, Mann–Whitney U test; ). With this protocol, we observed that the inhibition of M2→LS, but not M1→LS, showed a tendency to increase the duration of the transition (M2→LS on = 1.0 ± 0.06 vs M2→LS off = 0.8 ± 0.06, the paired mean difference was 0.08 [95.0%CI 0.02, 0.16] and p = 0.050, q = 0.4 of the two-sided permutation t test; , upper panels), that reached significance when compared with the eYFP-control group: unpaired mean difference between control and M2→LS was 0.18 [95.0%CI −0.09, 0.3] and p = 0.003, q = 0.03 of the two-sided permutation t test; , Δ on-off bottom panel). This effect was time-specific since neither the inhibition before nor the inhibition execution protocol on the M2→LS yielded a similar effect ( ). This effect on the transition raised the question of whether the M2→LS projections treat each subsequence as independent chunks or concatenate them into a single chunk once the whole sequence has been acquired ( ; ). We hypothesized that if the transition was affected by the light inhibition, the proportion of breaks in the sequences might increase as well. Contrary to this hypothesis, no modifications on the breakings of sequences were detected by this protocol ( ). We also observed no differences in the number of presses or sequence length ( ), supporting the idea that the M2→LS projections contribute to the transition intersubsequences without affecting the second part of the serial order sequence.
Altogether, the results from the recording and inhibition experiments reveal that the cortico-striatal projections to the LS from the premotor and motor cortices have specific contributions to the execution of sequences. Both projections contribute to the initiation, while M2→LS also contributes to the correct execution and transition between subsequences of self-paced sequences. Importantly the M2→LS contribution was essential during the beginning of the execution, suggesting its contribution is relevant before the initiation, at the beginning of execution, and decreasing in relevance as the sequence execution progressed.
### Premotor cortico-striatal projections innervate the direct pathway with a stronger synaptic weight than the indirect pathway
Finally, we investigated whether the M2→LS projections differentially impact the two subcircuits of striatal projection neurons, the direct or indirect striatal pathways ( ), by looking at synaptic connectivity. For these experiments, we crossbred Emx1-Cre (allowing us to express ChR2 in M2) with BAC D2-GFP mice. By injecting red retrobeads into the substantia nigra, we could identify in the striatum neurons from direct pathway (red retrobeads; dMSN), or the indirect pathway (GFP; iMSN). This allowed us to record from identified striatal neurons in brain slices ex vivo while stimulating the M2→LS ChR2 axons (Extended Data ). Using this procedure, we recorded pairs of neurons in the same brain slice in the LS (one iMSN and one dMSN; counterbalancing the recording order per slice). We evoked postsynaptic responses by activating the M2→LS ChR2 projections with brief pulses of light (1 ms; Extended Data ). From these experiments, we recorded a total of 6 pairs of striatal neurons (in 6 different slices from four animals). Postsynaptic evoked responses had a bigger amplitude in the dMSN than in the iMSN (dMSN: 290 ± 100 vs iMSN = 30 ± 9 pA, Mann–Whitney U test, p = 0.031; Extended Data ), with no difference in their latencies (dMSN: 3.3 ± 0.1 ms vs iMSN = 3.8 ± 0.1 ms, Mann–Whitney U test, p > 0.05; Extended Data ). These evoked postsynaptic responses were glutamatergic, being blocked by the AMPA receptor antagonist CNQX (Extended Data ). This finding suggests that the M2→LS projections may have stronger synaptic connections onto the dMSNs than onto the iMSNs in vivo when the animals are executing the serial order action sequences.
## Discussion
These results show that mice can execute serial order forced and self-paced sequences of lever presses ( ). Both the premotor (M2) and motor (M1) cortices are modulated during sequence execution ( ). M2 showed a bigger modulation during the initiation and execution in the units modulated before the start of the sequences ( ). Both M2 and M1 units showed regressions between their activity and the sequence duration before or during the execution ( ), even in photo-identified M2 and M1 cortico-striatal cells (Extended Data ). The M2→LS and M1→LS projections contribute to the proper initiation, but mainly M2→LS projections contribute to the structuring/execution of serial order self-paced sequences ( - ). The M2→LS synapses present bigger amplitude postsynaptic responses onto the direct versus the indirect pathway of striatal neurons (Extended Data ).
Mice had been previously been shown to develop and execute serial order operant tasks, including pressing two levers ( ; ). The task adapted here allows temporally separated manipulations within the different phases of the two subsequences (initiation, execution, and transition between subsequences). We first trained animals in forced sequences and later in blocks of forced–self-paced sequences as animals trained only in the latter dropped their behavior. The fact that animals were more efficient at executing forced than self-paced can be explained by the greater number of possible errors in self-paced sequences. However, a difference in the recruitment of brain structures in forced versus self-paced sequences is not excluded ( ; ).
To identify whether the cortico-striatal projections from M2/M1 contribute to sequence execution, we investigated whether neuronal activity in these structures was modulated during the execution of the sequences. We identified that M2 and M1 units showed activity modulations milliseconds before the start and during the execution of these serial order sequences ( , ), with mostly exclusive firing categories ( ). Importantly, the units from M2 that were modulated before the sequences started showed a bigger positive modulation during the execution and before starting the sequences than those from M1 ( ). This increased activity time before the execution of movements is attributed to time estimation or preparation to execute an action [e.g., the anticipatory activity (ramping) in M2 has a direct relationship with the time that animals wait to start an action or the duration of a stimulus ( ; )]. In this study, the animals had to approach the lever before the first press in the sequence. We could not therefore rule out that the activity before starting the sequences may be related to motor preparation, although the inhibition of this activity before pressing delayed the latency of both forced and self-paced sequences, but only impaired the execution of self-paced sequences. Together, these results show that units in M2/M1 are modulated during the execution of sequences. When evaluating whether these units’ activity may encode the execution parameters, we observed no significant differences in the proportions of units in M2 versus M1, as measured by regression analysis between neuronal activity and temporal parameters of the sequences ( ). These results show that both M2 and M1 contain units encoding the sequences’ execution ( ; ). However, it must be acknowledged that the lack of differences in some of the comparisons between M2 and M1 may be influenced by the small set of recorded cells (34 cells from M2 and 26 from M1).
To evaluate how the M2/M1 cortico-striatal projections contribute to the execution of serial order sequences, we performed two experiments. First, we recorded from cortico-striatal antidromic PID units (Extended Data ). Second, we performed state-dependent temporal inhibitions of the M2→LS or the M1→LS projections either before or during the execution of sequences ( - ). Although the photo-identification of cortico-striatal M2→LS or M1→LS was low (10 per structure), it allowed us to detect units in each of these cortices that presented significant regressions with temporal parameters of the execution of the sequences. However, there were too few photo-identified neurons to infer whether M2 or M1 has a more important contribution to sequences execution. The contribution of these cortico-striatal projections was thus mainly evaluated with state-dependent optogenetic inhibition.
To test the hypothesis that the M2/M1 cortico-striatal projections have time-specific contributions to sequences execution, we performed three protocols of inhibition while animals performed lever pressing sequences. Our inhibitions were all 2 s, below the safe limit of the biophysical constraints of using Arch for optogenetic inhibitions ( ).
Interestingly, the inhibition of either the M2→LS or M1→LS, before the initiation increased the latency to start the sequences ( ). This increase in latency was accompanied by an increase in the animals’ return to the magazine in the initiations that received light inhibition, as if the inhibition interrupted the proper initiation ( ; see ). These results are consistent with a model where the M2→LS and M1→LS cortico-striatal projections are required for the proper initiation of sequences, with the former likely setting up the parameters for the upcoming sequence (for further discussion, see below).
Remarkably, the three inhibition protocols revealed that the projections from M2→LS, but not the M1→LS, are temporally required for the execution of self-paced sequences. The inhibition of M2→LS before the start altered the sequence structure, increasing premature switches ( ) and decreasing the number of actions in the sequence ( ). Their inhibition during the beginning of the execution affected the first segment of the sequence (by decreasing the probability of executing Long-S1 sequences; ). Their inhibition during the transition increased the transition time between subsequences, although this last had the slightest effect ( ).
A previous experiment, which decreased the M2→striatal projections (by manipulating M2→striatal projections plus their collaterals to other brain regions) during the execution of a serial order short sequence reported that inhibition of M2→dorsolateral striatal cells, retrogradely labeled from the striatum, impaired the first step accuracy ( ), perhaps by increasing incorrect starts or promoting premature switches from S1 to S2. Here, the possibility of longer subsequences in the sequence allowed for measuring incorrect starts, breaks, premature switches, and the transition time between subsequences. Our inhibitions of M2→LS was spatially and temporally specific (if the optogenetic inhibition does not backpropagate as it has been documented; ), showed no effect on incorrect starts. Rather we observed an increase in premature switches when inhibiting before the start, a decrease in Long-S1 sequences when inhibiting during the execution, and slowness of the transition time when inhibiting around the transition.
Another important point reported here is the specificity of the effects mainly on self-paced sequences (see ). How is it that the premotor projections can differentiate between the execution of forced and self-paced sequences? A possible explanation is that the M2→LS projections could detect whether the animals were in blocks of self-paced sequences, engaging these projections with a bigger sensitivity to the inhibitions as a consequence of self-deciding the execution. This possibility is supported by previous studies showing that when animals engage in self-paced behaviors, as opposed to forced behaviors, some cortical subcircuits become engaged and more important ( ; ; although see ; ). An idea supported by the tendency of cortico-striatal M2 PID units to show a bigger proportion of units with significant regression in the self-paced sequences (Extended Data , columns i, iii). Importantly we observed that the inhibition of the M1→LS did not affect the execution of sequences, consistent with the idea that M1→LS activity is not required to execute sequences once these have been learned ( ).
Finally, another remarkable finding from the optogenetic inhibitions experiments is that the M2→LS projections may contribute as a serial driver of sequence execution ( ; ; ; ; ). Supporting this idea, their inhibition before the starting could restart the initiation, increasing premature switches during the execution, and decreasing the number of actions ( ). Conversely, their inhibition during the start of the execution decreased the probability of executing Long-S1 sequences only ( ).
In addition to the inhibition of cortico-striatal projections, we verified the LS requirement for the sequence execution and addressed the possibility that our cortico-striatal inhibitions may affect other brain areas besides the LS by performing two experiments: (1) we directly inhibited the lateral striatal neurons with the same protocols that inhibited the cortico-striatal projections (Extended Data ); and (2) we evaluated whether the manipulated region contains axons en passant from M2/M1 that reach other brain areas (e.g., thalamus/pons; Th/Pns; Extended Data ).
The first experiment’s results found that direct inhibition of LS during sequence execution decreased the length of correct sequences (Long-S1 sequence) and the number of presses (Extended Data ). The inhibition of the LS in the transition protocol increased the transition time between subsequences (Extended Data ). These effects were in line with the inhibition of M2→LS projections. However, the LS inhibition before the sequence start did not change the errors in self-paced sequences, contrary to the inhibition of the M2→LS projections which increased premature switches with this protocol. A possible explanation for this last difference is that M2→LS activity may be required to set the execution of sequences by impinging on specific striatal subcircuits ( ), However, our direct inhibition of the LS did not differentiate between striatal subcircuits. To address this, Extended Data shows that the M2→LS innervates, strongly the striatal projection neurons from the direct pathway. Such preferential innervation has been previously suggested ( ; ) to direct the structuring of sequences in serial order tasks ( ), although only after learning.
Lastly, the quantification of the cells that project from M2/M1 to other brain areas by crossing through the LS showed it to be a small percentage of neurons (M2/M1→Str→Th 3%; M2/M1→Str→Pns 3%; Extended Data ). This data, plus the evidence that Arch3.0 mainly has inhibitory actions through local inhibition rather than affecting the propagation of action potentials ( ), suggests that the effects observed here were predominantly from optogenetic inhibition of the cortico-striatal projections from M2 or M1 into the LS.
In conclusion, this study shows that the premotor cortico-striatal projections to the LS contribute to the initiation and execution of self-paced sequences. It supports a model in which both M2 and M1 contain activity modulated by sequence initiation and execution. However, the M2→LS cortico-striatal projections mainly contribute to the proper execution/structuring of self-paced sequences. Both, M2 and M1 cortico-striatal projections contribute to the initiation of sequences. Also, our findings support the idea that the premotor cortico-striatal projections to the LS are a serial driver for the execution of sequences ( ; ; ; ; ). Altogether the presented findings may have important implications for pathophysiological conditions whereby self-paced generation of actions is impaired.
|
The neurohypophysis (NH), located at the posterior lobe of the pituitary, is a major neuroendocrine tissue, which mediates osmotic balance, blood pressure, reproduction, and lactation by means of releasing the neurohormones oxytocin (OXT) and arginine-vasopressin (AVP) from the brain into the peripheral blood circulation. The major cellular components of the NH are hypothalamic axonal termini, fenestrated endothelia and pituicytes, the resident astroglia. However, despite the physiological importance of the NH, the exact molecular signature defining neurohypophyseal cell types and in particular the pituicytes, remains unclear. Using single-cell RNA sequencing (scRNA-Seq), we captured seven distinct cell types in the NH and intermediate lobe (IL) of adult male mouse. We revealed novel pituicyte markers showing higher specificity than previously reported. Bioinformatics analysis demonstrated that pituicyte is an astrocytic cell type whose transcriptome resembles that of tanycyte. Single molecule in situ hybridization revealed spatial organization of the major cell types implying intercellular communications. We present a comprehensive molecular and cellular characterization of neurohypophyseal cell types serving as a valuable resource for further functional research.
## Significance Statement
The neurohypophysis (NH) is a major neuroendocrine interface, which allows the brain to regulate the function of peripheral organs in response to specific physiologic demands. Despite its importance, a comprehensive molecular description of cell identities in the NH is still lacking. Utilizing single-cell RNA sequencing (scRNA-Seq) technology, we identified the transcriptomes of five major neurohypophyseal cell types in the adult male mice and mapped the spatial distribution of selected cell types in situ . We revealed an unexpected cellular heterogeneity of the NH and provide novel molecular markers for neurohypophyseal cell types with higher specificity than previously reported.
## Introduction
The pituitary, also dubbed the hypophysis, is the master endocrine gland that is localized at the base of the hypothalamus in all vertebrate species. It is composed of the adenohypophysis (AH) and the neurohypophysis (NH), also known as the anterior and posterior pituitary, respectively. The mammalian pituitary consists of an additional anatomically discernable tissue, the intermediate lobe (IL), which is located between the NH and AH. However, the IL is not as distinguished in the pituitary of human and some non-mammalian vertebrates, including zebrafish ( ; ; ; ). The hypothalamo-neurohypophyseal system (HNS) encompasses hypothalamic magnocellular neurons residing in the paraventricular nucleus (PVN) and supraoptic nucleus (SON) and project their axons into the NH. Thus, two neuropeptides, oxytocin (OXT) and arginine-vasopressin (AVP), are produced in magnocellular neurons, transported along neurohypophyseal-projecting axons and released into the general blood circulation through the neurohypophyseal capillary plexus ( ). Circulating OXT and AVP neurohormones affect the physiologic function of peripheral organs such as the kidney, mammary gland and the uterus. Specifically, AVP regulates osmotic balance and blood pressure ( ; ; ), while OXT is mainly known due to its effects on reproduction organs ( ).
Unlike the AH, which serves as a hormone-secreting gland, the NH is a neural tissue, which serves as a neuroendocrine interface between AVP and OXT axonal projections and the permeable capillary network of fenestrated endothelia ( ). This neurovascular interface also contains the pituicytes, specialized neurohypophyseal astroglia, which occupy ∼50% of the neurohypophyseal total volume ( ; ). Pituicytes engulf HNS axonal swellings and their terminal buttons and are in close contact with the basal laminar and vascular endothelia ( ; ). Based on their dynamic morphologic plasticity during lactation and in response to chronic dehydration, it has been suggested that the pituicytes mediate neurohormones passage through the fenestrated capillaries serving as a physical gateway between the axons and the perivascular space ( ; ). Recently, we reported that during development, pituicyte-derived factors regulate the decision of zebrafish NH vasculature to adopt a permeable endothelial fate instead of forming a BBB ( ). The early definition of pituicytes was based on histochemical staining with silver carbonate and hematoxylin and eosin ( ; ; ). Thus, different subtypes of pituicytes have been defined by their fibrous, ependymal (with cilia or microvilli), oncocytic morphologies or by ultrastructure of organelle contents, such as dark and pale pituicytes due to high/low density contents of cytoplasmic matrix and organelles and granular pituicytes containing numerous cytosegregosome type dense bodies ( ; ; ). However, there is very little knowledge of pituicyte-specific genes. Consequently, mammalian pituicytes have been so far labeled with astroglial markers, such as apolipoprotein E (APOE), GFAP, S100β, vimentin (VIM), and connexin43 (Cx43/GJA1), all of which are general astrocytic markers, which are also expressed in other cell types ( ; ; ; ; ). Moreover, defining and visualizing pituicytes by co-expression of the above genes is not informative as these markers only partially overlap ( ). Hence, the exact definition of pituicyte cell type and/or subtype remains ambiguous. Finally, other neurohypophyseal cell types might not have been detected in published bulk neurohypophyseal transcriptomic data ( ).
The recent technological revolution enables high-resolution studies for transcriptome patterns in heterogeneous cell populations. Single-cell RNA sequencing (scRNA-Seq) allows dissecting cell types that are previously hidden due to identical histology, same genetic marker and adjacent location within a complex tissue ( ). This technology enables hundreds and thousands of single cells being processed at once, therefore delivers high-throughput, and highly efficient analysis of cell heterogeneity. In this study, we used scRNA-Seq to unravel the cell heterogeneity of the NH. Seven major cell types in the NH and IL of adult male mouse were identified. We present a comprehensive view of the molecular landscape as well as spatial organization of NH and IL cell types, hence providing valuable resources for studying their specific cellular and physiologic functions.
## Materials and Methods
### Experimental design
Three-month-old male C57/BL6 and Cx3cr1 -GFP mice ( ) were used in this study. All experimental procedures were approved by the Weizmann Institute’s Institutional Animal Care and Use Committee (IACUC).
### Single-cell dissociation
Two independent groups of five C57/BL6 mice were sacrificed by decapitation and the NH were dissected and collected into ice-cold 1 ml of magnesium-free and calcium-free HBS-/- buffer (20 mM HEPES-buffered saline, 145 mM NaCl, 5.4 mM KCl, and 20 mM glucose, pH 7.2) ( ). NH tissues were then transferred to ice-cold PBS containing magnesium and calcium (HyClone, GE Healthcare), treated with 50 ng/µl Liberase TM (Roche) for 12 min at 37°C, and further dissociated by incubating in HBS-/- buffer containing 0.15 mg/ml Papain (Sigma) and 10 U/ml DNase I (Invitrogen) for 8 min at 37°C. The reaction was stopped by adding heat-inactivated fetal bovine serum (HI-FBS; HyClone) to reach final concentration of 5%. To obtain single-cell resuspension, the loosened tissues were collected and passed through a 40-µm nylon mesh in 800-µl resuspension buffer [Leibovitz L-15 with 0.3 mM glutamine (Gibco, Thermal Fisher), 0.5% of penicillin streptomycin solution (Gibco, Thermo Fisher), 1% HI-FBS, 0.04% BSA]. Cell number, survival rate, clarity, and singularity were checked by Trypan Blue staining followed by hemocytometer counting.
### scRNA-Seq
scRNA-Seq was performed with 10x Genomics Chromium Single Cell kit version 2. Two independent samples, each containing 600–800 cells/ml, which had ∼70% survival rate and very few debris were used to form droplets containing single cell and barcoded-beads. The targeted recovery was 4000 cells per sample. The subsequent cDNA synthesis and library preparations were conducted according to the manufacturer’s protocol (10x Genomics). Two libraries were then indexed and pooled for sequencing using a NextSeq 500 High Output v2 kit (75 cycles; Illumina) according to the manufacturer’s instructions. Four lanes were used with R1 26 cycles and R2 58 cycles.
### Data and software availability
The accession number for the NH single-cell transcriptome reported in this paper is Gene Expression Omnibus (GEO; ): GSE135704.
### Statistical analyses
Sequences data were demultiplexed using Illumina bcl2fastq. Each of the samples was analyzed by Cellranger (version 2.0.0), run with the option –force-cells = 1500 and using the 10X prebuilt mm10 reference database version 1.2.0. The outputs from CellRanger were further analyzed using the Seurat package V2.3 ( ) and R 3.5. Using Seurat, we performed gene filtering (gene must appear in three cells of a sample) and merging of the cells of both samples to one set. Cell filtering was based on the number of genes per cell (must be between 400 and 5000), the number of UMI counts per cell (between 1000 and 10,000), and the percentage of mitochondria genes lower than 0.25 percentage. Eleven clusters were created with 900 variable genes and 11 principal components (PCs). The cluster names were replaced with the cell type identity based on the differentially expressed genes (marker genes).
### Gene set enrichment analysis
To determine whether known biological functions or gene sets are overrepresented (enriched) in an experimentally-derived gene list, an overrepresentation analysis (ORA) ( ) was employed. The gene set associated with a cell type, which was downloaded from PanglaoDB database ( ) were compared to the differentially expressed pituicyte markers filtered with criteria of average_logFC ≥ 1 and p adj ≤ 0.05. To test for overrepresentation of successes in the sample, the hypergeometric p value was calculated using R function phyper with lower tail= false as the probability of randomly drawing k or more successes from the population in n total draws ( ). The FDR was achieved by adjusting the p value using Benjamini and Hochberg ( ). To further illustrate the above finding specific differentially expressed pituicyte markers were compared with filtering criteria of average_logFC ≥ 1 and p adj ≤ 0.05 to published scRNA-Seq gene lists of astrocytes, and tanycytes i.e., PanglaoDB and other studies ( ; ; ; ; ).
### Wholemount in situ hybridization (WISH) and immunostaining
Three-month-old C57BL6 mice were perfused and fixed by 2% PFA for 10 min and fixed in 4% PFA on ice for 20 min in the dark. WISH was performed as described in ( ; ) with prolonged proteinase K treatment of 45 min. Tissues were postfixed in 4% PFA for 20 min at room temperature and washed 3 × 15 min PBS-Tx (Triton X-100; 0.3%). Subsequent immunostaining of WISH samples was performed following re-blocking in blocking buffer (10% lamb serum, 0.3% Triton X-100, 1% DMSO in PBS) for 1 h. Primary antibody staining was performed at 4°C overnight. After 3× 30-min PBS-Tx wash, the samples were incubated with 1:200 secondary antibody at 4°C overnight, followed by 3× 30-min PBS-Tx wash and mounting in 75% glycerol. Imaging of WISH samples was performed using Zeiss LSM 800 confocal microscope with oil immersion 40× objective. Whole z-stack maximum intensity projections and cell number quantification of specific cell populations were generated by Fiji-ImageJ software.
### Cryotomy and fluorescent in situ hybridization (smFISH)
C57BL6/ Cx3cr1 -GFP transgenic mice were sacrificed by decapitation. The whole pituitary was quickly dissected and fixed in 1% PFA containing 30% sucrose overnight at 4 C. The fixed tissue was then washed and equilibrated in half Tissue-Tek O.C.T Compound (Sakura) and half 60% sucrose (final 30%) mixture before positioned inside a plastic mold with only O.C.T compound and frozen by burying in dry ice powder. After the whole block turned opaque, it was stored at –80°C in a sealed plastic bag in the dark. Before cryotomy, the embedded O.C.T block was first equilibrated inside the Cryostat machine (Leica) to –25°C for 30 min followed by cryo-sectioning (7 µm) and slice collection on 22 × 22-mm glass coverslips #1 (Thermo Scientific Menzel), precoated with 0.01% L-lysine (Sigma), and stored at –80°C in a Parafilm sealed six-well plate in the dark for up to a month before further digestion and prehybridization steps. smFISH was conducted as described in ( ) with the exception that the formamide concentration was increased to 30% for prehybridization and washing. Tissue sections were mounted on Prolong Gold antifade mountant (Thermo Fisher) and images were captured using a wide-field fluorescent microscope (Nikon Eclipse Ti-E) with a cooled CCD camera equipped with oil immersion 60× objective.
### Vibratome sections
Pituitary from Cx3cr1 -GFP mouse was dissected on ice and fixed in 4% PFA overnight at 4°C. After washing, the pituitary was embedded in 3% Nobel Agarose (BD Biosciences) on ice; 50-µm coronal sections were cut using a Leica VT1000 S vibrating blade microtome (Leica) and then mounted with Aqua-Poly/Mount (Polysciences). The sections were then imaged using a Zeiss LSM 800 confocal microscope.
## Results
### scRNA-Seq revealed seven cell types in the NH and IL
The pituitary is located within a bony structure of the mouse skull, dubbed sella turcica, allowing accurate surgical isolation of this tissue. In particular, the medially located NH can be readily observed owing to its conspicuous white color, due to the high density of neurohypophyseal axons and pituicytes. We took advantage of these anatomic features to dissect neurohypophyseal tissue from three-month-old C57/BL6 male mice and thereafter performed scRNA-Seq analysis. Notably, the isolated tissue contained residual tissue from the adjacent intermediate pituitary lobe (IL), hence we took into consideration that our NH tissue preparation will contain some IL cells ( ; Extended Data ).
Single-cell RNA-Seq reveals seven cell types of dissected mouse NH. A , Schematic representation of the scRNA-Seq procedure. Neurohypophyseal tissues were dissected from five C57BL6 adult male mice and pooled. Two independent pools were separately subjected to single-cell dissociation, single-cell capturing, and library preparation using the 10x chromium platform. The two libraries were then indexed and combined for sequencing using NextSeq 500 High Output v2 kit (75 cycles). B , The two libraries were pooled and mapped on the tSNE plot, showing cell clusters of IL cells, T-cell like, VLMC, epithelial like cells, vascular endothelia macrophage/microglia, and pituicyte. Each dot represents one cell, and cells with the same color belong to one cell type. C , Dendrogram showing the distance matrix from the PCA space of the average cell among the seven cell types. The length of the path between each two cell types indicates the relativeness between them. D , A table summarizing the number of cells, average number of genes and UMIs found in each cell type.
We collected two pools of dissected neurohypophyseal tissue, each has been derived from five mice. Single cells from the dissociated tissue were thereafter captured using the 10x chromium gel beads in a droplet, followed by independent library preparation for each pool. The two sets of libraries were indexed and sequenced together ( ). The low variation between the pools was detected in the PC analysis (PCA) plot containing the two first PCs and in the tSNE plot using Seurat R package ( ). The two data sets were pooled and cell clusters were built using the 900 most variable genes using FindClusters function in Seurat package using 11 PCs with resolution 1.0 and analyzed together to create the tSNE plot ( ; Extended Data ). The normalized differentially expressed genes of each cluster (Extended Data , ) were used to identify seven major cell types, which were designated based on expression of published marker genes and following comparisons to existing single-cell database ( ). Thus, we compared our gene lists to the mouse brain atlas from the Linnarsson Lab ( ), the PanglaoDB database ( ), the cell type function from Allen Brain Atlas ( ), mouse vascular and vascular associated cell single-cell database ( ; ), and the DropViz web tool ( ). We also compared our data to published scRNA-Seq of anatomically adjacent tissues, such as the hypothalamus and the median eminence ( ; ; Extended Data , ). The identified NH cell types were labeled as: pituicyte, macrophage/microglia, vascular endothelia, T-cell like and vascular and leptomeningeal cells (VLMCs). As expected, due to the nature of the dissection procedure mentioned above, we also identified IL cells. The latter was identified by comparing to recently published whole mouse pituitary single-cell transcriptomes ( ; ; ). To determine the relativeness of the clustered cell types, we used the BuildClusterTree function in Seurat R package to generated dendrogram, representing a phylogenetic tree relating the “average” cell from each identity class ( ). The number of cells, as well as mean number of genes and average number unique molecular identifiers (UMIs) representing each of the designated cell types are shown in . Notably, the cell number does not necessarily reflect the compositional proportion in the tissue but probably randomized sampling in single-cell capturing, varied resilience of different cell types to dissociation procedure and cell type-specific RNA stability.
Following the identification of NH and IL cell types, we searched for sets of genetic markers characterizing each cell type. We generated a heatmap showing cluster analysis of the top twenty differentially expressed genes representing the transcriptomic profile of the various NH and IL cell types and then selected three feature genes, which represent each cell type ( ). These included known markers for VLMC cells ( Ogn , Lum , and Dcn ), fenestrated vascular endothelia ( Emcn , Flt1 , and Plvap ), T-cell like ( Ms4a4b and Cd3d ), and macrophage/microglia ( Ctss , C1qa , and Cx3cr1 ) ( ; ; ; ). In the case of three of the identified cell types, epithelial-like cells, pituicytes, and IL cells, there was no published database and therefore they were designated based on the top differentially expressed markers. Thus, the epithelial cell markers Krt18 , Krt8 , and Clu were top-ranked in the so-called epithelial-like cells, and the melanotrope markers Pomc and Pcsk2 were used to designate IL cells. To define the pituicyte cell type we first used Vegfa and Gja1 , which were previously associated with this cell type ( ; ). Next, we performed an unbiased bioinformatics analysis by comparing our pituicyte transcriptome to PanglaoDB, a public database for exploration of mouse and human scRNA-Seq data ( ). We employed ORA, which is a widely used approach to determine if known biological functions or gene sets are overrepresented in an experimentally-derived gene list ( ). Our unbiased comparison of the pituicyte to all PanglaoDB gene sets revealed that the pituicyte cluster is highly enriched in tanycyte (FDR = 1.20E-21) followed by astrocytes (FDR = 1.18E-07) and Bergmann glia (FDR = 5.63E-06; Extended Data ). To further illustrate the above finding, we compared the specific differentially expressed pituicyte markers with other published scRNA-Seq data of tanycytes (40% shared markers) and astrocytes (12% shared markers) in addition to PanglaoDB ( ; ; ; ; ; Extended Data ). Therefore, the unique differentially expressed featured genes we assigned for these cell types are novel markers. Thus, the novel markers Lcn2 , Cyp2f2 , and Krt18 represented epithelial-like cells; Pcsk2 , Scg2 , and Chga marked IL cells, and finally, Col25a1 , Scn7a , and Srebf1 were selected as pituicyte panel of markers ( ). The specificity of the selected marker genes is exemplified in in which a featured gene from each cluster is highlighted in the tSNE plot showing distinct distributions of different cell types ( ). A violin plot showing the normalized log-transformed single-cell expression of selected featured genes in the different cell types is shown in .
Heatmap of differentially expressed genes in neurohypophyseal and IL cell clusters. Heatmap showing scaled gene expression of the top twenty genes (square brackets) representing each of the seven cell types found in the NH and IL. Each column display gene expression of an individual cell and genes are listed in the rows. Selected marker genes are underlined in red and enlarged on the side.
Featured genes representing the landscape of the seven neurohypophyseal and IL cell types. A , Distribution of featured genes from each cell type embedded in tSNE plots. The gene expression scale was color-coded with high expression level in deep blue, low expression in gray. B , Violin plots displaying normalized log-transformed expressions of each featured gene distributed across all the seven clusters. EL, epithelial-like cells; M/m, macrophage/microglia; P, pituicyte; TL, T-cell like; VE, vascular endothelia.
### Novel pituicyte genes display higher specificity than commonly used markers
We report five selected differentially expressed genes, Srebf1 , Rax , Scn7a , Adm , Col25a1 , and Col13a1 , which showed robust expression in the majority of pituicyte population ( ). Four of these genes, Srebf1 , Rax , Adm , and Col25a1 were robustly expressed in the pituicyte population. Srebf1 displayed residual expression in a small number of epithelial-like cells but was not differentially expressed in this cluster ( ; Extended Data and ).
Novel pituicyte markers show higher specificity and robustness compared to previously used markers. A , Violin plots displaying expression distributions of novel pituicyte marker genes in seven pituitary cell types seven clusters. Srebf1 , Rax , Scn7a , Adm , Col25a1 , and Col13a1 were selected from this single-cell RNA-Seq data and mapped onto the violin plots. The y -axis represents the normalized log-transformed expression of respective genes. Each dot represents a cell and the shape of the violin represents the proportion of cells being enriched compared to the rest of cells in a given cluster. B , Previously published pituicyte markers Apoe , Vim , Gfap , S100β , Gja1 (Cx43) , and Vegfa were mapped onto the violin plots within the seven identified cell types. EL, epithelial-like cells; M/m, macrophage/microglia; P, pituicyte; TL, T-cell like; VE, vascular endothelia.
We noticed that the novel pituicyte genes revealed by scRNA-Seq displayed higher specificity than previously published pituicytes markers ( ; ; ; ; ). Thus, violin plots of our scRNA-Seq indicated that two commonly used pituicyte markers Gfap and S100β displayed low normalized log-transformed expressions in the pituicyte population. Furthermore, Apoe , which is often used as pituicyte and astrocyte marker displayed low cell-type specificity, as it was detected in all neurohypophyseal types except for T-cell like. The other three reported pituicyte markers Gja1/Cx43 , Vegfa , and Vim ( ; ; ) displayed higher normalized pituicyte expression and were somewhat more specific than Apoe ( ). Notably, although Vim displayed some expression in the pituicyte cells, it did not pass the differentially expressed criteria in the pituicyte cluster when compared to other cell types (Extended Data and ).
We next examined whether the novel pituicyte markers identified by scRNA-Seq are expressed in the mouse NH by in situ hybridization. The selected pituicyte marker Col25a1 ( ) with robust normalized expression (adjusted p = 8.38E-82, average ln fold change = 1.67) was subjected to wholemount mRNA in situ hybridization, followed by immunostaining with an antibody against the previously published pituicyte marker Vim. This analysis showed that Vim immunoreactivity is detected in a subset of Col25a1 -positive cells ( ; Extended Data ; ). This analysis was in agreement with our scRNA-Seq bioinformatic analysis ( ), suggesting that some of the commonly used pituicyte markers also label other NH cell types.
Expression of the novel pituicyte marker Col25a1 , in the NH. A , Validation of the scRNA-Seq results using wholemount staining of dissected NH derived from a C57/BL6 adult mouse. Dissected NH was subjected to fluorescent mRNA in situ hybridization with an antisense Col25a1 probe, followed by immunostaining with an antibody directed to the Vim protein and visualized by confocal microscopy. The top panels display different magnifications (scale bars, 100 µm) a single confocal optical plane of Col25a1 , Vim , and the nuclei dye, DAPI. Highly magnified field (scale bars, 10 µm) of views showing a representative Col25a1 ; Vim pituicyte (I) and another Col25a1 ; Vim neurohypophyseal cell (II). B , Numbers of different subpopulation of cell expressing Col25a1 and/or Vim were analyzed in 15 randomly chosen areas of interest (between 18,133 and 40,429 µm ). The average cell numbers and ratios, as well as the individual counting in each region of interest, are presented.
### Spatial organization of neurohypophyseal cell types
To better understand the spatial organization of neurohypophyseal cell types, we analyzed the expression of selected genetic markers representing the major NH cell types and localized the expression on a horizontal section of whole mouse pituitary ( ). We performed single molecule smFISH on a pituitary derived from a transgenic macrophage/microglia reporter ( ) as well as wholemount mRNA in situ hybridization combined with antibody staining ( ). Our scRNA-Seq analysis indicated that Srebf1 is a novel pituicyte marker displaying limited expression in the epithelial-like cells, while Cyp2f2 was highly expressed in epithelial-like cells ( , ). Accordingly, Srebf1 was prominently expressed in the NH ( ; Extended Data , ), while Cyp2f2- expressing cells were mostly located at the boundary between the IL and the AH ( ; Extended Data ). Notably, Cyp2f2 mRNA signals were much weaker in the NH compared to the IL and the AH boundary suggesting that some epithelial-like cells are also found in the NH ( ; Extended Data , ). This conclusion was further confirmed using smFISH to probe another specific epithelial-like featured gene, Lcn2, which was mainly expressed by cells located at the IL and AH boundary (Extended Data ).
Spatial distribution of pituicyte, macrophage/microglia and epithelial-like cells in the NH and IL. A , A brightfield image of a horizontal section of adult mouse pituitary showing the locations of the NH, IL, and AH. The white boxes in the brightfield image mark the locations of specific pituitary subdomains shown in the fluorescent images below (scale bar, 100 µm). B , Different fields of views (marked by roman numbers) of horizontal section (7 µm) of pituitaries derived from three-month-old Cx3cr1 -GFP macrophage/microglia transgenic reporter mouse, which were subjected to smFISH with antisense probes directed to Srebf1 (I), Cyp2f2 (II), or multiplexed smFISH of Srebf1 and Cyp2f2 on Cx3cr1 :GFP mouse (III) to observe the relative location of selected cell types. A high-magnification image of the region delineated with the white dashed box is shown. White dotted line in III’ marks the boundary between IL and NH. Note that the smFISH probe of epithelial-like cell marker, Cyp2f2 , labels the border between the IL and the AH, as well as IL cells. Arrows indicate background autofluorescent signals of circulating erythrocytes. Scale bars, 20 µm (I, II) and 10 µm (III).
Neurohypophyseal VLMCs are associated with fenestrated vascular endothelia. Confocal Z-stack (maximum intensity projection) of dissected NH, which was subjected to wholemount FISH with an antisense RNA probe directed to the VLMC marker, Lum , followed by immunostaining with an antibody directed to Plvap protein, which is a marker of fenestrated endothelia (scale bars, 100 µm). The bottom panels (labeled I–III) display high-magnification single plane confocal images of the respective regions delineated in white boxes in the top right panel (scale bars, 20 µm).
We next performed simultaneous labeling of macrophage/microglia, pituicyte, and epithelial-like cells by performing smFISH of Srebf1 and Cyp2f2 probes on pituitaries of transgenic Cx3cr1 :GFP reporter mice, labeling macrophage/microglia ( ). We observed that the Cx3cr1 : GFP-positive macrophage/microglia were distributed throughout the whole pituitary, including the NH, IL, and AH ( ; Extended Data ). These macrophages/microglia were intermingled with both Srebf1 ; Cyp2f2 pituicytes and Cyp2f2 epithelial-like cells suggesting a possible cross-talk between pituitary cells and these macrophages/microglia ( ).
Our scRNA-Seq analysis also detected an NH cell population, which co-expressed Pdgfra and Lum ( ; Extended Data ). We assumed that this cell population is similar or identical to the so-called VLMC, which has been found to localized on blood vessels of the brain ( ; ; ). We, therefore, examined the tissue distribution of VLMC cells and fenestrated neurohypophyseal vascular endothelia, which express the Plvap protein ( ; ). This analysis confirmed that as in the case of brain vasculature, VLMCs were in close association with the fenestrated endothelia of the NH ( ). Finally, although we have found a small population of T-cell like cells in the NH, immunostaining of the T-cell-specific cell surface marker, Cd3 , revealed low abundance of Cd3- positive cells in the NH ( ; Extended Data ). It is likely that these T cells are not a resident NH population but rather a transient population, which is transported from the blood. The above in situ hybridization analyses confirmed our featured gene designation determined by scRNA-Seq.
Taken together, our gene expression analysis of NH and IL reveals a comprehensive view of neuro-IL cell types in adult male mice. This study provides an important resource for specific functional studies and possible crosstalk between the various NH cell types.
## Discussion
The NH is a major neuroendocrine interface, which allows the brain to regulate the function of peripheral organs in response to specific physiologic demands. Despite its importance, a comprehensive molecular description of cell identities in the NH is still lacking. Recent studies revealed cell heterogeneity of whole pituitary gland using scRNA-Seq, however, these studies did not separate the NH from the adjacent AH and very few NH cells with limited sequence information were reported ( ; ; ). Here, utilizing scRNA-Seq technology, we identified the transcriptomes of five major neurohypophyseal cell types and two IL cell populations in the adult male mice. Using selected featured genetic markers, we mapped the spatial distribution of selected cell types in situ .
The identified differentially expressed gene clusters revealed by scRNA-Seq correspond to previously characterized cell types. Thus, previous studies reported the appearance of pituicyte ( ; ; ; ; ), macrophage/microglia ( ; ; ), and fenestrated endothelia ( ) in the NH. The identification of IL and epithelial-like cells in the present study is in agreement with other reports of these cells in the mammalian pituitary ( ; ) and also matches recent scRNA-Seq analyses of whole mouse pituitary ( ; ; ).
The novel pituicyte markers identified in our study showed more specific and robust expression than previously published pituicyte markers. Among them, Srebf1 , Col13a1 , Adm , Scn7a , and Col25a1 were not reported to be expressed by pituicyte. Vegfa was reported as pituicyte marker in both mice and zebrafish ( ; ). Srebf1 protein is involved in sterol biosynthesis process, this may be relevant to the lipid droplets that were found in ultra-structure studies of pituicyte ( ; ; ). Other prominent pituicyte markers we identified, such as Rax , Scn7a , Col25a1 , and Adm were reported as hypothalamic tanycyte markers ( ; ; ; ; ). Our finding that Rax , Scn7a , Col25a1 , and Adm are expressed in pituicytes is in line with the notion that tanycytes and pituicytes are of a common astrocytic lineage ( ; ; ). Specifically, Rax is a general tanycyte marker ( ; ; ), Scn7a , Col25a1 , and Adm were reported as β2 tanycyte markers ( ; ). Finally, Col25a1 was found to be enriched in the NH according to the Bgee database ( ).
Our novel pituicyte markers displayed greater specificity (i.e., adjusted p ≤ 0.05 for differential expression), higher expression level (average ln fold change ≥ 1) and robustness (i.e., abundance in pituicytes) compared to the most commonly used markers. Thus, as we previously showed in the case of zebrafish pituicytes ( ), we found that Apoe is broadly expressed in multiple mouse NH cell types. However, although Vim and Gfap displayed relatively low mRNA expression levels in our scRNA-Seq analysis, their protein immunoreactivity was readily detectable in the NH. This could be due to the inherently shallow sequencing method for 10x Genomics platform. The astroglial protein S100β is also used to label pituicytes ( ). It was reported that S100β is highly abundant when compared to Vim and Gfap cells ( ; ). However, in our study, S100β was not among the top differentially-expressed pituicyte genes but was found to be exclusively expressed in the VLMC cell type. In view of the gene coverage limitation of the 10x Genomics platform, S100β might have been missed in our analysis, hence, future studies should be aware of our findings regarding its expression in VLMC. Another known pituicyte-specific marker, namely Gja1 , also known as Cx43 , displayed robust specific expression in our mouse pituicyte cluster. This is in agreement with the reported findings in rat and zebrafish ( ; ).
We identified VLMC as a new neurohypophyseal cell type which is marked by the prominent expression of Pdgfra and Lum. We further showed that VLMC is associated with Plvap fenestrated neurohypophyseal capillaries. In agreement with our findings, Pdgfra ; Lum VLMC population was found in the mouse brain as vascular-associated cell type ( ) or as fibroblast-like cells that are loosely attached to vessels and located in between smooth muscle cells and astrocyte end-feet ( ). Although VLMC express some markers of oligodendrocyte precursor cells (OPCs), such as Pdgfra , they are distinct from OPCs and oligodendrocyte lineages ( ).
Importantly, previous reports have described the existence of OPCs in the NH ( , ; ). We did not detect OPCs in the present study, however, this could be due to the low abundance of these cells in our tissue. Alternatively, because Virard et al. relied on Pdgfra as a sole OPC marker, it is possible that they misidentified VLMCs as OPCs. Notably, Virard et al. reported that these Pdgfra cells were shown to be pituicyte progenitors in their study ( ). Similarly, other studies reported that VLMC display multipotent stem cell niche function in the CNS and other organs suggesting that they may play similar roles in NH function ( ; ). Further studies are required to determine whether neurohypophyseal OPCs are in fact VLMCs and whether VLMCs are pituicyte progenitors.
Although pericytes have been previously reported to be associated with neurohypophyseal capillaries ( ; ), we did not detect them in the present study, possibly due to the fact that isolating pericytes requires different tissue dissociation conditions. It is also possible that other minor neurohypophyseal cell populations have been missed, which may be revealed if more cells would be sampled.
Macrophage/microglia were found in our study as prominent NH resident cells. Previous reports showed that neurohypophyseal microglia in rat endocytose and digest axonal terminus ( ), whereas the pituicyte envelops the buttons of axons ( ) and provide cues for the permeable endothelial fate ( ). Our finding that macrophage/microglia are closely located to the pituicytes in the NH is in agreement with such functional cooperation between these two cell types.
In summary, our transcriptome analysis of individual cells derived from NH and IL tissues of adult male mice have revealed the cellular heterogenicity of the NH and provide novel molecular markers for the major cells in those tissues. We present a valuable resource that will serve as the basis for further functional studies.
|
Visual Abstract
Blocking inhibitory factors within CNS demyelinating lesions is regarded as a promising strategy to promote remyelination. Bone morphogenetic protein 4 (BMP4) is an inhibitory factor present in demyelinating lesions. Noggin, an endogenous antagonist to BMP, has previously been shown to increase the number of oligodendrocytes and promote remyelination in vivo. However, it remains unclear how BMP4 signaling inhibits remyelination. Here we investigated the downstream signaling pathway that mediates the inhibitory effect that BMP4 exerts upon remyelination through pharmacological and transgenic approaches. Using the cuprizone mouse model of central demyelination, we demonstrate that selectively blocking BMP4 signaling via the pharmacological inhibitor LDN-193189 significantly promotes oligodendroglial differentiation and the extent of remyelination in vivo . This was accompanied by the downregulation of transcriptional targets that suppress oligodendrocyte differentiation. Further, selective deletion of BMP receptor type IA (BMPRIA) within primary mouse oligodendrocyte progenitor cells (OPCs) significantly enhanced their differentiation and subsequent myelination in vitro . Together, the results of this study identify that BMP4 signals via BMPRIA within OPCs to inhibit oligodendroglial differentiation and their capacity to myelinate axons, and suggest that blocking the BMP4/BMPRIA pathway in OPCs is a promising strategy to promote CNS remyelination.
## Significance Statement
Blocking inhibitory factors within central demyelinating lesions is a promising strategy to promote remyelination. Previous studies have established that exogenous bone morphogenetic protein (BMPs) inhibit oligodendrocyte differentiation during CNS development and after injury. Here, we demonstrate that blocking endogenous BMP4 signaling via a selective pharmacological approach promotes oligodendroglial differentiation and the rate of remyelination after a central demyelinating insult in vivo. Using in vitro analysis, we identify that oligodendrocyte progenitor cell (OPC)-expressed BMP Type I receptors mediate this effect. Together, our data propose that blocking the BMP4 signaling pathway at the Type I receptors in OPCs is a promising strategy to promote CNS remyelination.
## Introduction
In central demyelinating diseases such as multiple sclerosis (MS), oligodendrocytes (OLs) are targeted through inflammatory activity and the myelin sheath surrounding axons is degraded ( ; ). The degree of remyelination within demyelinating lesions is variable; although MS lesions remyelinate relatively efficiently early on in disease, at later stages many lesions remain chronically demyelinated ( ). These chronically demyelinated lesions typically contain oligodendrocyte progenitor cells (OPCs) and premyelinating OLs that have “stalled” in their differentiation, implicating blocked OL differentiation as a major contributing factor to remyelination failure ( ; ). Although the full complement of factors that inhibit OL differentiation and remyelination in the context of MS are yet to be completely elucidated, they most likely include a variety of inhibitory signals present within the lesion environment as well as an absence of positive signals ( ; ; ). Thus, blocking the action of inhibitory factors is regarded as a leading strategy to promote endogenous CNS remyelination ( ).
The bone morphogenetic proteins (BMPs) are a group of secreted proteins that are part of the larger transforming growth factor-β (TGF-β) superfamily ( ) and play critical roles in neural development and gliogenesis ( ; ). Of the 20 BMPs, BMP4 has a prominent role in promoting astroglial and inhibiting oligodendroglial specification ( ; ). In vitro , BMP4 exerts stage-specific inhibitory effects on OPCs ( ), in particular inhibiting the production of myelin proteins by immature OLs ( ). In vivo, transgenic overexpression of BMP4 led to an increase in the number of astrocytes and a decrease in the number of oligodendrocytes in the murine CNS ( ). In the context of demyelinating disease, BMP4 mRNA is detected in human demyelinated MS lesions ( ), and is expressed by astrocytes, microglia, and infiltrating immune cells ( ). Astrocytes also express a high level of BMP4 in chronic lesions that have failed to remyelinate ( ). Through using the cuprizone-induced murine model of CNS demyelination, we have previously found that BMP4 mRNA is upregulated in the mouse corpus callosum (CC) following a demyelinating insult in vivo ( ). Furthermore, we demonstrated that inhibiting BMP4 signaling following cuprizone-induced CNS demyelination via infusion of its extracellular antagonist noggin resulted in more mature oligodendrocytes and more remyelinated axons ( ; ). However, in addition to BMP4, noggin also inhibits other BMPs such as BMP2, 7, 13, and 14 ( ). Due to the promiscuous inhibitory effect of noggin and the potential effects it exerted on oligodendroglia, astrocytes, and microglia, the precise influence that the inhibition of BMP4 exerts on remyelination and the cell type mediating the effect remain unclear.
BMP4 signals through membrane-bound receptor complexes composed of two type I receptors and two type II receptors. While several type I receptors exist, BMP4 has the greatest affinity for the BMP type I receptors BMPRIA (also known as ALK3) and BMPRIB (also known as ALK6; ; ). In the presence of BMP4, BMPRIA and BMPRIB initiate signaling via phosphorylation of SMAD1, SMAD5, and SMAD8 ( ) and the pharmacological inhibitor LDN-193189 selectively blocks phosphorylation SMAD1/5/8 ( ). To specifically interrogate the influence that BMP4 signaling exerted on remyelination, we infused LDN-193189 into the brain following cuprizone-induced demyelination and found it significantly enhanced oligodendroglial differentiation and their subsequent remyelination following the demyelinating insult in vivo . This finding is also supported in vitro in which LDN-193189 significantly enhanced OPC differentiation and myelination. Further, by using a tamoxifen-dependent inducible conditional knockout (KO) mouse strategy ( Pdgfra-CreER :: Bmpr1a ) to specifically ablate BMPRIA expression within OPCs, we identified that selectively deleting of BMPRIA in OPCs significantly potentiated their differentiation into mature oligodendrocytes and increased myelin formation in vitro . Together, our findings indicate that BMP4 acts on OPC-expressed BMPRIA receptors to inhibit oligodendroglial differentiation and myelination, and that blocking BMPRIA signaling OPCs is a promising strategy to promote CNS remyelination.
## Materials and Methods
### Animals and reagents
All animal procedures were performed in accordance with the Florey Institute of Neuroscience and Mental Health Animal Ethics Committee regulations. Female mice (7–8 weeks old) were used for in vivo cuprizone experiments and postnatal day 5 (P5) to P7 mice of either sex were used for in vitro experiments. C57BL/6 mice were purchased from the Animal Resource Center (Canning Vale, WA, Australia). Pdgfra-CreER :: Bmpr1a mice were generated by crossing Pdgfra-CreER mouse line (provided by Dr. Kaylene Young, University of Tasmania, Hobart, TAS, Australia; ) with Bmpr1a mouse colony (also known as Alk3 ; provided by Professor Yuji Mishina, University of Michigan, Ann Arbor, MI; ). Pdgfra-CreER :: Bmpr1a mice have a tamoxifen-inducible deletion of the Bmpr1a allele from the start of the sequence to the end of exon 2, rendering it untranscribable ( ). All animals used for this study were bred at the Core Animal Services facility of the Florey Institute of Neuroscience and Mental Health. All chemicals were obtained from Sigma-Aldrich, unless otherwise indicated.
### Cuprizone protocol
Cuprizone-mediated demyelination was induced by feeding 8- to 10-week-old female mice (C57BL/6) powdered feed (Barastoc) containing 0.2% cuprizone (w/w: bis-cyclohexanoneoxaldihydrazone) for 5 weeks, as previously described ( ; ). Mice were then returned to a normal diet for either 0 or 1 week, according to the experimental paradigm. During the 5 week demyelination phase, feed was refreshed every 3 d, with ∼20 g provided per mouse for this period. Mice were weighed daily to monitor extreme fluctuations in weight and to ensure no mouse lost >15% of its initial weight during the protocol. Unchallenged control mice were fed identical feed without added cuprizone.
### Intracerebroventricular infusion
Following cuprizone feeding, animals received either LDN-193189 (400 ng/d; Stemgent) or artificial CSF (aCSF) via intracerebroventricular osmotic pumps (catalog #1007D, Alzet). The concentration of LDN-193189 was based on our previous study using noggin ( ). Mice were deeply anaesthetized using 2.5% isoflurane and attached to a stereotactic frame. The scalp was cut sagittal to the cervical spine. The pumps were used in conjunction with Alzet Brain Infusion Kit III to implant a cannula into an entry point drilled 0.5 mm anterior to bregma, 0.7 mm laterally from the longitudinal midline and at a depth of ∼1-2 mm. Canullae were fused to the skull using Araldite, and the incision was sutured with Vicryl veterinary sutures and disinfected using Betadine iodine solution. Mice were allowed to recover for >30 min at 30°C before returning to the cage. Mice were monitored daily to observe any symptoms of distress or infection. After 7 d of continuous infusion, animals were killed and the brain removed for immunohistochemical and histologic analysis.
### Post-cuprizone tissue collection
Following cuprizone withdrawal, mice were transcardially perfused using 0.1 mouse tonicity PBS (MT-PBS) as a buffer and 4% PFA (in MT-PBS, 15 ml/mouse) as a fixative. Brains were dissected and postfixed overnight with 4% PFA in MT-PBS and rinsed the following day with MT-PBS before being cut coronally into 1 mm sections. For electron microscopy, sections containing the most caudal region of the CC (approximately −2.12 mm from bregma) were trimmed to expose the splenium of the caudal CC and placed in Karnovsky’s buffer (4% PFA, 2.5% glutaraldehyde in 0.1 sodium cacodylate) overnight before being rinsed three times in 0.1 sodium cacodylate. For immunohistochemical analyses, sections containing the caudal corpus callosum (approximately −1.12 mm from bregma) were placed in 30% sucrose (in MT-PBS with 0.1% sodium azide) overnight. Sucrose-treated sections were frozen in Tissue Tek Optimum Cutting Temperature (O.C.T., Sakura) solution using chilled isopentane and stored at −80°C.
### Immunohistochemistry
Coronal brain sections were cut at 10 or 12 μm thin and blocked for 1 h in antibody diluent (10% normal goat serum, 0.3% Triton X-100 in MT-PBS) at room temperature (RT) before exposure to primary antibodies diluted in antibody diluent overnight at 4°C. The following primary antibodies were used at a dilution of 1:200: rat anti-myelin basic protein (MBP; catalog #MAB386, Abcam), rabbit anti-OLIG2 (catalog #ab9610, Millipore), rat anti-CC1/APC (catalog #D35078, Calbiochem), mouse anti-platelet-derived growth factor receptor α (PDGFRα; catalog #AF1062, R&D Systems), mouse anti-glial fibrillary acidic protein (GFAP; catalog #MAB360, Millipore), and goat anti-IBA-1 (ionized calcium binding adaptor molecule 1; catalog #ab5076, Abcam). Cryosections were then rinsed with MT-PBS three times for ∼5 min followed by the appropriate fluorophore-conjugated secondary antibodies (all 1:500 in antibody diluent; Thermo Fisher Scientific) for 60 min at RT. Sections were rinsed twice in MT-PBS before adding Hoechst (1:10,000 in MT-PBS; catalog #33342, Invitrogen) for 10 min. Cryosections were rinsed twice in MT-PBS, and a coverslip was mounted with Cytomation fluorescence mounting medium (Dako). Six sections per animal from a minimum of three animals per group were analyzed, and images captured by a Carl Zeiss LSM 780 confocal fluorescent microscopy. All images were acquired using the same settings and were analyzed by an operator blinded to conditions using FIJI (ImageJ 1.51K, National Institutes of Health) software ( ). For OLIG2 /CC1 /PDGFRα cell counts, cells were counted from the entire visible corpus callosum per image field with the same size of area. For MBP immunostaining, a central area of 200 μm was measured for integrated density (the product of the mean gray value of each pixel, ranging from 0 to 255, and the total area) using the “Measure” function in FIJI. For GFAP and IBA-1 immunostaining, the entire corpus callosum was measured using the “Trace” function.
### Spectral confocal reflection microscopy
Spectral confocal reflection (SCoRe) imaging was performed on brain sections to assess the extent of myelin damage in cuprizone mice using published methods ( ; , ). Briefly, mice were perfused with 4% PFA, and their brains were dissected, frozen, and cryosectioned at 12 μm. Coronal sections of caudal brains were imaged via a Zeiss 780 LSM Confocal Microscope with a water-immersion objective [Zeiss W Plan-Apochromat 20×/1.0 numerical aperture (NA) differential interference contrast M27, 70 mm] using 458, 561, and 633 nm laser wavelength through the Tunable Lazer In Tune 488-640 filter/splitter wheel and a 20/80 partially reflective mirror. The reflected light was collected using three photodetectors set to collect light through narrow bands defined by prism and mirror-sliders, centered around the laser wavelengths 488, 561, and 633 nm. Sections were immersed in MT-PBS and a 20× dipping objective was equipped before imaging. The midline corpus callosum was located, and a 3 × 2 tile scan image was taken of each section. The channels from each photodetector were then additively combined as a one-color composite. Myelinated area was calculated using ImageJ by first applying a Z -stack transformation and then setting a threshold of 50 pixels. Measurements of the resulting area were obtained with the “Measure” function and divided by the total area of the region of interest (ROI). The percentage area of positive signal was computed for each image. For quantification, a minimum three separate ROIs per image and three images per tile (using a 20×/1.0 NA objective at a z -depth 4 μm from the tissue surface) per treatment group were used and statistically analyzed.
### Transmission electron microscopy
Mouse caudal CC samples were embedded in resin for 5 d before trimming and sectioning using an ultramicrotome. Semi-thin sections (0.5 μm) were taken and imaged using toluene blue staining to identify ROI. Ultra-thin sections (70 nm) were then taken and imaged using a transmission electron microscopy (TEM). Images were taken at 5000× and 10,000× magnification per animal using a JEOL 1011 transmission electron microscope. Three 10,000× images were taken per hexagonal bounding grid corresponding to a size of 250 μm , with six distinct fields of view were imaged at 10,000× magnification per animal. Images were used to count myelinated axons, measure axon diameters, and g-ratios in FIJI. For g-ratio analysis, a minimum of 90 axons per animal from a minimum of three mice per group were measured.
### Primary mouse OPC culture
Oligodendrocyte progenitor cells were isolated from P5 to P6 wild-type or transgenic mouse pups using a previously published protocol ( ). Cultures were grown on poly- -lysine (PDL)-coated vessels in defined serum-free media and supplied daily with Platelet-Derived Growth Factor AA (PDGF-AA) (10 ng/ml; PeproTech), Neurotrophin 3 (NT-3; 1 ng/ml, PeproTech), and ciliary neurotrophic factor (CNTF; 10 ng/ml; PeproTech). For the differentiation assay, PDGF is withdrawn from OPC culture, and cells were cultured in Sato media containing OL differentiation factor thyroid hormone T3 (3,3′,5-Triiodo-L-thyronine sodium; 4 ng/ml in Sato media; Sigma-Aldrich), CNTF (10 ng/ml), forskolin (5 µM), and NT-3 (1 ng/ml). For small molecule inhibitor experiments, OPCs were either cultured in the differentiating condition (see above) with LDN-193189 (0.2 µ ; Stemgent) or vehicle (DMSO) being added 30 min before BMP4 addition (1 ng/ml; catalog #314-BP, R&D Systems). In some cultures, OPCs were isolated from Pdgfra-CreER :: Bmpr1a (Cre[+]) and Bmpr1a control (Cre[−]) mice. These OPCs were treated with 4-hydroxy-tamoxifen (referred to as “4OHT,” 500 nM in EtOH; Sigma-Aldrich) to induce the knockout of BMPRIA or an equal volume of vehicle (ethanol). For the differentiation assay, OPCs were treated with either BMP4 (1 ng/ml) or vehicle (0.1% BSA in D-PBS) with the addition of differentiation Sato media containing T3. For some experiments, 4OHT or vehicle (ethanol) were added 24 h before BMP4 addition (1 ng/ml; catalog #314-BP, R&D Systems). After a set time point as indicated, cells were fixed in 4% PFA for 20 min followed by immunocytochemical staining (see below). For differentiation assays, three technical replicates and a minimum of three mice per condition or genotype were used.
### Dorsal root ganglion/OPC coculture
Dorsal root ganglion (DRG)/OPC cocultures were established based on published techniques ( ). Briefly, OPCs were isolated as detailed above and seeded onto coverslips containing purified DRGs at a density of 2 × 10 OPCs per 22 mm poly-ornithine (Sigma-Aldrich)/PDL-coated coverslip and incubated overnight to facilitate attachment. DRG-OPC cocultures were maintained for 14 d in a defined coculture media containing a 1:1 ratio of Sato medium/Neurobasal medium (Thermo Fisher Scientific) with 2% NeuroCult SM1 supplement (Stem Cell Technologies). Media were changed every 2–3 d. For small-molecule inhibitor experiments, cells were cultured either with LDN-193189 (0.2 µ ; Stemgent) or vehicle (DMSO) for 30 min before BMP4 addition (1 ng/ml; catalog #314-BP, R&D Systems) at each feed. For transgenic experiments, OPCs isolated from Pdgfra-CreER :: Bmpr1a and Bmpr1a control (Cre[−]) mice were treated with 4OHT or vehicle control (ethanol) for the first 48 h following coculturing with neurons. After 14 d, cocultures were immunostained or protein was extracted for Western blotting as described below.
### Immunocytochemistry
After fixation with 4% PFA for 18 min, cells were rinsed three times in MT-PBS. Cells were blocked with 10% normal goat serum with 0.3% Triton X-100 in MT-PBS for 60 min at RT, followed an incubation with primary antibodies against GFAP (1:200, mouse, catalog #MAB360, Millipore; 1:200, rabbit, catalog #Z03374, DAKO), MBP (1:50, mouse, catalog #MAB381, Millipore; 1:100, rat, catalog #ab980, Millipore), or rabbit anti-Neurofilament (1:200; catalog #AB1987, Millipore). Cells were then rinsed with MT-PBS followed by the appropriate fluorophore-conjugated secondary antibodies (all 1:500 in antibody diluent, Thermo Fisher Scientific) for 60 min at RT. Cells were rinsed twice in MT-PBS before adding Hoechst (1:10,000 in MT-PBS; catalog #33342, Invitrogen) for 10 min. Cells were rinsed twice in MT-PBS and mounted with Cytomation fluorescence mounting medium (Dako) on SuperFrost Plus glass slides (Thermo Fisher Scientific). Six fields per culture, and three technical replicate from a minimum of three animals per condition or genotype were analyzed, and images captured by a Carl Zeiss Axioplan 2 epifluorescence upright microscope.
### Immunocytochemical quantification
For OPC culture images, all Hoechst nuclei were counted using Adobe Photoshop (version CS5, Adobe), and the morphology of each Hoechst cell was designated as an astrocyte (GFAP ), immature oligodendrocyte (MBP ), or mature oligodendrocyte (MBP ), or was unclear (MBP/GFAP ). These populations (excluding the “unclear” cells) were then graphed as a proportion of all Hoechst cells. For the DRG/OPC coculture analysis, an average length for a clearly defined segment was subjectively defined at the start of counting using the ImageJ measure tool, and then the same length is used to count further segments. This was consistently applied throughout all treatments by one counter over one session.
### Western blotting analysis
Total protein of OPC/DRG cocultures was extracted using TNE buffer supplemented with proteinase inhibitor (Roche), separated by SDS-PAGE (200 V; ∼30-40 min) and transferred to PVDF membrane using an iBlot quick transfer dry blot system (Life Technologies). Protein blots were blocked with 5% nonfat milk powder in Tris-buffered saline/Tween 20 (TBST; 50 m Tris, 150 m NaCl, 0.05% Tween 20, all from Sigma-Aldrich) for 5-10 min, followed by three rinses with TBST. Blots were subsequently probed with antibodies against myelin proteins MBP (1:50; catalog #AB980, Millipore Bioscience Research Reagents), Myelin oligodendrocyte glycoprotein (MOG) (1:50; catalog #MAB5680, Millipore), or BMPRIA (1:200; catalog #38560, Abcam) overnight at 4°C. An antibody against β-actin (1:5000 in TBST + 2% BSA; catalog #A5441, Sigma-Aldrich) was also added as an internal loading control. Following three rinses with TBST, blots were incubated with HRP-conjugated secondary antibodies (1:5000; Cell Signaling Technology).
### RNA isolation and quantitative real-time PCR analysis
Following differentiation assay, OPCs were rinsed once with cold D-PBS and lysed using a cell scraper with addition of 600 µl RLT-plus buffer (Qiagen) supplemented with 1% 2-mercaptoethanol (Sigma) as an RNase inhibitor. Pure OPC RNA was acquired by following RNeasy Plus Mini protocol (Qiagen). RNA was reverse-transcribed using Applied Biosystems reagents and following manufacturer’s protocol. Following synthesis of cDNA, samples were loaded undiluted into 96-well plates and SYBR Green Master Mix (Applied Biosystems) was added along with primers. The plate was sealed with optical film (Applied Biosystems) and centrifuged for 1 min at 1000 rpm. It was then loaded into an Applied Biosystems ViiA 7 quantitative real-time PCR (qRT-PCR) system. Average expression of housekeeping gene 18S was used to normalize gene expression using the ΔΔCt method. Primer sequences used were shown in (all primers are specific for Mus musculus ).
Primer sequences used for qRT-PCR
### Analyzing multiple transcriptional changes using RT profiler PCR array
Purified mRNA reverse transcribed using the RT First Strand kit (catalog #330401, Qiagen) according to the manufacturer instructions. A mouse TGF-β/BMP Signaling Pathway RT Profiler PCR Array (catalog #PAMM-035C, SABiosciences) was used to assess the expression of 84 genes specific to TGF-β/BMP signaling activity. Reverse-transcribed cDNA was added to SYBR Green ROX Master mix (catalog #330520, Qiagen) as per manufacturer instructions and loaded into the 96-well plate PCR array. Samples were run on an Applied Biosystems ViiA 7 qRT-PCR system (experimental setup settings were provided by SABiosciences). Average transcription of housekeeping genes provided in the PCR array was used to normalize gene expression using the ΔΔCt method. Data were analyzed using an online software program provided by the manufacturer. Data are reported as changes in fold regulation, defined as equal to the fold change when the fold change value is positive, and the negative inverse of the fold change when the fold change value is negative. A full list of genes analyzed using this method can be found at .
### Statistical analysis
All statistical tests were performed using GraphPad Prism 7 (GraphPad Software). Assessors were blinded to conditions, groups, or genotypes during analysis. All data are presented as the mean ± SEM.
## Results
### LDN-193189 infusion promotes remyelination following cuprizone-induced demyelination in vivo
To investigate the influence that BMP4/BMPRI signaling exerts on remyelination, we subjected C57BL/6 mice to cuprizone-induced demyelination as published previously ( ). Mice were fed cuprizone for 5 weeks to induce demyelination in several white matter tracts of the brain including the CC. Following cuprizone withdrawal, mice were infused with either LDN-193189 (400 ng/d), a previously characterized inhibitor of BMPRIA and BMPRIB receptor signaling ( ), or vehicle (0.1% DMSO in aCSF) for 7 d and allowed to recover. A parallel cohort of control mice were fed cuprizone and killed at the end of a 5 week period (with no recovery) to assess the extent of demyelination.
The extent of demyelination in the medial caudal corpus callosum of 5-week-old cuprizone-fed mice (no recovery), and the extent of remyelination in cuprizone-fed mice following 7 d infusion with vehicle or LDN-193189 after cuprizone withdrawal was assessed in three ways. We first performed immunohistochemical analysis of the myelin protein marker MBP, as an indicator of myelination. Unchallenged age-matched mice were used as healthy controls to assess the basal level of myelination. While there were clear qualitative effects on MBP staining following cuprizone exposure ( , top panels) and LDN infusion ( , bottom panels), the assessment of the intensity of MBP staining revealed no significant difference between the groups ( ). This could be due to the presence of myelin debris (positive for MBP) after cuprizone-induced demyelination. We did observe a trend difference between vehicle- and LDN-infused mice following 1 week of recovery from cuprizone, but this was not significant ( , right histogram; p = 0.21 ; , statistics). We next used SCoRe imaging to assess the extent of remyelination. SCoRe imaging is a label-free (antibody-free) technique allowing for high-resolution quantitative in vivo imaging of substantial areas of myelinated white matter tracts such as the CC ( ). Using the SCoRe imaging, at the end of 5 weeks of cuprizone feeding, we observed a significant reduction (>10 fold) in the percentage of myelinated area in the corpus callosum of cuprizone-fed mice compared with healthy control mice [ , top panels (quantified in D ); p = 0.0047 ]. When assessing the 1 week recovery groups, we found that mice infused with LDN-193189 for 7 d showed a significant increase (approximately twofold) in the myelinated area compared with vehicle-infused control mice [ , bottom panels (quantified in D ); p = 0.014 ], indicating a greater extent of remyelination.
Inhibiting BMP4/BMPRI signaling following demyelination promotes remyelination in vivo . A , Representative MBP IHC images showing myelin protein in the caudal corpus callosi of healthy control (control) and 5 week cuprizone-challenged mice (Cuprizone 5w, top panels); and 5 week cuprizone-challenged mice followed by 1 week of recovery with vehicle (Vehicle recovery) or LDN-193189 (LDN recovery) infusion (bottom panels). B , Quantification of integrated density of MBP immunostaining. No significant differences were observed between control and cuprizone-fed mice, or between vehicle- and LDN-infused mice. C , Representative SCoRe images to identify myelin in the caudal corpus callosi of healthy control (control) and 5 week cuprizone-challenged mice (Cuprizone (5w), top panels); and 5 week cuprizone-challenged mice followed by 1 week of recovery with vehicle (Vehicle recovery) or LDN-193189 (LDN recovery) infusion (bottom panels). D , Quantification of myelinated area (SCoRe signal that is pixelated) as a percentage of the total area measured. The SCoRe signal is significantly reduced in 5 week cuprizone-challenged mice [Cuprizone (5w) compared with healthy control (Ctrl) mice, confirming demyelination ( B )]. LDN-193189-infused mice display a significantly greater SCoRe signal than the vehicle-infused control group ( C ), indicating greater remyelination. E , TEM cross-sectional images of caudal corpus callosum axons of 5 week cuprizone-challenged mice followed by 1 week of recovery with vehicle (Vehicle recovery) or LDN-193189 (LDN recovery) infusion. F , A scatterplot comparison of g-ratio distribution relative to axonal diameter. LDN-infused mice had a significantly higher average g-ratio than vehicle-infused controls ( p = 0.016). G , Proportion of total myelinated axons in the caudal corpus callosum of vehicle- and LDN-infused mice following 5 weeks of cuprizone administration. A trend, but a nonsignificant increase, was observed in LDN-infused mice compared with vehicle controls. H , The average g-ratio of axons in the caudal corpus callosum of vehicle- and LDN-infused mice following 5 weeks of cuprizone treatment. Mice treated with LDN-193189 after 5 weeks of cuprizone had more thinly myelinated axons (high g-ratio) in the corpus callosum compared with vehicle-infused mice. I , Number of axons in the corresponding g-ratio range for vehicle- versus LDN-infused mice following 5 weeks of cuprizone treatment. EM analysis indicated a higher number of axons with thinner myelin in the LDN-treated group, indicating greater remyelination ( N = 4-6 animals/group for SCoRe; N = 3 animals/group for EM). * p < 0.05, **** p < 0.0001. Scale bars: SCoRe images, 50 µm; TEM images, 2 µm.
Statistics
To ascertain the effect of LDN-193189 on the extent of remyelination and ultrastructure of myelinated axons, sagittal sections of caudal corpus callosum were assessed using TEM. Comparing the raw counts of total myelinated axons per image field in the corpus callosum of mice treated with either LDN-193189 or vehicle revealed a trend increase in the percentage of myelinated axons compared with the control group [ (quantified in G ); p = 0.20 ]. However, when we compared g-ratios (as an indicator of myelin thickness), both the average g-ratio ( ) and distribution of g-ratios relative to axonal diameter ( ), were greater in mice infused with LDN-193189 compared with vehicle-infused controls, indicative of thinner myelin ( ; p = 0.016b2). Thinner myelin sheaths are likely to have been recently myelinated after a demyelinating insult, as they have not completed the full number of wraps around the axon compared with myelin sheaths formed during development ( ). Importantly, analysis of the number of myelinated axons grouped by the range of g-ratios demonstrated that the LDN-infused group had significantly more myelinated axons with g-ratios >0.81 ( ; p = 0.035 ) compared with the control group, indicative of more axons with thin myelin sheaths. Therefore, our EM results together with the SCoRe imaging data collectively suggest that LDN infusion significantly enhances the extent of remyelination, resulting in more remyelinating axons than the control group.
### LDN-193189 infusion promotes oligodendrocyte differentiation following demyelination in vivo
Having shown that infusion of LDN-193189 significantly enhanced the extent of myelin repair in vivo , we next sought to determine the effect that infusion exerted on oligodendroglial populations. To address this, we assessed the number of OLIG2 oligodendroglia as well as the proportion of OLIG2 /CC1 mature OLs and OLIG2 /PDGFRα OPCs in the medial caudal corpus callosum ( ). As expected, there was significantly fewer OLIG2 oligodendroglial cells in the corpus callosum of mice treated with cuprizone for 5 weeks versus control mice [ (quantified in B ); p = 0.030 ]. Interestingly, there was no significant difference in the number of OLIG2 cells between the vehicle- and LDN-treated mice [ (quantified in C ); p = 0.26 ], suggesting that LDN infusion does alter the overall number of oligodendroglial lineage cells during remyelination. Consistent with previous studies ( ; ), there was a significant reduction in the proportion of OLIG2 /CC1 mature OLs at the peak of demyelination (5 weeks of cuprizone) compared with non-cuprizone-challenged healthy control mice [ (quantified in D ); p = 0.0002 ], which is accompanied by a significantly higher proportion of OLIG2 /PDGFRα OPCs [ (quantified in F ); p = 0.0025 ]. Interestingly, after 1 week of recovery following cuprizone withdrawal, LDN-193189-infused mice had a significantly higher proportion of OLIG2 /CC1 mature OLs compared with the vehicle infused mice [ (quantified in E ); p = 0.0059 ]. This is accompanied by fewer OPCs in these animals compared with the vehicle control group [ (quantified in G ); p = 0.016 ]. Thus, our results show that blocking BMP4/BMPRI signaling enhances the differentiation of OPCs into mature OLs during remyelination in vivo . Coupled with the SCoRe and TEM analysis, it suggests that inhibiting BMPRIA/B signaling with LDN-193189 leads to a greater number of OPCs contacting axons, differentiating, and forming new myelin; this subsequently leads to a greater number of axons with high g-ratios, indicative of remyelination.
Inhibiting BMP4/BMPRI signaling following demyelination promotes oligodendrocyte differentiation in vivo . A , Representative micrographs of immunostaining in the caudal corpus callosi of healthy control mice (control), mice subjected to 5 weeks of cuprizone treatment (Cuprizone 5w), and mice subjected to 5 weeks of cuprizone with either vehicle (Vehicle recovery) or LDN-193189 (LDN recovery) infusion for 1 week, and immunostained with OLIG2 and either PDGFRα or CC1. B , C , Analysis of OLIG2 cell number in healthy control mice (control), mice subjected to 5 weeks of cuprizone treatment (Cuprizone 5w), and mice infused with either vehicle (Vehicle recovery) for 1 week or LDN-193189 (LDN recovery) for 1 week. As expected, the total number of OLIG2 cells is significantly decreased after 5 weeks of cuprizone treatment compared with controls. D , Quantification of the proportion of OLIG2 /CC1 mature oligodendrocytes showing a significant reduction at the end of cuprizone feeding. E , LDN-193189-infused mice have a significantly higher proportion of mature oligodendrocytes compared with the vehicle control group following 1 week recovery. F , Quantification of the proportion of OLIG2 /PDGFRα OPCs showing a significant increase at the end of cuprizone feeding. G , LDN-193189-infused mice have a significantly smaller fraction of OPCs compared with the vehicle control group following recovery ( N = 4-6 animals/group). * p < 0.05, ** p < 0.01, *** p < 0.001. Scale bar, 50 µm.
It has been previously identified that exogenous BMP4 promotes astrogliogenic effect in vitro and in vivo, whereas blocking its signaling inhibits this effect ( ; ; ). Thus, we next investigated whether LDN-193189 infusion also affected astrocytes ( ). Immunostaining of caudal corpus callosum sections of normal control mice for GFAP showed a low level of positive immunostaining. As astrocytes can both proliferate and ramify in response to injury ( ), we assessed the integrated density of GFAP fluorescence of the section and observed a substantial increase in GFAP immunofluorescence signal at the end of 5 weeks of cuprizone feeding compared with healthy controls [ (quantified in C )]. This is expected, as astrogliosis is observed from 3 to 4 weeks after cuprizone ( ). Interestingly, the administration of LDN-193189 resulted in no significant effect on GFAP immunofluorescence intensity compared with the vehicle control [ (quantified in C ); p = 0.85], suggesting that blocking BMP4/BMPRI signaling via LDN-193189 infusion exerted little influence on astrogliosis during remyelination in vivo .
Inhibiting BMP4/BMPRI signaling exerts no influence on astrocytes or microglia in vivo. A , B , Representative micrographs of immunostaining in the caudal corpus callosi of healthy control mice (control), mice subjected to 5 weeks of cuprizone treatment (Cuprizone 5w), and mice subjected to 5 weeks of cuprizone treatment with either vehicle (Vehicle recovery) or LDN-193189 (LDN recovery) infusion for 1 week, and immunostained with GFAP ( A ) or IBA-1 ( B ). C , Quantification of the integrated density of GFAP immunofluorescence. There is no significant change in GFAP immunofluorescence at peak demyelination (Cuprizone 5w; left) or following the infusion of LDN-193189 (LDN recovery) for 1 week compared with control groups (Control, Vehicle; right panel). D , Quantification of the integrated density of IBA-1 immunofluorescence. There is a significant increase in IBA-1 immunofluorescence in the corpus callosum at peak demyelination (Cuprizone 5w; left); however, there is no significantly different increase in IBA-1 immunofluorescence between vehicle (Vehicle recovery) or LDN-193189 (LDN recovery) infusion during 1 week of recovery after cuprizone treatment (right; N = 4-6 animals/group). **** p < 0.0001. Scale bar, 50 µm.
As microglia represent a considerable proportion of cells in the corpus callosum during cuprizone-induced demyelination ( ), we then assessed whether LDN-193189 infusion affected microglia by quantifying the degree of IBA immunofluorescence in the corpus callosum ( ). As expected, there is a significant increase in the integrated density of IBA-1 immunofluorescence in the caudal corpus callosum of mice at the peak of demyelination (following 5 weeks of cuprizone) compared with healthy controls [ quantified in D ); p < 0.0001 ], indicating a dramatic increase in the inflammatory response to demyelination. However, there was no significant difference in the integrated density of IBA-1 immunofluorescent between mice infused with LDN-193189 and vehicle following 1 week of recovery [ (quantified in D ); p = 0.61 ], suggesting that LDN-193189 exerted no significant influence on microglia during remyelination in vivo. Together, these data suggest that blocking BMP4/BMPRI signaling in the murine cuprizone model of demyelination exerts little effect on either astrocytes or microglia, but rather selectively enhances OPC differentiation to promote myelin repair in vivo .
### Inhibiting BMP4/BMPRI signaling promotes oligodendroglial differentiation and myelination in vitro
The in vivo data suggest that LDN-193189 is exerting its effects selectively on OPC differentiation to promote remyelination. To further establish whether LDN-193189 mediates its promyelinating effect directly on oligodendroglia, we used in vitro OPC monocultures and myelinating cocultures to examine the effect of BMP4 and LDN-193189 on OPC differentiation and myelination, respectively. To assess differentiation, isolated primary mouse OPCs were exposed to T3 to initiate differentiation, in the presence of either LDN-193189, BMP4, both (LDN BMP4, with BMP4 being added after 30 min after LDN-193189), or vehicle for 72 h ( ). The majority (∼70%) of vehicle-treated OPCs differentiated into MBP mature oligodendrocytes [ (quantified in B , E )], characterized by a flat morphology as the cells extended their developing myelin sheath across the 2D surface of the coverslip. This contrasted with the immature phenotype, where the processes of differentiating oligodendrocytes have extended, but have not begun spreading out and fusing. Concordant with previous studies ( ; ), OPCs treated with BMP4 primarily (∼70%) differentiated into GFAP astrocytes compared with vehicle control OPC cultures [ (quantified in B , C ); p < 0.0001 ]. While LDN-193189 treatment alone did not significantly influence OPC differentiation at the basal level, it significantly blocked the astrogliogenic effect that BMP4 exerted on the OPC cultures, as evidenced by significantly more oligodendrocytes (both immature and mature phenotypes) in LDN plus BMP4-treated cultures than BMP4 alone cultures [ (quantified in B , D , E ); p < 0.0001 ]. These data demonstrate that blocking BMPRI signaling in OPCs reduces the astrogliogenic effect of BMP4 and promotes the differentiation of OPCs into mature oligodendrocytes, suggesting that BMP4 signals via BMPRI receptors in OPCs to exert an inhibitory effect on oligodendrocyte differentiation.
BMP4 signals via BMPR1 in OPCs to enhance oligodendrocyte differentiation and reduce astrogliogenesis in vitro. A , Representative micrographs of immunostaining of differentiated OPC cultures for MBP and GFAP under untreated (Control) conditions, or following treatment with BMP4, LDN-193189 (LDN), or both BMP4 and LDN-193189 (LDN BMP4). B , Quantification of cell phenotypic distribution for each condition based on GFAP expression and MBP morphology. MBP cells were classified as either mature (flattening of branched extracellular membrane) or immature (branched morphology but not fused layers). C , Quantification of the proportion of GFAP cells in the cultures. BMP4 significantly increased the proportion of GFAP cells compared with untreated (Control) cultures. While LDN-193189 (LDN) alone exerted no significant effect, pretreatment with LDN before BMP4 (LDN BMP4) significantly abrogated effect of BMP4 on astrocytes. D , Quantification of the proportion of immature oligodendrocytes in the cultures. Treatment with BMP4 or LDN-193189 (LDN) exerted no significant effect, whereas pretreatment with LDN-193189 before BMP4 (LDN BMP4) significantly increased the proportion of immature oligodendrocytes. E , Quantification of the proportion of mature oligodendrocytes in the cultures. Treatment with BMP4 significantly blocked OPC differentiation, whereas LDN-193189 (LDN) alone exerted no significant effect. Pretreatment with LDN before BMP4 (LDN BMP4) significantly abrogated the effect of BMP4 on oligodendrocyte differentiation ( N = 4 animals/group). **** p < 0.0001. Scale bar, 20 µm.
We next assessed whether the effect that LDN-193189 exerts on potentiating OL differentiation also enhances myelination using the well established DRG neuron/OPC myelinating coculture assay ( ; ). Consistent with a previous report ( ), there is significantly fewer MBP myelinated axonal segments (approximately threefold reduction) in exogenous BMP4-treated cocultures compared with vehicle-treated control cocultures ( ; p < 0.0001 ), suggesting that BMP4 inhibits myelination in vitro . Importantly, this BMP4-induced inhibitory effect on myelination is blocked by pretreatment with LDN-193189 before BMP4 exposure ( ; p < 0.0001 ), suggesting that BMP4 signals via BMPRI to exert this inhibitory effect. Interestingly, LDN-193189 treatment alone also resulted in a significant increase in the number of myelinated segments compared with baseline vehicle controls ( ; p = 0.019 ), suggesting there is some endogenous BMP4 present in the cocultures. Together, our results suggest that BMP4 signals via BMPRI in OPCs to inhibit their differentiation into mature OL and subsequent myelination.
BMP4 signals via BMPR1 in OPCs to promote myelin formation in vitro . A , Representative micrographs of myelinating DRG/OPC cocultures treated with vehicle (control), BMP4, LDN-193189 (LDN), or
both LDN-193189 and BMP4 (LDN+BMP4) for 14 d and immunostained for MBP and Neurofilament. Arrows indicate MBP myelin segments colabeled with NFL axons. B , Quantification of the number of MBP myelinated axonal segments per field from these cocultures. BMP4 treatment significantly reduced the number of MBP myelin segments compared with cocultures, which is blocked by the pretreatment of LDN-193189 (LDN+BMP4; N = 4 independent cocultures/group). * p < 0.05, **** p < 0.0001. Scale bar, 30 µm.
### Inhibiting BMP4/BMPRI signaling in OPCs alters the expression of the transcriptional repressor Id4
Previous research strongly suggests that BMP4 inhibits the differentiation of oligodendrocyte-lineage cells by upregulating Id4, a transcription factor that inhibits OL differentiation ( ). To understand whether the effect observed on OPC differentiation and myelination was mediated, at least partially, by Id4, we used qRT-PCR to examine changes in transcription levels of Id4 as well as Gfap , Mbp , and myelin regulatory factor ( Myrf ) in OPCs treated with BMP4 and/or LDN-193189. To do this, we repeated the differentiation assay in OPC monocultures in the presence or absence of LDN-193189 and BMP4 over various time points ( ). We found there was a significant increase in the level of Id4 transcription in BMP4-treated OPCs compared with control untreated cultures at 2 h (approximately fivefold; ; p < 0.0001 ), which peaked at 24 h (approximately sixfold; ; p < 0.0001 ). Interestingly, this BMP4-induced increase in Id4 transcription is abolished by pretreatment with LDN-193189 at both the 2 and 24 h time points [ ; p = 0.0014 (2 h); p < 0.0001 (24 h)]. BMP4 treatment also led to a significant increase in Gfap transcription at 24 h ( ; p = 0.014 ), which was attenuated by the pretreatment with LDN-193189, but not significantly ( ; p = 0.094 ). Further, BMP4 treatment significantly reduced the expression of Mbp and Myrf transcripts at the 24 h mark compared with vehicle treated cultures [ ; p = 0.016 ( Mbp ), p = 0.0053 ( Myrf )]. Collectively, these data suggest that BMP4 signals to BMPRI in OPCs to upregulate Id4 , coinciding with an increase in Gfap transcription and downregulation of myelin genes Mbp and Myrf .
Inhibiting BMP4/BMPRI signaling in OPCs alters the expression of transcription factor Id4 and Gfap , but not Mbp or Myrf . A , B , qRT-PCR analysis of Id4 and Gfap transcript levels from OPCs cultured in differentiation media and treated with LDN-193189, BMP4, or both, or vehicle (control) over various time points. *BMP4 significantly increased the level of Id4 transcripts at 2 and 24 h compared with the control, and this upregulation is blocked by pretreatment with LDN-193189 before BMP4 exposure. * Gfap expression was also significantly reduced by pretreatment of OPCs with LDN-193189 before BMP4 exposure. C , D , qRT-PCR analysis of myelin protein gene Mbp and key myelination transcription factor Myrf from OPCs treated with LDN-193189, BMP4, or both, or vehicle over 24 h. BMP4 significantly reduced the expression level of both Mbp and Myrf genes, with this effect reduced by LDN-193189 pretreatment; N = 3 independent cultures/group). * p < 0.05, ** p < 0.01.
We further explored the downstream transcriptional effects that BMP4 and LDN-193189 exerted on OPCs using the RT2 PCR Profiler Array Kit measuring the transcription of 84 genes related to the TGF-β/BMP signaling family. To address this, OPC monocultures were treated with either LDN-193189, BMP4, both (LDN BMP4) or vehicle and allowed to differentiate for 24 h before RNA analysis. We compared changes in transcription within the following three comparisons: (1) control OPCs versus BMP4-treated OPCs; (2) control OPCs versus LDN-193189-treated OPCs; and (3) BMP4-treated OPCs versus LDN-193189 BMP4-treated OPCs. A summary of genes with a significant fold regulation of greater than two is presented in . We found that BMP4-treated OPCs significantly increased the transcription of several TGF-β target genes, as well as Id1 and Id2 . Interestingly, the genes of several BMP signaling regulatory proteins such as noggin, BAMBI, and BMP binding endothelial regulator (BMPER) were also upregulated, suggesting the possibility that exogenous BMP4 treatment of OPCs also activates intrinsic self-feedback mechanisms to modify the levels of BMP4 signaling. The transcription of Bmp4 itself was downregulated by exogenous BMP4 treatment in OPCs. Interestingly, BMP4 treatment significantly upregulates Smad1 but not Smad5 ; this is reversed in OPC cultures pretreated with LDN-193189 before BMP4 exposure. Smad2 , which is not typically used by BMP4 ( ), was also downregulated, suggesting that Smad5 may also be similarly unused by BMP4 in OPCs.
Summary of differentially regulated BMP/TGF-β signaling pathway genes in OPCs cultured in LDN-193189, BMP4, or both, or vehicle for 24 h in differentiating conditions
Furthermore, we found that OPCs treated with LDN-193189 significantly downregulated Id1 and Id2 , as well as levels of the BMP antagonist noggin. Pretreatment of OPCs with LDN-193189 before BMP4 exposure reversed the transcriptional levels of several genes differentially regulated by BMP4 treatment, including Bmper , Bambi , and Emp1. Levels of Id1 and Id2 were not significantly downregulated as a result of LDN-193189 pretreatment, in contrast with decreased Id4 transcription in OPC cultures pretreated with LDN-193189 before BMP4 exposure (identified by an individual Id4 qRT-PCR). Together, the data suggest that BMP4 inhibits OPC differentiation and their subsequent capacity to myelinate axons via signaling through BMPR1 and regulating an array of downstream signaling molecules and transcription factors in OPCs.
### Deleting OPC-expressed BMPRIA receptors promotes differentiation and myelination in vitro
LDN-193189 is known to disrupt BMP4 signaling by inhibiting both BMPRIA and BMPRIB, and, while mouse OPCs express both BMPRIA and BMPRIB, BMPRIA is expressed at a substantially higher level than BMPRIB ( ). Thus, it remains unclear whether the aforementioned effect of LDN-193189 on OPC differentiation and myelination are mediated via BMPRIA, BMPRIB, or both. Further, it also remained possible that BMP4 signaling in neurons may influence myelination in the coculture setting. To unequivocally determine whether BMP4 selectively signals to BMPRIA in OPCs to regulate their differentiation and myelination, we adopted a genetic approach, specifically deleting BMPRIA from OPCs. The BMPRIA KO mice are embryonic lethal: thus, we generated Pdgfra-CreER :: Bmpr1a conditional KO mice, allowing 4OHT-dependent Bmpr1a deletion in Pdgfra- expressing OPCs. We first confirmed 4OHT-mediated knockout of Bmpr1a in primary OPCs using PCR. OPCs were isolated from Pdgfra-CreER :: Bmpr1a (Cre[+]) and Bmpr1a control (Cre[−]) mice, treated with 4OHT followed by RNA extraction. PCR analysis confirmed the expression of Cre-recombinase in the 4OHT-treated cells, as well as deletion of exon 2 of the Bmpr1a sequence ( , ΔBMPRIa panel), while Bmpr1b transcription was unaffected. Analysis of 18S confirmed similar levels of RNA were analyzed ( , 18s panel).
BMP4 signals via BMPR1A in OPCs to potentiate oligodendrocyte differentiation and reduce astrogliogenesis in vitro. A , Representative micrographs of immunostaining of differentiated OPC cultures (isolated from Pdgfra-CreER :: Bmpr1a mice) for MBP and GFAP under untreated (Control) conditions, or following treatment with BMP4, 4OHT, or both BMP4 and 4OHT ( 4OHT BMP4). B , Quantification of cell-phenotypic distribution for each condition based on GFAP expression and MBP morphology, as described above ( ). C , Quantification of the proportion of GFAP cells in the cultures. BMP4 significantly increased the proportion of GFAP cells compared with untreated (Control) cultures, whereas 4OHT alone exerted no significant effect. Pretreatment with 4OHT before BMP4 ( 4OHT BMP4) significantly attenuated the effect of BMP4. D , Quantification of the proportion of immature oligodendrocytes in the cultures. Treatment with BMP4, 4OHT, or both BMP4 and 4OHT ( 4OHT BMP4) exerted no significant effect. E , Quantification of the proportion of mature oligodendrocytes in the cultures. Treatment with BMP4 significantly decreased OPC differentiation, whereas 4OHT alone exerted no significant effect. Pretreatment with 4OHT before BMP4 ( 4OHT BMP4) significantly attenuated the inhibitory effect of BMP4 on OPC differentiation. F , PCR analysis of 4OHT-treated OPCs to assess Bmpr1a knockout. Pdgfra-CreER :: Bmpr1a and Cre[−] OPCs were isolated and treated with 4OHT for 24 h and analyzed for the transcription of a sequence corresponding to Bmpr1a - ex2 , rendering the resulting protein untranscribable; N = 4 animals/group). * p < 0.05, **** p < 0.0001. Scale bar, 20 µm.
To investigate the effect that BMPRIA signaling exerts on OPC differentiation, cells were isolated from Cre[+] and Cre[−] control mice and exposed to 4OHT or vehicle for 24 h, followed by a 72 h differentiation assay in the presence or absence of BMP4. Cultures were assessed for the proportion of postmitotic OLs and astrocytes via immunostaining for MBP and GFAP, respectively ( ). Consistent with previous results ( ), in the control condition, the majority (>60%) of OPCs differentiated into mature myelinating OLs after 72 h at the basal level [ (quantified in B , D , E )]. As expected, exogenous BMP4 significantly inhibited OPC differentiation compared with the vehicle control, with the vast majority (∼80%) of cells being GFAP astrocytes in BMP4 alone-treated cultures after 72 h [ (quantified in B , C ); p < 0.0001 ]. Treatment with 4OHT exerted no effect on the proportion of OLs [ (quantified in B , D , E ); p = 0.711 ] or astrocytes [ (quantified in B , C ); p > 0.999 ], but, importantly, it resulted in significantly more OL differentiation and less astrogliogenesis following BMP4 treatment ( ; p = 0.011 ), potentiating astrogliosis ( ; p < 0.0001 ). These results collectively suggest that BMP4 signals via BMPRIA within OPCs to inhibit their differentiation.
To investigate whether BMPRIA also mediates the subsequent capacity to myelinate axons, we repeated the myelinating cocultures containing OPCs isolated from Cre[+] and Cre[−] control mice. Cocultures were exposed to 4OHT or vehicle for 24 h and maintained for 14 d followed by immunocytochemical and biochemical analyses of myelination in vitro. We found that, in cocultures containing BMPRIA-null OPCs (isolated from Cre[+] mice), 4OHT treatment resulted in significantly more MBP myelinated axonal segments compared with vehicle-treated control cultures ( ; p = 0.0098 ). Concordant with this, Western blot analysis of myelin proteins MBP and MOG shows that there was qualitatively more myelin protein expression in 4OHT-treated cocultures compared with vehicle controls ( ). In contrast, 4OHT exerted no effect on myelin formation in cocultures containing OPCs from Bmpr1a control (Cre[−]) mice ( ; p = 0.92 ). Together, our data suggest that selectively blocking BMP4 signaling in OPCs through ablating BMPRIA promotes oligodendroglial differentiation, reduces astrogliogenesis, and leads to a greater extent of myelination in vitro , indicating that BMP4 selectively signals via BMPRIA in OPCs to block oligodendroglial differentiation and myelination.
BMP4 signals via BMPR1A in OPCs to promote myelination in vitro . A , Representative micrographs of immunostaining for MBP and Neurofilament (NFL) in myelinating cocultures containing OPCs isolated from Pdgfra-CreER :: Bmpr1a mice. The cocultures were treated with or without 4OHT for 24 h before 14 d of myelination. Arrows indicate MBP myelinated axons segments colabeled with NFL axons. B , Quantification of MBP myelinated axonal segments from these cocultures. 4OHT-induced BMPRIA ablation in OPCs causes significantly more MBP myelinated axonal segments compared with controls. C , Western blot analysis of BMPRIA and myelin proteins (MOG and MBP) in sister cocultures from A and B , treated with either 4OHT or vehicle. Treatment with 4OHT substantially reduced BMPRIA expression and leads to qualitatively more myelin proteins (MBP and MOG) expression compared with controls. D , Representative micrographs of immunostaining for MBP and NFL in myelinating cocultures containing OPCs isolated from Bmpr1a (Cre control) mice. The cocultures were treated with or without 4OHT for 24 h before 14 d of myelination. Arrows indicate MBP myelin segments colabeled with NFL axons. E , Quantification of MBP myelinated axonal segments from cocultures. Treatment with 4OHT did not exert a significant effect on myelination in the Cre cocultures ( N = 3 independent cultures/treatment group). ** p < 0.01. Scale bar, 30 µm.
## Discussion
Identifying the mechanisms that inhibit OL differentiation and remyelination is crucial for developing future strategies that directly target myelin repair in MS. Here we have identified that inhibiting BMP4/BMPRI signaling following cuprizone-induced central demyelination significantly enhances oligodendroglial differentiation and promotes myelin repair in the brain in vivo . We have further determined that BMP4 signals to BMPRIA receptors in OPCs to inhibit OL differentiation and myelination in vitro . Together, the results of this study identify that inhibiting BMP4/BMPRIA signaling within OPCs promotes CNS remyelination via potentiating OL differentiation, and that blocking this pathway within OPCs is a potential strategy to enhance remyelination.
### Disrupting BMP4/BMPR1 signaling promotes remyelination via potentiating oligodendrocyte differentiation in vivo
The results of this study strongly support a role for blocking BMP4/BMPR1 receptor signaling in promoting CNS remyelination. BMP4/BMPRI signaling is upregulated during the remyelinating phase after myelin injury ( ), and blocking BMP signaling via noggin significantly enhances remyelination following demyelination in vivo ( ). While these studies firmly identify BMP signaling as refractory to remyelination, the fact that noggin promiscuously inhibits multiple BMPs, and thus signaling through several receptor classes, ultimately means that the molecular mechanisms mediating this effect remain to be elucidated. In this study, we took advantage of pharmacological developments in small-molecule inhibitors of the TGF-β signaling pathway and adopted an approach more specific to BMP4/BMPRI signaling ( ). LDN-193189 primarily inhibits BMPRIA and BMPRIB, with some inhibition of ACVRL1 (ALK1) and ACVR1 (ALK2) demonstrated in C2C12 osteoblast and chondroblast cell lines ( ). The mechanism of inhibition involves competitive binding of the compound to the kinase domain of the type I subunits, preventing phosphorylation of downstream SMAD molecules and restricting the signaling cascade ( ). Concordant with previous studies ( ; Karni A, Amir Levi Y, Urshansky N, Bernadet-Fainberg K (2013) World Intellectual Property Organization international patent application, PCT/IL2013/050503), here we have shown that inhibiting BMP4/BMPRI signaling with LDN-193189 significantly increased remyelination after central demyelination. This beneficial effect is achieved via selectively promoting OL differentiation, as evidenced by significantly more mature OLs after LDN-193189 administration, whereas the number of other glial cells such as astrocytes and microglia remained unchanged. This is also supported by the analysis of cultured primary OPCs in which LDN-193189 significantly potentiated OL differentiation and their subsequent myelination, and, importantly, blocked the astrogliogenic effect of BMP4 on OPCs.
It is interesting that LDN-193189 did not exert any significant effect on astrocytes during remyelination in vivo, whereas in our previous studies noggin infusion significantly inhibited the proliferation of GFAP astrocytes ( ; ). One consideration regarding the different astroglial effect is likely the timing of infusion. In this study, LDN-193189 was administered following a 5 week cuprizone challenge (i.e., the first week after cuprizone was withdrawn) to assess its effect on early myelin repair. However, in our previous studies, noggin was infused into the murine corpus callosum during the final third of a 6 week cuprizone challenge, when there is ongoing demyelination ( ; ). Thus, the role of BMP4 in relation to astrocytes may be proliferative in the context of acute CNS injury and be more apparent earlier in the course of disease. Potentially, astrocyte proliferation and gliosis may be modified by inhibiting BMP4 signaling activity at specific stages during demyelination and remyelination. Collectively, the results of this study, together with our previously published work, indicate that the major role of BMP4 is in promoting astrogliogenesis/astrocyte proliferation when there is active demyelination, but that it has relatively little effect on astrogliosis during remyelination following CNS injury. Additionally, the differential effects of noggin and LDN-193189 on OPCs may be due to the broader inhibitory effect of noggin. During remyelination, we found that LDN-193189 inhibition of BMP4/BMPRI/SMAD signaling selectively promotes OPC differentiation but has no effect on the generation of astrocytes. In contrast, studies using noggin to inhibit the generation of astrocytes may be achieving this through by inhibiting other BMP signaling pathways. This also is supported by in vitro evidence, in which noggin inhibits astroglial production in vitro ( ), whereas in this study, we found that LDN-193189 or deleting BMPRIA receptor exerted little effect on astrocytes in OPC cultures where exogenous BMP4 is absent (although this may also be due to subtleties in culturing conditions). Our results together with previous data suggest that the influence of BMP4 signaling effects on OPCs is context dependent, promoting astrogliogenesis when there is active demyelination while inhibiting the differentiation of OPCs during remyelination following CNS injury.
### BMP4 signals via BMPRIA in OPCs to inhibit oligodendrocyte differentiation and myelination
Consistent with previous studies ( ; ), we found that exogenous BMP4 promoted the majority of OPCs to differentiate into GFAP -expressing astrocytes, while inhibiting BMP4/BMPRI signaling using LDN-193189 before BMP4 exposure is sufficient to block this effect and enhance myelination in vitro . Transcriptional analysis of OPCs revealed that LDN-193189 significantly downregulated the expression of Id family genes including Id4 , which strongly inhibits oligodendrocyte differentiation in vitro ( ). The resulting culture environment was such that astrogliogenesis was mostly inhibited, but residual BMP4 signaling activity prevented full differentiation of OPCs into mature OLs. It is speculated this may be due to the following two separate mechanisms: an Id4-mediated sequestering of OL transcription factor OLIG2; and synergy of BMP4-activated SMADs with the astrogliogenic pathway JaK–STAT (Janus kinase–signal transducer and activator of transcription). The action of LDN-193189 likely affects both pathways, as BMP4-induced phosphorylation of SMADs occurs upstream of both mechanisms. Different minimum thresholds of SMAD activation for each mechanism may mean that LDN-193189 has varying efficacy for inhibiting the separate effects of BMP4 signaling in OPCs. Here, we note again that in vivo we did not observe decreased GFAP immunostaining in mice infused with LDN-193189 following cuprizone challenge in the corpus callosum compared with vehicle-infused mice. Thus, the environmental context in which OPCs are interacting with BMP4 likely influences the specific mechanism of action directing differentiation of these cells. Notably, BMP4 treatment in vitro exerted a marked inhibitory effect on the expression of MBP proteins, while a relatively less robust effect was observed on MBP transcription. The precise reason behind this relatively different transcriptional and translational regulation of MBP is unclear, but suggests that BMP4 signaling exerts greater influences that target translational regulation of MBP expression. Gene function is ultimately determined by the level of protein expression. In our study, the strong effect that BMP4 exerts on suppressing MBP expression is consistent with its marked influence on inhibiting the differentiation of OPCs into mature oligodendrocytes.
Data obtained from the myelinating cocultures was in accordance with that obtained from the OPC monocultures, with BMP4 decreasing and LDN-193189 increasing myelination, respectively. Importantly, LDN-193189 blocked the inhibitory effect of BMP4 on myelination in vitro ( ). One interesting observation was the significantly higher number of MBP myelinated axonal segments in cocultures treated with LDN-193189 compared with controls. This is likely due to the increased levels of endogenous BMP4 expressed by neurons and OPCs in the coculture setting. Indeed, OPCs themselves express a high level of BMP4 as they begin to differentiate ( ).
Historically, related but individual roles for BMPRIA and BMPRIB have been well identified in the regulation of various aspects of chondrogenesis and osteogenesis ( ). Precisely understanding the differential influences that BMPRIA and BMPRIB receptor signaling exerts in the context of oligodendrocyte differentiation is a key step toward identifying the most suitable therapeutic targets for promoting myelination and remyelination. However, the effects of BMP4 signaling on oligodendrocyte differentiation have been inconsistent in the field, largely due to the complexity in the nature of BMP4 signaling, the genetic tools being used to target mixed cell lineages, and a diverse range of the ages and regions of the animals being analyzed. Given that global genetic knockout of BMP4 and its receptors is embryonic lethal ( ; ), conditional genetic ablation driven by the expression of lineage markers offers a more nuanced approach to understanding BMP4 signaling in oligodendrocyte development. Previous to this study, used Cre- loxP -mediated transgenic excision of the Bmpr1a gene from cells expressing BRN4, a broad neural transcription factor activated in early embryogenesis. This was crossed with a conventional Bmpr1b KO mouse to generate mice with a Bmpr1a-Bmpr1b double KO in the neural tube by embryonic day 10 (E10.5). This leads to loss of BMPRIA/BMPRIB function in all subsequent spinal cord and hindbrain cells, causing several developmental defects and lethality at P0. Cultures of Bmpr1a-Bmpr1b double KO OPCs did not display phospho-SMAD1/5/8 immunoreactivity when treated with 50 ng/ml BMP4, suggesting a complete loss of the SMAD-dependent BMP4 signaling pathway in these mice. While the number of astrocytes in the spinal cord decreased at P0 compared with controls, disrupted BMP4 signaling through BMPRIA/B does not appear to affect the total number of spinal cord OPCs. Intriguingly, while the number of immature O4 oligodendrocytes was unchanged, the number of mature oligodendrocytes expressing common myelin proteins, including MBP, was reduced at P0. Counterintuitively, this suggests that some level of BMP4 signaling through BMPRIA/B is required for oligodendrocyte maturation in the spinal cord and hindbrain ( ), either through a direct effect or in combination with other synergistic pathways regulating oligodendrocyte development. Importantly, the lack of BMP4 signaling did not appear to affect the number of OPCs specified, conflicting with previous research indicating an inhibitory effect on OPC specification from neural stem cells in vitro ( ) and in overexpression studies in vivo ( ).
A further study by , deleted BMPRIA only from neural precursor cells expressing OLIG1 from E13.5 in the neural tube, which can differentiate into neurons, astrocytes, or oligodendrocytes. This did not affect the subsequent number of OPCs at birth or at P20 ( ). However, at P20, there was an increase in mature oligodendrocytes in the BMPRIA KO group; this was at odds with the previous study, where mature oligodendrocytes were reduced at the much earlier time point. This study did not discount the possibility of increased compensatory signaling through BMPRIB, as phospho-SMADs 1, 5, and 8 were still detected. A third study by deleted Bmpr1a in Emx-1-Cre -expressing neural stem cells (NSCs) of the murine telencephalon. These cells develop into neurons, astrocytes, and oligodendrocytes in the telencephalon, with Cre recombination occurring at the peak of neurogenesis but preceding gliogenesis in the mouse. It was found that subsequent astrocytes derived from these NSCs aberrantly expressed vascular endothelial growth factor at P10, leading to the disruption of cerebrovascular angiogenesis as well as impaired blood–brain barrier formation ( ). Interestingly, while previous studies using Olig1 - Cre -driven Bmpr1a deletion showed increases in mature O4 oligodendrocytes at P20, no differences in O4 cells were observed at P20 in this study. In addition, compared with the earlier study deleting both Bmpr1a and Bmpr1b from BRN4-expressing cells in which GFAP astrocytes are reduced, no such decreases were observed here.
In summary, embryonic overexpression of BMP4 before or during gliogenesis clearly decreases subsequent oligodendrogliogenesis; the inhibition of BMP4 signaling embryonically using noggin has the opposite effect and increases the number of oligodendrocytes. However, See et al., demonstrated that inhibiting BMP4–SMAD signaling by deleting BMPRIA/B before OPC specification reduces the number of mature oligodendrocytes at P0 ( ). Importantly, this was not due to reduction in the number of OPCs specified, as no changes in the number of OPCs were detected. Additionally, the study by found that reduction, but not complete suppression, of BMP4 signaling through BMPRIA deletion in E13.5 neural precursor cells has no effect on the number of OPCs at P0. However, deleting BMPRIA at E13.5 increases mature oligodendrocyte number by P20. The reasons for this remain unclear. However, observations from all three studies suggest that BMP signaling through BMPRIA/BMPRIB does not play a role in the specification of OPCs from NSCs, but has a strong negative effect on subsequent OPC differentiation ( ). Only one study specifically targeted oligodendrocyte lineage cells using an Olig1 - Cre driver; however, this targets all oligodendrocytes as well as some neuronal populations. Before our study presented here, the effect of inhibiting BMPRIA in postnatal, lineage-committed OPCs had not been examined.
Using a conditional and inducible transgenic approach to specifically ablate BMPRIA expression in OPCs, we have identified that BMPRIA has a critical role in mediating the inhibitory BMP4 signal in OPC lineage progression within the postnatal CNS. We used the Pdgfra-CreER driver of Cre expression to specifically ablate the expression of Bmpr1a in postnatally derived OPCs, rather than in neural progenitor cells, or in all oligodendrocyte lineage cells, as seen with the more commonly used Olig2-Cre driver. As PDGFRα is downregulated in OPCs before differentiation ( ; ), BMPRIA expression is ablated before the differentiation process occurring. This allowed us to examine the influence of BMPRIA specifically on this process, in the absence of any confounding effects of coincident deletion in mature oligodendrocytes. We found that OPCs with a BMPRIA deletion significantly attenuated the inhibitory effect of BMP4 on OPC differentiation into mature oligodendrocytes, as seen in OPC cultures treated with LDN-193189. Moreover, we found that 4OHT-treated myelinating cocultures containing BMPRIA KO OPCs showed an increased capacity to myelinate, suggesting that inhibiting BMP4/BMPRI signaling in OPCs promotes the basal level of myelination. Noticeably, the magnitude of the effect of disrupting BMPRIA expression in OPCs was lower than that seen in experiments where the signaling of BMPRI receptors is inhibited pharmacologically using LDN-193189, both at the transcriptional and protein level. For instance, the astrogliogenic effect of BMP4 treatment (as measured by the differentiation of OPCs into GFAP-expressing astrocytes) was ∼25% less in BMPRIA KO OPCs compared with control cultures, in contrast to a near-total reduction in the LDN-193189-treated OPCs. Similarly, a greater number of OPCs differentiated into either immature or mature oligodendrocytes in the LDN-193189-treated OPCs compared with the BMPRIA-null OPCs. This differential effect may be due to the latency of the turnover and replacement of functional BMPRIA receptors. The rate of BMP receptor turnover is governed by either clathrin-dependent or caveolin-dependent endocytosis, depending on whether the BMP ligand initially binds to the type I subunit or to a preformed complex of type I/type II subunits ( ). Second, it is possible that the Pdgfra-CreER Cre driver used did not generate a full knockout of Bmpr1a . We observed residual BMPRIA protein expression in 4OHT-treated DRG/OPC cocultures using Western blotting (although this may have been contributed by DRG neurons). The original study characterizing the Pdgfra-CreER Cre driver found ∼45–50% successful recombination of floxed DNA regions ( ). Thus, there is likely to be remaining BMPRIA expression on the OPC cell surface that may have attenuated the observed effect of inhibiting BMPRIA signaling on OPC differentiation. Further, LDN-193189 inhibits BMPRI receptors including BMPRIA and BMPRIB, whereas BMPRIB remains active in BMPRIA KO OPCs. This finding suggests that BMPRIB may also play a role in mediating the BMP4-induced inhibitory effect on oligodendrocyte differentiation and myelination in the postnatal CNS, which warrants future investigation. Both the use of LDN-193189 and OPC-targeted transgenic ablation of BMP receptor subunits may enhance the current state of knowledge regarding the role of BMP4 signaling on embryonic oligodendrocyte development, as detailed above.
In summary, our results show that inhibiting BMP4/BMPRI signaling in OPCs promotes remyelination following myelin injury in vivo . This beneficial effect is likely mediated by potentiating OPCs differentiation into mature myelinating oligodendrocytes. Further, we have identified that BMPRIA in OPCs plays a critical role in mediating the inhibitory effect of BMP4 on OPC differentiation and myelination. Together, our work presented here indicates that targeting BMP4/BMPRIA signaling in OPCs is a potential strategy for enhancing remyelination following a demyelinating insult.
|
Visual Abstract
Huntington’s disease (HD) patients suffer from a progressive neurodegeneration that results in cognitive, psychiatric, cardiovascular, and motor dysfunction. Disturbances in sleep/wake cycles are common among HD patients with reports of delayed sleep onset, frequent bedtime awakenings, and fatigue during the day. The heterozygous Q175 mouse model of HD has been shown to phenocopy many HD core symptoms including circadian dysfunctions. Because circadian dysfunction manifests early in the disease in both patients and mouse models, we sought to determine if early intervention that improve circadian rhythmicity can benefit HD and delay disease progression. We determined the effects of time-restricted feeding (TRF) on the Q175 mouse model. At six months of age, the animals were divided into two groups: ad libitum (ad lib) and TRF. The TRF-treated Q175 mice were exposed to a 6-h feeding/18-h fasting regimen that was designed to be aligned with the middle of the time when mice are normally active. After three months of treatment (when mice reached the early disease stage), the TRF-treated Q175 mice showed improvements in their locomotor activity rhythm and sleep awakening time. Furthermore, we found improved heart rate variability (HRV), suggesting that their autonomic nervous system dysfunction was improved. Importantly, treated Q175 mice exhibited improved motor performance compared to untreated Q175 controls, and the motor improvements were correlated with improved circadian output. Finally, we found that the expression of several HD-relevant markers was restored to WT levels in the striatum of the treated mice using NanoString gene expression assays.
## Significance Statement
Huntington’s disease (HD) is a genetically caused disease with no known cure. Lifestyle changes that not only improve the quality of life but also delay disease progression for HD patients are greatly needed. In this study, we found that time-restricted feeding (TRF) improves activity/rest rhythms in the Q175 mouse model of HD. This treatment also improved motor performance and heart rate variability (HRV) in the HD mice. Finally, TRF altered the expression of HD relevant markers in the striatum. Our study demonstrates the therapeutic potential of circadian-based treatment strategies in a preclinical model of HD.
## Introduction
Huntington’s disease (HD) is caused by an expanded CAG repeat within the first exon of the Huntingtin ( Htt ) gene. The mutated HTT protein leads to dysfunction of a large range of cellular processes, including cytoskeletal organization, metabolism, and transcriptional activities ( ; ; ). As result, HD patients suffer from progressive neurodegeneration that inflicts cognitive, psychiatric, cardiovascular, and motor dysfunction. The genetic components greatly determine the age of symptom onset and the severity. Generally, the longer the CAG repeat, the earlier the age of onset and the greater the severity of the symptoms ( ). Still, even among patients with the same CAG repeat length, large variabilities in the onset of symptoms (around a decade) and their severity have been reported ( ). In addition, studies have shown that environmental factors also affect the disease progression ( ). Those reports raise the possibility of environmental modifiers to the disease and suggest that lifestyle changes can increase the health span of the patients. This possibility is important to pursue as there are no known cures for HD.
Disturbances in the timing of sleep, typified by frequent bedtime awakenings, prolonged latency to fall asleep, and more naps during the awake phase, are extremely common in HD and often become apparent years before the onset of classic motor symptoms ( ; ; ). Similarly, mouse models of HD also exhibit a disrupted circadian rest/activity cycle that mimics the symptoms observed in human patients ( ; ; ). This body of work supports the hypothesis that circadian dysfunctions may interact with HD pathology and exacerbate the symptoms. To test this hypothesis, we have been using the Q175 knock-in model of HD. In previous work ( ), we have characterized the impact of age (3, 6, 9, and 12 months) and gene dosage (Het and Hom) on the degradation of circadian rhythms in locomotor activity and other HD core symptoms. Recently, a detailed RNA-seq analysis of striatum, cortex, and liver of the Q175 line has been published ( ); therefore, we have a good understanding of the transcriptional changes that occur with age in this model. Finally, recent work has carefully characterized age-related changes in the electroencephalogram (EEG) in both Hom and Het Q175 ( ). This wealth of data makes the Het Q175 an ideal preclinical model to examine the impact of circadian interventions on disease trajectory.
The central circadian clock responsible for the generation of daily rhythms is localized in the suprachiasmatic nucleus (SCN) in the hypothalamus. While lighting conditions are a critical environmental input to this timing system, a body of recent work has lead us to appreciate that the feed/fast cycle is also a powerful regulators of the circadian system ( ). While progressive, age-related SCN dysfunction has been reported in HD mouse models ( ), a time-restricted feeding (TRF) regimen promises therapeutic potential and can benefit even SCN-lesioned mice ( ; ). For example, mice under TRF consume equivalent calories from a high-fat diet as those with ad libitum (ad lib) access yet are protected against obesity, hyperinsulinemia, and inflammation and have improved motor coordination ( ). In the present study, we examined the impact of imposing a 6-h feeding/18-h fasting regimen that was aligned to the middle [zeitgeber time (ZT) 15-21] of the period when mice normally active (ZT 12-24). The treatment was applied to Q175 Hets starting when the mutants were six months of age and ending when they were nine months. We selected this age range because the Het Q175 start to show disrupted sleep/wake cycles and motor symptoms are just beginning.
## Materials and Methods
The work presented in this study followed all guidelines and regulations of the UCLA Division of Animal Medicine that are consistent with the Animal Welfare Policy Statements and the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association.
### Animals
The Q175 mice used in this study were males on the C57BL6/J background. They arose from a spontaneous expansion of the CAG repeat in the CAG140 transgenic knock-in line ( ). The mice were heterozygous (Het) for the Q175 allele with an average of 189 ± 3 CAG repeats. Mutant mice were obtained from The Jackson Laboratory from a colony managed by the CHDI Foundation. The animals were singly housed within light-tight chambers with independently controlled lighting conditions: 12 h of light followed by 12 h of dark (12/12 h LD). The chambers were in the same animal housing facility with controlled temperature and humidity, and each chamber held eight cages of mice, grouped together by feeding treatment. All animals received cotton nestlets, and water was made available at all times. To confirm the effect of timed feeding on daily rhythms and motor performance, we also examined WT mice at nine months of age.
### TRF
Mice were first entrained to a 12/12 h LD cycle for a minimum of two weeks before any treatment. Experimental animals were randomly assigned to one of two feeding conditions: food available ad lib and food available for 6 h during the middle of the active phase during ZT 15-21. By definition, ZT 12 referrers to when the lights go off when the mice are in an LD cycle. Experimental mice were singly housed in cages with a custom made programmable food hopper that could temporally control access to food (Diet Teklad 7013: fat, 18 kcal%; caloric density, 3.13 kcal/g) and prevent food consumption during restricted times. These cages were also equipped with an infrared (IR) motion detector to give us the ability to measure cage activity. The mice were held in these conditions for a total of three months (from six to nine months of age).
### Monitoring of cage locomotor activity
Experimental mice were singly housed in cages with the food hopper as well as IR motion sensors. The locomotor activity recorded as previously described ( ). Mice were entrained to a 12/12 h LD cycle for a minimum of two weeks before data collection. Locomotor activity data were recorded using Mini Mitter data loggers in 3-min bins, and 10 d of data were averaged for analysis. We used the 10 d of activity data collected just before the motor performance tests during the final two weeks of the TRF schedule. The data were analyzed to determine the period and rhythmic strength as previously described ( ; ). The periodogram analysis uses a χ test with a threshold of 0.001 significance, from which the amplitude of the periodicities is determined at the circadian harmonic to obtain the rhythm power. The amount of cage activity over a 24-h period was averaged over 10 d and reported here as the arbitrary units (a.u.)/h. The number of activity bouts and the average length of bouts were determined using Clocklab (Actimetrics), where each bout was counted when activity bouts were separated by a gap of 21 min (maximum gap: 21 min; threshold: 3 counts/min). The onset variability was determined using Clocklab by drawing the best-fit line over the 10 d, and averaging the differences between activity onset and best-fit regression of each day.
### Monitoring of immobility-defined sleep behavior
Immobility-defined sleep was determined as described previously ( ; ). Mice were housed in see-through plastic cages containing bedding (without the addition of nesting material) and the food hopper. A side-on view of each cage was obtained, with minimal occlusion by the food bin or water bottle, both of which were top-mounted. Cages were side-lit using IR-LED lights. Video capture was accomplished using surveillance cameras with visible light filters (Gadspot Inc) connected to a video-capture card (Adlink Technology Inc) on a Dell Optiplex computer system. ANY-maze software (Stoelting Co) was used to track the animals.
Immobility was detected when 95% of the area of the animal stayed immobile for >40 s, as was previously determined to have 99% correlation with simultaneous EEG/EMG-defined sleep ( ; ). Continuous tracking of the mice was performed for a minimum of five sleep-wake cycles, with randomized visits (one to two times per day) by the experimenter to confirm mouse health and video recording. The 3rd and 4th sleep-wake cycles were averaged for further analysis. Immobility-defined sleep data were exported in 1 min bins, and total sleep time was determined by summing the immobility durations in the rest phase (ZT 0-12) or active phase (ZT 12-24). An average wave form of hourly immobile-sleep over the two sleep-wake cycles was produced during the final week of TRF. Variability of awake time was determined using Clocklab to draw the best-fit line over the sleep cycles, and the differences between sleep offset and best-fit regression of each sleep cycle were averaged.
### Rotarod test. Accelerating version
The rotarod apparatus (Ugo Basile) is commonly used to measure motor coordination and balance. This apparatus is, in essence, a small circular treadmill. It consists of an axle or rod thick enough for a mouse to rest over the top of it when it is not in motion and a flat platform a short distance below the rod. The rod is covered with smooth rubber to provide traction while preventing the mice from clinging to the rod. In this study, mice were placed on top of the rubber covered rod. When the mice moved at the pace set by the rotation rate of the rod, they would stay on top of it. When mice no longer move at the selected pace they dropped a short distance to the platform below. The time a mouse remained on the rod, before dropping to the platform was called the latency to fall. Following a 15-min habituation to the testing room, mice were placed on the slowly rotating rod. The rod gradually accelerated from 5 rpm to 38 rpm over the course of the trial. The length of time the mouse stays on the rod was recorded. A two-day protocol for the accelerating rotarod tests was used. On the first day, the mice were trained on the rotarod over five trials. The maximum length of each trial was 600 s, and mice were allowed to rest for a minimum of 60 s between trials. On the second day, mice were tested on the rotarod and the latency to fall from the rotarod was recorded from five trials. Mice were again allowed to rest for a minimum of 60 s between trials. Data from each mouse were analyzed after averaging the times from all five trials. The apparatus was cleaned with 70% alcohol and allowed to dry completely between trials. A dim red-light (2 lux) was used for illumination during active phase testing (night).
### Challenging beam test
The challenging beam test is a modified version of the beam traversal test first described by Goldberg and colleagues ( ), and was used to characterize the motor deficits of Q175 mutant mice in previous studies ( , ). The beam narrows in four intervals from 33 mm > 24 mm > 18 mm > 6 mm, with each segment spanning 253 mm in length. Apparatus and methods used are similar to those described by Fleming and colleagues ( ). The home cage of each mouse is put on the end of the beam as the motivating factor. In this study, animals were trained on the beam for five consecutive trials on two consecutive days. During each trial, each mouse was placed on the widest end of the beam and allowed to cross with minimal handling by the experimenter. On the testing day, a metal grid (10 × 10-mm spacing, formed using 19-gauge wire) was overlaid on the beam. This overlaid grid increased the difficulty of the beam traversal task and provided a visual reference for foot slips made while crossing the grid. Each mouse was subjected to five consecutive trials, which were recorded by a camcorder under dim red-light conditions (2 lux), supplemented with IR lighting for video recording. The videos were scored post hoc by two independent observers for the number of missteps (errors) made by each mouse. The observers were masked as to the treatment group of the mice that they were scoring. An error was scored when any foot dipped below the grid. The number of errors was averaged across the five trials per mouse to give the final reported values. The apparatus was cleaned with 70% alcohol and allowed to dry completely between trials. A dim red-light (2 lux) was used for illumination during active phase testing (night).
### Automatic outputs. Core body temperature (CBT), heart rate (HR), and HR variability (HRV)
For the telemetry measurements, methods employed were similar to those previously described ( ; ). Two groups (ad lib and TRF) of Het Q175 mice ( n = 7/group) were surgically implanted with a wireless radio-frequency transmitter (ETA-F20, Data Sciences International). Mice were singly housed in cages with the food hopper. Cages were placed atop telemetry receivers (Data Sciences International) in a light and temperature-controlled chamber. Standard rodent chow was provided for both groups. Data collection began two weeks after surgery. HR was extrapolated from ECG waveforms using the RR interval.
Data collection and analysis were performed as described previously ( ). Data were extracted in 20-s intervals then filtered to remove extreme noise. Remaining valid data segments were averaged into 1-h bins across the 24-h cycle. Mean normal to normal intervals (NN, in ms) and SD of all NN intervals (SDNN, in ms) were calculated for the time domain analysis.
### NanoString analysis of gene expression
Tissue collection and data analysis were performed as described previously ( ). Four weeks after the final behavioral tests were performed, the Q175 mutants were anesthetized with isoflurane before dissection of the striatum at ZT 15. The brain tissue samples were flash frozen and stored at −80°C before NanoString analysis. The NanoString analysis was performed by LabCorp using a custom CodeSet designed to interrogate 100 transcripts previously implicated in transcriptional changes in the striatum of Q175 mice ( ). The signal intensity of individual genes was normalized by adjusting to internal positive standards within each sample. Eight housekeeping genes were included in the CodeSet: Gins1, Myh15, Pank2, Poc1b, Pum2, Slc25a15, Ssrp1, and Utp3 . The expression levels for each probe within a sample were scaled using the geometric mean of the eight housekeeping genes for each sample. Each mouse was an individual sample as tissue did not need to be pooled. The fold change of signal intensity was derived by comparing the normalized means between the ad lib group and the TRF group.
### Pathway analysis
To study the HD-changed gene expression data in the context of biological networks, the gene expression data of TRF-treated Q175 and untreated Q175 control samples were analyzed with the Ingenuity Pathway Analysis (IPA) system (Ingenuity Systems). Datasets containing gene identifiers and corresponding expression values were uploaded in the application. Each gene identifier was mapped to its corresponding gene object in Ingenuity Pathways Knowledge. A cutoff of corrected p value (i.e., q value = 0.005) was set to identify genes whose expression was significantly different as a result of the treatment. These genes were overlaid onto a global molecular network developed from information contained in the Ingenuity Pathways Knowledge Base. Functional analysis using the IPA program identified the biological functions that were most significant to the dataset (uncorrected Fisher’s exact test p < 0.05).
### Statistical analysis
We were interested in determining if TRF can delay the progression of symptoms in the Q175 mouse model; Therefore, treated Q175 mice (TRF group) were compared to age-matched untreated Q175 mice (ad lib group) in all experiments. The sample size per group was determined by both our empirical experience with the variability in the prior measures in the Q175 mice ( ) and a power analysis (SigmaPlot, SYSTAT Software) that assumed a power of 0.8 and an α of 0.05. For the behavioral measures, the analysis was done by two observes masked as to the experimental condition and their values averaged. To assess the impact of TRF after three months, we applied a t test for the analysis. To determine the impact of the treatment on temporal activity, sleep, CBT, HR, and HRV waveforms, we used a two-way repeated measures ANOVA (two-way RM ANOVA) with treatment and time as factors. To determine the impact of the treatment on errors made in each beam of the challenging beam test, we used a two-way ANOVA with treatment and beam # as factors. F values are reported as F (degrees of freedom between groups, degrees of freedom within groups). Pairwise multiple comparison procedures were made using the Holm–Sidak method. Correlations between circadian parameters and motor performance were examined by applying Pearson correlation analysis. Statistical analysis was performed using SigmaPlot. The dataset was examined for normality (Shapiro–Wilk test) and equal variance (Brown–Forsythe test). The power of the statistical tests is reported in . Between-group differences were determined significant if p < 0.05. All values are reported as group mean ± SEM.
List of distribution, statistical test, and power for each dataset analyzed in this study
## Results
By using the programmable food hopper, we could temporally control access to food (ZT 15-21) and prevent food consumption for the rest of the daily cycle. During this 6-h interval, the mice would eat as much as they wanted and the amount of food consumed daily did not vary between the Het Q175 groups (ad lib: 2.8 ± 0.4 g; TRF: 2.8 ± 0.2 g, t = −0.13, p = 0.900, t test ). At the time when we performed the recordings and motor assays, the body weights were not different in Q175 mice under TRF compared to age-matched controls (ad lib: 23.9 ± 0.4 g; TRF: 24.5 ± 0.4 g, t = −1.03, p = 0.320, t test ).
### TRF increased the amplitude of diurnal rhythms in Het Q175 line
At early disease stage (nine months of age), the TRF-treated group showed greatly improved circadian locomotor activity rhythms ( ), evidenced by the stronger rhythmic power (ad lib: 32.1 ± 2.2; TRF: 43.4 ± 2.9, t = −3.12, p = 0.008, t test ) and lower activity onset variability (ad lib: 27.3 ± 4.6 min; TRF: 15.8 ± 2.4 min, t = 2.2, p = 0.045, t test ) than the control group. The amount of cage activity was also increased under the TRF regimen (ad lib: 75.3 ± 5.9 a.u./h; TRF: 160.7 ± 21.1 a.u./h, t = 42, p = 0.005, t test ). These increases in rhythm power and activity amount coincided with a decreased total number of activity bouts (ad lib: 10. 8 ± 0.9; TRF: 7.9 ± 0.6, t = 2.6, p = 0.021, t test ). A temporal activity wave form indicated more robust activity levels in the TRF-treated group at night when the mice should be active ( ). A two-way RM ANOVA revealed a significant effect of time ( F = 70.07, p < 0.001), treatment ( F = 10.82, p = 0.005), and a significant interaction between the two factors ( F = 8.24, P < 0.001). A further examination of activity bouts at night (ZT 12-24) revealed that the TRF group had longer bout lengths (ad lib: 60.6 ± 17.5 min; TRF: 128.8 ± 27.8 min, t = 48, p = 0.038, t test ) without a significant increase in the number (ad lib: 7.6 ± 0.6; TRF: 5.4 ± 0.9, t = 2.15, p = 0.05, t test ), suggesting that the robust amplitude of diurnal rhythms in the TRF group was due to the consolidated and high amount of locomotor activity during the active phase ( ). Under TRF, the activity parameters in the Q175 mice were no longer significantly different from WT ( , ). These findings demonstrate that TRF treatment significantly improved the activity rhythms of the HD mutant mice.
Locomotor activity rhythms were improved by the TRF regimen. A , Examples of cage activity rhythms recorded from Q175 mutants under control (left) and TRF (right) conditions. The activity levels in the actograms were
normalized to the same scale (85% of the maximum of the most active individual). Each row represents two consecutive days, and the second day is repeated at the beginning of the next row. The orange bar on the top of actograms indicates the time when food hopper is opened. B , The strength of the activity rhythm is indicated by the power (%V) of the χ periodogram analysis. C , The averaged level of cage activity. D , The averaged variation in onset from the best-fit regression line. E , Average waveforms from 10 d of cage activity (1-h window) are shown and SEs across animals are indicated. F , The number of activity bouts (separated by a gap of 21 mins or more) during rest phase (ZT 0-12), active phase (ZT 12-24), and 24 h are reported as the level of fragmentation of the circadian activity cycle. Black bars represent Q175 mutants under ad lib condition, and orange bars represent Q175 mutants under timed feeding condition. G , The average length of activity bouts during their active phase. The white/black bar on the top of actograms ( A ) and waveforms ( E ) indicates the 12/12 h LD cycle. The temporal activity wave form was analyzed using a two-way RM ANOVA with time and treatment as factors. Other comparisons between Q175 cohorts were made using a t test. Asterisks represent significant differences due to TRF regimen compared to ad lib controls ( p < 0.05); n = 8/group.
Comparisons of age-matched WT under ad lib conditions to Q175 mice under ad lib or TRF regimen ( n = 8/group)
### TRF shifted the timing but not the total amount of sleep behavior in the Het Q175 mice
The immobility-defined sleep behavior was measured using video recording in combination with automated mouse tracking analysis software. During the 6 h when food was available at night, the TRF-treated Q175 mice slept less than untreated Q175 controls ( ). A two-way RM ANOVA was used to analyze the temporal pattern of sleep (1-h bins) of each group. The analysis revealed significant effect of time ( F = 36.575, p < 0.001) and significant interaction between time and treatment ( F = 2.23, p = 0.002), but the effect of treatment did not reach a significant level ( F = 2.033, p = 0.155). No significant changes were detected in the total amount of sleep time over a 24-h cycle (ad lib: 722.5 ± 25.6 min; TRF: 686.3 ± 28.4 min, t = 0.95, p = 0.36, t test ; ). No significant difference was found in the total number of sleep bouts over a 24-h cycle (ad lib: 8.2 ± 0.4; TRF: 9.3 ± 0.8, t = 58, p = 0.33, t test ). The sleep bouts at night were significantly shorter in the TRF group than the control group (ad lib: 160.4 ± 31.6; TRF: 65.5 ± 7.9, t = 93, p = 0.007, t test ), suggesting that TRF group had shorter naps than the control group in their active phase ( ).
TRF prevented disease-caused awakening time without altering the amount of sleep behavior. Video recording in combination with automated mouse tracking analysis software was used to measure immobility-defined sleep. A , Running averages (1-h window) of immobility-defined sleep in Q175 mutants with ad lib (black) and timed feeding (orange) are plotted. The white/black bar on the top of wave form indicates the 12/12 h LD cycle. B–F , Quantification of the immobility-defined sleep rhythms. The temporal sleep wave form was analyzed using a two-way RM ANOVA with time and treatment as factors. Other comparisons between Q175 cohorts were made using a t test. Asterisks represent significant differences due to TRF regimen compared to ad lib controls ( p < 0.05); n = 8/group.
The TRF treatment advanced the phase when the Q175 mice transitioned from sleep to awake states (ad lib: ZT 12.6 ± 0.2 h; TRF: ZT 11.9 ± 0.1 h, t = 3.84, p = 0.002, t test ; ). The TRF group also exhibited a more precise awakening time than the Q175 control mutants (ad lib: 37.7 ± 6.3 min; TRF: 19.3 ± 5.4 min, t = 2.21, p = 0.044, t test ; ). Under TRF, the beginning of activity and the cycle-to-cycle variability in sleep behavior in the Q175 mice were no longer significantly different from WT ( , ). Overall, these findings demonstrate that the TRF regimen improved sleep behavior in Q175 mice.
Comparisons of age-matched WT under ad lib to regimen ( n = 8/group)
### TRF improved autonomic outputs in the Het Q175 mice
It has been shown that dysfunction in the circadian regulation of autonomic outputs can be detected early in disease progression in the Q175 mice ( ). In the present study, we measured the impact of TRF on activity, CBT, HR, and HRV measured simultaneously in freely moving Q175 mice ( ). The TRF Q175 mice exhibited higher levels in activity, CBT, and HR at some phases of the daily cycle ( ). TRF also reduced the inappropriate activity during the daytime (ZT 0-12) when mice are normally less active (ad lib: 618.6 ± 96.6 a.u.; TRF: 308.4 ± 33.9 a.u., t = 3.03, p = 0.010, t test ). A two-way RM ANOVA was applied on the activity wave form and significant effects of time ( F = 21.86, p < 0.001), treatment ( F = 23.81, p < 0.001) and interaction ( F = 3.68, p < 0.001) were detected. In addition, the daily 24-h averaged CBT was not significantly different between the two groups (ad lib: 37.1 ± 0.1°C.; TRF: 36.7 ± 0.3°C, t = 3.03, p = 0.17, t test ). The TRF-treated group showed a lower CBT at the dark/light transition (ZT 23-2; ). A two-way RM ANOVA confirmed significant effects of time ( F = 28.64, p < 0.001) and treatment ( F = 7.65, p = 0.006) without an interaction between the two factors ( F = 1.05, p = 0.398). Despite no difference in the daily 24-h averaged HR (ad lib: 405. 9 ± 8.0 bpm; TRF: 424.1 ± 10.2, t = −1.4, p = 0.190, t test ), the amplitude of the rhythm (max/min ratio) was improved by the TRF regimen (ad lib: 1.5 ± 0.02 bpm; TRF: 1.6 ± 0.03 bpm, t = −2.18, p = 0.049, t test ; ). The TRF group exhibited higher HR (ZT 13-17) when the food was available. As measured by two-way ANOVA , significant effects of time ( F = 10.21, p < 0.001) and treatment ( F = 11.39, p < 0.001) were detected. But no interaction between the two factors ( F = 1.52, p = 0.06) was detected. Finally, the TRF-treated group exhibited higher levels in HRV in the rest phase as well as the beginning of active phase than the Q175 control group ( ). The TRF-treated Q175 mice had significantly higher 24-h averaged HRV than the control Q175 mice (ad lib: 13.7 ± 0.8 msec.; TRF: 17.0 ± 1.0 msec, t = −2.5, p = 0.028, t test ). A two-way RM ANOVA confirmed significant effect of time ( F = 8.23, p < 0.001) and treatment ( F = 39.6, p < 0.001) without a significant interaction ( F = 1.33, p = 0.140). Overall, the TRF regimen improved the daily rhythms in physiologic, autonomically-driven outputs.
Autonomic output rhythms were improved by the TRF regimen. The autonomic outputs from ad lib (black circles) and TRF (orange triangles) Q175 mice were recorded simultaneously using telemetry device. A–D , Hourly running averages of activity ( A ), CBT ( B ), HR ( C ), and HRV from both groups are plotted ( D ). E , The HR rhythm amplitude, determined by the ratio of max and min of the day, in control and TRF-treated Q175 mice. F , The 24-h averaged HRV in control and TRF-treated Q175 mice. The temporal waveforms of autonomic outputs were analyzed using a two-way RM ANOVA with time and treatment as factors. Other comparisons between Q175 cohorts were made using a t test. Asterisks represent significant differences due to TRF regimen compared to ad lib controls ( p < 0.05); n = 7/group.
### TRF improved motor performance in the Het Q175 mice
One of the defining symptoms of HD is the incidence of movement disorders in early-stage patients and we hypothesized that TRF may improve the motor symptoms. To test this hypothesis, we assessed motor performance using two tests that have been shown to detect motor coordination deficits in Q175 mice: the accelerating rotarod ( ) and challenging beam tests ( ). The Q175 mice on TRF had a longer latency to fall compared to age-matched Q175 ad lib-fed mutants (ad lib: 256 ± 30.4 min; TRF: 420.1 ± 32.2 min, t = −3.7, p = 0.002, t test ). In addition, the treated Q175 mice made significantly fewer errors compared to control Q175 mice (ad lib: 7.4 ± 0.5; TRF: 4.9 ± 0.5, t = 3.23, p = 0.006, t test ). Breaking down the errors made by beam width, the two-way ANOVA revealed a significant effect of treatment ( F = 15.22, p < 0.001), effect of beam width ( F = 26.17, p < 0.001), and interaction between the two factors ( F = 3.924, p = 0.013). Post hoc analysis indicates that the main difference between treated and control Q175 mice were the errors in the narrowest beam (ad lib: 3.4 ± 0.5; TRF: 1.8 ± 0.2, t = 4.84, p < 0.001, t test).
TRF improved motor performance in the Q175 HD model. A , The accelerating rotarod test revealed that the TRF treatment improved motor performance by showing longer latency to fall. B , The challenging beam motor test indicated that the TRF treatment improved performance (fewer errors) by making fewer errors when the mice crossed the balanced beam. C , The circadian parameters and the performance in the two motor tests of individual mouse in ad lib group (black circles) and TRF group (orange triangles) are plotted in a 3D-XYZ grid. In this XYZ grid, there are two distinctive clusters, suggesting that the mouse with stronger circadian rhythms performed better in both motor tests. Comparisons between Q175 cohorts were made using a t test. Asterisks represent significant differences due to TRF regimen compared to ad lib controls ( p < 0.05). The correlations between circadian parameters and motor performance are described in the text; n = 8/group.
The TRF-treated Q175 mice which showed the most improved circadian output also had better performance in the two motor tests ( ). In a XYZ grid composed of key activity rhythms parameters and performance of motor tests, there were two distinctive clusters which indicated that the mice with improved locomotor activity rhythm performed better in both motor tests. The correlation analysis indicated that the rhythmic power tended to be positively correlated with the amount of time staying on the accelerating rotarod (coefficient = 0.54, p = 0.17) and was negatively correlated with numbers of errors made crossing the narrowest beam (coefficient = −0.52, p = 0.04) in the TRF group. This correlation was not detected in the Q175 control group (coefficient = 0.16 and 0.13, respectively). Similarly, the TRF-treated group showed a negative correlation between their cage activity level and beam crossing errors (coefficient = −0.51, p = 0.01). This correlation was, again, not detected in the Q175 control group (coefficient = −0.06). These data indicate that the TRF-driven improvement in activity rhythms is correlated with the reduction in beam crossing errors.
### Expression of multiple HD markers in striatum were altered by TRF
Striatum is one of the key brain structures of the cortical-basal ganglia circuit controlling motor function, and it has been shown to be particularly vulnerable in HD. Previous work has identified HD-driven changes in transcription in the striatum of the Q175 mouse ( ). Using NanoString technology, we examined the impact of TRF on changes in gene expression of HD markers in the striatum of the Q175 mice as previously described ( ). The expression patterns were compared to Q175 ad lib controls ( ). The TRF regimen altered expression of immediate early genes such as Arc , Erg1,2,4 , and Fos , as well as receptors for neurotransmitters such as acetylcholine, histamine, 5HT, tachykinin, and dynorphin ( ). The IPA analysis tool was applied to the total dataset ( ) to identify corresponding enriched pathways and biofunctions ( ). The top canonical pathways identified included (in descending order of significance): G protein-coupled receptor (GPCR) signaling, cAMP-mediated signaling, and glutamate receptor signaling. The top upstream regulators included BDNF, CREB1, and HTT. Hence, the TRF treatment significantly altered the patterns of expression of genes linked to HD and modulated multiple transcriptional pathways.
Top 5 HD markers in the striatum of Q175 altered by the TRF treatment
Altered expression level of multiple HD markers in the striatum of the Q175 HD model. A , Differentially expressed genes in the striatum observed between TRF group and ad lib group using NanoString (find all gene expression data in ). The same Q175 mice that underwent activity/sleep monitoring and behavioral tests were allowed to recover for four weeks from manipulations before tissue collection. The signal intensity of individual genes was normalized by adjusting to internal positive standards within each sample (see Materials and Methods). B , Enriched functional clustering in the striatum using the IPA analysis tool (based on data in ; uncorrected Fisher’s exact test p value < 0.05). The clusters of interest with statistical significance are picked and enriched biofunctions in those picked clusters are shown (in descending order of significance). The picked clusters include Behavior ( p = 2.72E-17, color orange), Cell-to-cell signaling and interaction ( p = 1.02E-17, color blue), inflammatory response ( p = 2.87E-04, color pink), and neurologic disease ( p = 8.74E-14, color green).
Top 10 canonical pathways and upregulators identified using IPA analysis in striatum of Q175 under TRF regimen
Full dataset of expression of HD markers in the striatum of Q175 that are tested by using NanoString technology. Bold text indicates significant difference between ad lib and TRF feeding protocols
## Discussion
A range of circadian deficits in the mouse models of HD have been reported, detailing the impact on rhythms in behavior and physiology ( ; ; ; ; ; ). The findings suggest that the most common sleep-related clinical complaints of HD patients (i.e., difficulty falling asleep, frequent awakenings during sleep, and difficulty staying awake during the active cycle) are due, at least in part, to the disease-induced dysfunction in the circadian system. These findings raise the possibility of treating HD symptoms by improving the regularity/robustness of circadian rhythms in activity and rest ( ; ).
In the present study, the Het Q175 mice were allowed access to their food (standard chow, 6 h) nightly for three months starting at an age before the onset of motor symptoms. We confirmed that the animals consumed similar amounts of food and the body weights were not significantly decreased by this feeding regimen. We demonstrate that the nightly TRF regimen improved the daily activity rhythm with increases in the rhythmic strength as measured by power of the periodogram and decreases in cycle-to-cycle variability in activity onset. Prior work in WT mice did not find an impact of TRF on locomotor activity patterns ( ). While we are not sure of the difference, we did evaluate older mice (six months) who may be already exhibiting some age-related decline in locomotor activity rhythms. The TRF treatment also advanced the time that the mice ended their sleep phase without changes in total amount of sleep per cycle. Critically, the TRF regimen also improved performance of the HD mutant mice on two different motor tests.
The beneficial impact of TRF on motor performance could be dependent on or independent from the improvements in circadian output. We examined this issue by taking advantage of the animal-to-animal variation in the impact of the treatment on circadian and motor function. Using our most sensitive motor assay (i.e., challenge beam test), we found that the improved circadian behavior was correlated with improved motor function in the TRF group (coefficient = −0.52, P = 0.04). This finding leads us to conclude that improved circadian timing underlies the improved motor function in the treated mice. Furthermore, a variety of different approaches aiming to boost circadian output have now been found to improve motor functions in different HD mouse models. There is evidence that improving the sleep/wake cycle with sleep-inducing drugs ( ; ), stimulants ( ; ), bright light and restricted wheel access ( ), and blue light ( ) can treat HD symptoms. This body of work supports our general hypothesis that TRF improves circadian robustness and acts via this mechanism to delay disease symptoms in HD.
Our data clearly demonstrate that the benefits of TRF extend to physiologic measures such as HRV. Cardiovascular events are a major cause of early death in the HD population ( ; ) and the dysfunctional autonomic nervous system may be linked to the increased cardiovascular susceptibility. HRV measures the variation in the beat-to-beat (R-R) interval. It reflects the dynamic balance of sympathetic and parasympathetic control of heart function, and displays a robust circadian rhythm. A prior study demonstrated that the Q175 mice exhibit a loss of circadian control in HRV day/night differences, as well as an overall decrease in HRV over a 24-h period when compared to WT controls ( ). It is worthwhile to note that a similar decrease in HRV has also been reported in HD patients beginning during the presymptomatic stage of disease progression ( ; ). Reduced HRV is generally considered an indication of poor cardiovascular health and a predictor for cardiovascular disease and mortality ( ). To our knowledge, this is the first study showing that a TRF regimen can improve HRV in a disease model.
Prior work in Drosophila has also demonstrated the benefits of TRF in ameliorating age-related cardiovascular decline ( ). In this model, TRF downregulates expression of gene involved in mitochondrial electron transport while increasing expression of a cytoplasmic chaperonin ( ). This study also found that mutations in circadian clock genes prevented the benefits of TRF. TRF improved the amplitude of the day/night rhythms in many circadian-regulated transcripts. In mice, genetic disruption of the circadian clock results in a variety of cardiovascular deficits ( ; ). Together, this work suggests that TRF can work in concert with the photic regulation of the circadian system to boast the amplitude and perhaps the phasing of the molecular clock-work.
Lifestyle interventions have been suggested to be preventative and therapeutic for diseases associated with aging, such as type-2 diabetes, cardiovascular disease and increasingly neurodegenerative disorders. For example, caloric restriction (CR) has consistently been found to prolong life span and protect against a variety of pathologic conditions ( ; ). Conceptually, the TRF regimen used in the present study is quite distinct from CR. While CR focuses on overall, dramatic reduction in energy intake, TRF emphasizes the temporal pattern of fasting without a reduction in overall energy intake. Mechanistically, TRF may activate the same beneficial biochemical pathways as CR ( ; ) but would likely be easier to implement in a patient population ( ; Marder et al., 2009). In humans, the time of food availability would be during the day when food is normally consumed while the fast would be extended past the normal night. Prior studies have demonstrated the benefits of an 8:16 feed/fast cycle in improving the metabolic state and motor coordination of mice without altering caloric intake or nutrient composition ( ; ). In the HD-N171-82Q mouse model, CR improves motor performance and survival while reducing cell death ( ). Prior work in the R6/2 HD model has shown that TRF can restore HD-driven disruption in circadian gene expression in the liver ( ) and improve locomotor activity as well as exploratory behavior in the open field without increasing life span ( ). Together, these data suggest that feeding schedules could play a role in the treatment of HD and could lead to the development of new treatment options for neurodegenerative disorders.
The mechanisms underlying the beneficial effects of the TRF regimen on Q175 mouse model are uncertain and likely mediated by multiple pathways. Our data indicate that the TRF treatment changes the transcriptional environment in a brain region intimately involved in HD, i.e., the striatum. We used the NanoString technology with the IPA platform to analyze the transcriptional changes evoked by TRF. We found that >50% of genes (13/24) that had been shown downregulated in Q175 controls in a prior study (comparison with age-matched WT controls ( ) were upregulated by this treatment ( ), suggesting our circadian manipulation may exert beneficial effects through these pathways ( ). For example, striatal histamine receptor H3 ( Hrh3 ) may connect improved circadian rhythms to improved motor functions. Hrh3 , a GPCR, is strongly expressed in the cortico-striatal circuits controlling motor behavior ( ). Prior work found a significant reduction in Hrh3 radioligand binding in tissue of HD patients ( ) suggesting a central role of the histaminergic system in this basal ganglia disorder. Histamine is a well-known regulator of the sleep-wake cycle ( ; ) and specifically, H3R modulates striatal neurons through its regulation of glutamate ( ), GABA ( ; ), and dopamine ( ; ) release. In a recent study, we found that daily treatment with an H3R inverse agonist improved several behavioral measures in the Q175 mice including activity and sleep rhythms, exploratory behavior, mood ( ). GPCR signaling and glutamate receptor signaling are the top three pathways identified in the IPA analysis as being regulated by TRF. Unfortunately, the feeding schedule did not reduce the levels of mutant Htt ( ). Nevertheless, identifying treatments that improve the standard of living for HD patients remains an important goal. Future work will need to specifically evaluate the role of the histaminergic system in mediating the benefits of TRF for the sleep-wake cycles as well as motor performance.
## Conclusion
Imposed feeding cycles have the capacity to synchronize or increase the amplitude of circadian oscillations throughout the body. Disturbances in the sleep/wake cycle are by now a well-established symptom of neurodegenerative diseases, and here we show that we can treat the HD symptoms by controlling the timing of food availability. The results presented in our preclinical study suggest that a TRF regimen could be a useful management tool for neurodegenerative disease patients. More generally, the present study adds to a growing body of evidence that improvements in “circadian hygiene” through attention to regularity in environmental signaling, including timed feeding, leads to improvements in health outcomes for a wide range of human diseases including neurodegenerative disorders.
|
Subsets and Splits