Clinical and Preclinical Evidence for M1 Muscarinic Acetylcholine Receptor Potentiation as a Therapeutic Approach for Rett Syndrome

Rett syndrome (RTT) is a neurodevelopmental disorder that is characterized by developmental regression, loss of communicative ability, stereotyped hand wringing, cognitive impairment, and central apneas, among many other symptoms. RTT is caused by loss-of-function mutations in a methyl-reader known as methyl-CpG-binding protein 2 (MeCP2), a protein that links epigenetic changes on DNA to larger chromatin structure. Historically, target identification for RTT has relied heavily on Mecp2 knockout mice; however, we recently adopted the alternative approach of performing transcriptional profiling in autopsy samples from RTT patients. Through this mechanism, we identified muscarinic acetylcholine receptors (mAChRs) as potential therapeutic targets. Here, we characterized a cohort of 40 temporal cortex samples from individuals with RTT and quantified significantly decreased levels of the M1, M2, M3, and M5 mAChRs subtypes relative to neurotypical controls. Of these four subtypes, M1 expression demonstrated a linear relationship with MeCP2 expression, such that M1 levels were only diminished in contexts where MeCP2 was also significantly decreased. Further, we show that M1 potentiation with the positive allosteric modulator (PAM) VU0453595 (VU595) rescued social preference, spatial memory, and associative memory deficits, as well as decreased apneas in Mecp2+/- mice. VU595’s efficacy on apneas in Mecp2+/- mice was mediated by the facilitation of the transition from inspiration to expiration. Molecular analysis correlated rescue with normalized global gene expression patterns in the brainstem and hippocampus, as well as increased Gsk3β inhibition and NMDA receptor trafficking. Together, these data suggest that M1 PAMs could represent a new class of RTT therapeutics.


Introduction
Loss-of-function mutations in the Methyl-CpG-Binding Protein 2 (MeCP2) gene are linked to a neurodevelopment disorder known as Rett syndrome (RTT) [1]. MeCP2 is a methyl reader protein that epigenetically regulates gene transcription either positively or negatively, depending upon its binding partners [2]. Pathogenic mutations in MeCP2 primarily either destabilize the protein, disrupt its ability to bind methylated DNA, or prevent its association with the nuclear receptor corepressor (NCoR) complex [3,4]. RTT patients exhibit a myriad of symptoms that include developmental regression, loss of communicative ability, impaired social engagement, compulsive hand wringing, and respiratory dysfunction [5,6].
Similar to other neurological disorders, RTT therapeutic targets are traditionally identified through expression or functional analysis in Mecp2 knockout mice. Motivated by 1 the high failure rate of promising preclinical compounds to translate into effective therapeutics [7], we recently adopted the alternative strategy of performing transcriptional profiling in autopsy samples from RTT patients. Through this mechanism, we sought to identify potential targets that would originate from a position of translational relevance that could be back-modeled in rodents for safety and efficacy studies [8,9]. Using this approach, we previously identified conserved disruption of mRNA from four of the five subtypes of muscarinic acetylcholine receptors (mAChRs) in nine motor cortex and six cerebellar RTT samples compared relative to age, sex, and postmortem interval matched controls [8]. As a proof of concept, we demonstrated that a positive allosteric modulator (PAM) of the mAChR subtype 4 (M 4 ) improved cognitive, social, and respiratory symptom domains in Mecp2 +/mice.
Decreased cholinergic tone is a well-established aspect of RTT pathophysiology in patients which predates the discovery of MeCP2 as the causative gene. In RTT autopsy samples, both decreased choline acetyltransferase (ChAT) activity and radiolabeled vesamicol binding have been reported in the putamen and in the thalamus [10,11]. Similar results are also observed in Mecp2 +/rats, where decreased acetylcholine levels are observed at symptomatic ages while all other neurotransmitters remain unaffected [12]. Data from RTT mouse models agree with these findings, as loss of Mecp2 in cholinergic neurons evokes comparable anxiety, motor, social, cognitive, and cardiac phenotypes to what is observed with global Mecp2 knockout [13][14][15]. Notably, these phenotypes also appear to be responsive to compounds that increase cholinergic tone such as donepezil, PNU282987, and dietary choline [13,14,16], as well as to viral delivery of MeCP2 to cholinergic neurons in RTT model mice [15]. These data reinforce the hypothesis that methods of targeting cholinergic neurotransmission may have utility as RTT therapeutics and align with our data in autopsy samples that support the use of compounds that specifically potentiate mAChR signaling.
The clinical development of mAChR modulators has a robust and complex history, highlighted by seminal studies in the 1980s showing efficacy in cognitive and social symptom domains with administration of the M 1 -preferring, nonselective agonist xanomeline in patients with Alzheimer's disease (AD) [17,18]. While hypercholinergic adverse effects plagued mAChR discovery programs, recent advancements have shown promise in clinical trials and provided renewed hope [19]. In this study, we use a series of 40 temporal cortex samples from RTT autopsies to show that decreased M 1 mAChR expression is both highly conserved and exhibits a linear relationship with MeCP2 expression. Further, we demonstrate that administration of the selective M 1 PAM VU0453595 (VU595) normalizes social interaction phenotypes, rescues associative and spatial learning defects, and decreases apneas in RTT model mice while producing no overt adverse effects. We also show that treatment with VU595 normalizes global gene expression patterns in Mecp2 +/animals and rescues pathways associated with Gsk3β inhibition and NMDA receptor (NMDAR) trafficking. Together, these data support the continued development of M 1 mAChR positive allosteric modulators as potential RTT therapeutics.

Study Design
Human samples were obtained from the NIH NeuroBioBank and the University of Maryland Brain and Tissue Bank under PHS contract HHSN-271-2013-00,030. The tissues were postmortem and fully de-identified and, as such, are classified as exempt from Human Subject Research regulations. Animal work was conducted under the oversight of the Loyola University and Vanderbilt University Institutional Animal Care and Use Committees.
Mice were assigned to dosing groups at random and phenotyping or molecular quantitation was either performed by a blinded researcher or by automated software. Statistics were carried out using Prism 6.0 (GraphPad) and Excel (Microsoft). All data shown represent mean ± SEM. Statistical significance between groups was determined using two-tailed unpaired or paired student's t tests and one-or two-way analysis of variance (ANOVA), with Bonferroni's or individual student's t test post hoc analysis, as specified in each figure legend. Mice sacrificed for gene expression were euthanized using CO 2 inhalation at a flow rate of 2 L per minute, as recommended by the AVMA.

Selection of Animal Model
The selection of mouse model (Mecp2 +/tm1.1 bird ), sex, and sample size was based on the standards established by the National Institute of Mental Health and RTT research community. Heterozygous female mice were chosen to achieve construct validity, as the overwhelming majority of RTT patients are female and MeCP2 mut/+ . A robust literature also establishes RTT-like phenotypes in this model (face validity), albeit milder than what is observed in humans, and variable in both age of onset and symptom severity, due to random x inactivation [20][21][22]. To account for variability, we used N values of > 10 mice per genotype and treatment group, compared test mice to their own baseline where possible, and performed assays at 20 weeks, when phenotypes become reproducibly observed in virtually all Mecp2 +/mice within our colony [8,9,23].

RNA-cDNA Preparation
The demographics of autopsy samples used in these studies are summarized in Table S1. Briefly, Brodmann Area (BA) 38 (temporal cortex) samples were received from N = 40 RTT samples with an age of 22.9 ± 2.1 years and a PMI of 20.2 ± 1.5 h. N = 12 controls were an age of 21.8 ± 2.4 years and a PMI of 19.3 ± 2.0 h. Approximately 1 g of the temporal cortex was impact-dissociated under dry ice and then pulverized using mortar and pestle under liquid nitrogen. Total RNA and cDNA were prepared from 200 mg of tissue using standard trizol-chloroform methodology. Brain samples from 20-week-old Mecp2 +/-(tm1.1 Bird) mice were prepared using identical methodology.

mRNA and Protein Analysis
Quantitative real-time (qRT)-PCR was performed on Bio-Rad CFX96 instrumentation using Thermo Fisher Assay on Demand primer-probe kits. The assay IDs used are CHRM1 (Hs00265195_s1), CHRM2 (Hs06634238_s1), CHRM3 (Hs00327458_s1), CHRM4 (Hs00265219_s1), CHRM5 (Hs00255278_s1), and Chrm1 (Mm00432509_s1). Human and mouse samples were normalized to the internal control G6PD (Hs00166169_m1; Mm00656735_g1). qRT-PCR data was analyzed using the delta-delta Ct method. Differential RNA-sequencing was performed as previously described [8]. Differentially expressed genes were processed by Reactome.org for pathway enrichment analysis, and false discovery rates were calculated as a function of mapped genes.

Compound Administration and Phenotyping
In the studies described herein, 20-week-old Mecp2 +/and Mecp2 +/+ females were used which represent an age where consistent phenotypes are observed in our colony. Based on the pharmacokinetics and T max of the molecule [24], either vehicle (10% Tween 80) or 10 mg/kg of VU0453595 (VU595) was administered to Mecp2 +/and Mecp2 +/+ female mice (20w) via intraperitoneal (ip) injection 30 min before the start of the phenotypic assay and/or tissue harvest. The order phenotyping was performed was open field, three-chamber social interaction, novel object recognition, whole-body plethysmography, and contextual fear conditioning, with a minimum of 5 days washout between assays.

Open Field
Spontaneous locomotion and anxiety were measured using the open field assay. Test mice were injected with either VU595 or vehicle and placed in the activity monitoring chamber. Exploratory and locomotor behavior was then monitored using Activity software to quantify beam breaks in the X, Y, and Z planes.

3-Chamber Social Preference Assay
Control and test mice were placed in a standard 3-chamber apparatus and allowed to habituate for 7 min. A novel mouse (stranger 1, 6-week-old C57B6, female) that was restrained in a wire cage was then placed in one of the end chambers, and an empty cup was placed in the opposing end chamber. The test mouse was then allowed to explore all 3 chambers for 7 min before being returned to its home cage. After 1 h, the test mouse was returned to the 3-chamber apparatus, and the ability to distinguish between stranger 1 and a novel stranger (stranger 2, 6-week-old C57B6, female) was quantified for 7 min using AnyMaze™ software.

Novel Object Recognition Assay
Novel object recognition (NOR) was conducted as described in [25]. Briefly, test mice were placed inside a chamber with 2 identical objects (a slide box or ball) and allowed to explore for 10 min. The animals were then returned to their home cage. After 1 h, test mice were returned to the chamber a final time for 5 min, and 1 of the 2 objects was replaced with a novel object. The test was video recorded and the seconds spent directly sniffing each object were scored by a blinded reviewer. Discrimination index was defined as (Time novel -Time familiar ) / (Time novel + Time familiar ). Direct sniffing in the novel object recognition assay was defined as an approach where the mouse's nose either made contact or came in close proximity to the object.

Contextual Fear Conditioning
Animals were treated with VU595 and fear-conditioned on day 1 of the task, and the percent of time spent freezing was assessed 24 h later. On the conditioning day, mice were placed into an operant chamber with a shock grid (Med Associates Inc.) in the presence of a 10% vanilla odor cue. Following a 3-min habituation period, mice were exposed to two 1-s 0.7-mA foot shocks spaced 30 s apart. After 24 h, mice were placed back into the same shock chamber with a 10% vanilla odor cue, and the percent of time spent freezing during a 3-min testing period was assessed by Med Associates software.

Whole-Body Plethysmography
Unrestrained Mecp2 +/and Mecp2 +/+ mice were placed in a whole-body plethysmograph recording chamber (Buxco, 2-site system) with a continuous inflow of air (1 L/min). Following a habituation period of 30 min, a baseline recording was established for 30 min. Mice were then removed from the chamber, injected with VU595 or vehicle, and reacclimated for 30 min, and respiratory measurements were made for an additional 30 min. Analysis was performed using FinePointe Research Suite (v2.3.1.9). Apneas, defined as pauses spanning 2 × the average expiratory time of the previous 2 min, were quantified using the FinePointe apnea software patch, followed by manual spot-checking of the larger data set. Only periods of motion-free recording were analyzed. All filters were applied while the researchers were blinded to the genotype and treatment group.

M 1 Expression is Decreased in 40 RTT Temporal Cortex Autopsy Samples
Our previous work used RNA-sequencing (RNA-seq) to identify conserved disruption of mAChRs in 9 motor cortex and 6 cerebellar RTT autopsy samples [8]. While among the highest-powered transcriptional profiling studies conducted using RTT autopsy samples, the potential for variability across a larger population must be considered when determining the value of a prospective therapeutic target. To address this, we obtained 40 temporal cortex samples from patients clinically diagnosed with RTT as well as 12 age, sex, and postmortem interval matched controls (Table S1). While fully subtype-selective mAChR antibodies do not exist [26], we isolated protein and compared total (M 1 -M 5 ) mAChR levels via fluorescent Western blot with an M 1 -preferring antibody. These experiments showed a significant decrease in mAChR levels in the temporal cortex of RTT patients relative to matched controls ( Fig. 1A-B).
We next performed qRT-PCR to quantify mAChR subtype expression in this larger sample set. As shown in Fig. 1C and Supplemental Fig. 1, we observed a significant reduction in M 1 , M 2 , M 3 , and M 5 expression. Remarkably, we did not see a significant change in M 4 expression in the temporal cortex (Fig. 1D), which stands in contrast to what was observed in the motor cortex and cerebellum [8]. To determine whether the expression of any of the mAChR subtypes correlated with MeCP2 levels, we quantified MeCP2 mRNA expression and compared it relative to the M 1 -M 5 mRNA levels within each patient. These experiments showed a significant linear relationship between MeCP2 and M 1 mRNA, such that M 1 levels are preferentially decreased in a context in which MeCP2 levels are also decreased ( Fig. 1E-F). This relationship was not observed with any of the other mAChRs in RTT patients; however, it was observed with M 2 and M 5 expression in controls (Fig. S1). Although limited by the availability of human samples, using the quantified decrease in M 1 expression, we back-calculated a value of 86.7% power for this analysis (ClinCalc), which exceeds the standard 80% used in many clinical studies.

M 1 Expression is Decreased in the Brainstem of Mecp2 +/-Mice
When coupled with the recent advancements in M 1 positive allosteric modulator (PAM) drug discovery, the finding that M 1 expression is decreased in the brains of RTT patients is potentially salient. To determine whether this observation also extended to mouse models of RTT, we isolated synaptosomes from the cortex, hippocampus, and brainstem of female 20-week-old Mecp2 +/and Mecp2 +/+ mice and performed Western blotting with an M 1 -preferring antibody. In contrast to what was observed in human samples, significantly decreased mAChR expression was only observed in the brainstem of Mecp2 +/mice (Fig. 1G), potentially indicative of temporal, spatial, and/or species-specific factors contributing to the regulation of M 1 expression.

M 1 Potentiation Rescues Social and Cognitive Deficits in Mecp2 +/-Mice
Given the significant decrease in M 1 expression in clinical and preclinical sample sets, we next sought to determine whether M 1 potentiation could improve RTT-like phenotypes in Mecp2 +/mice. Mecp2 +/mice and Mecp2 +/+ littermate controls were treated with either vehicle (10% tween 80) or the M 1 -PAM VU595 and progressed through a battery of tests encompassing the major RTT symptom domains, including open field (motor and anxiety), three-chamber social interaction (social), novel object recognition (spatial memory), and contextual fear conditioning (associative memory). Mice were treated (ip) 30 min before each test, with a 5-day washout period between assays. The strategy of using a comprehensive phenotypic assessment was chosen to control for the broad expression pattern of the M 1 receptor and the potential for VU595's PAM activity in brain regions with normal expression to either evoke adverse effects, improve phenotypes independent of M 1 expression, and/or influence interdependent phenotypes (i.e., anxiety and fear conditioning, motor function and sociability).
In the open field assay, we observed no difference in spontaneous locomotion between Mecp2 +/mice and littermate controls, regardless of treatment ( Fig. 2A-B). This finding also extended to time spent in the center, where no changes were observed as a function of drug or genotype. In the three-chamber social interaction assay, there was no effect of genotype or drug on the sociability phase of the assay, with all test groups showing a preference for the Student's t test. E-F Linear regression analysis. A correlative analysis shows a significant linear relationship between M 1 and MeCP2 transcript levels, which is not observed with M 4 . Linear regression. G Western blot. Relative to Mecp2 +/+ controls, synaptosome preparations from 20-week-old Mecp2 +/mice show a significant reduction in mAChR levels in the brainstem. Two-way ANOVA with Tukey post hoc analysis. N = 5/treatment/genotype. *p < 0.05, **p < 0.01 stranger 1 mouse over the empty cup (Fig. 2C). Although not definitive, this finding is important in light of the sensoryperception deficits in RTT model mice [27][28][29][30][31], as it establishes that the capacity of test mice to detect the stranger mouse is comparable to controls. Similar to other studies [8,9], vehicle-treated Mecp2 +/mice failed to show a preference for the novel mouse in the second phase of this assay, while littermate controls spent significantly more time with the stranger 2 mouse, regardless of treatment (Fig. 2D). VU595 treatment significantly improved this parameter in Mecp2 +/mice and restored social preference to levels comparable to littermate controls.
We next assessed spatial memory phenotypes using novel object recognition, where Mecp2 +/+ controls showed significantly increased preference for the novel over the familiar object, and vehicle-treated Mecp2 +/mice did not discriminate between the two (Fig. 2E). Following VU595 administration, there was no longer a significant difference between RTT and control test groups, indicating that spatial memory was improved. To determine whether these effects also extended to associative learning, we next progressed mice through a contextual fear assay. Following a mild foot shock, vehicle-treated Mecp2 +/mice presented with a significant decrease in time spent freezing compared relative to vehicle-treated controls when re-exposed to the test chamber (Fig. 2F). In contrast, VU595-treated Mecp2 +/mice were not distinguishable from controls indicative of a modest improvement in associative memory.

VU595 Improves Respiratory Phenotypes in RTT Mice
Using the decrease in M 1 expression in the brainstem of Mecp2 +/mice as rationale, we next assessed VU595's efficacy on respiratory phenotypes using whole-body plethysmography (WBP). Following 30 min of baseline recording, Neither genotype nor VU595 administration (10 mg/kg, ip) had an impact on spontaneous locomotion (A) and anxiety phenotypes (B) relative to vehicle Mecp2 +/+ control mice. C-D Three-chamber social interaction assay. No genotype or compound effect was observed in the sociability phase of the assay (C), as all mice demonstrated a preference for the stranger 1 mouse over the empty cup. When exposed to a familiar (stranger 1) and a novel (stranger 2) mouse, Mecp2 +/+ mice showed a preference for social novelty, independent of treatment. Conversely, vehicle-treated Mecp2 +/mice did not distinguish between stranger 1 or stranger 2, and VU595 treatment restored preference for the novel mouse. Two-way ANOVA with Tukey post hoc analysis. E Novel object recognition. DI, discrimination index. Vehicle-treated Mecp2 +/mice did not show a preference for the novel object over a familiar object, indicative of a deficit in spatial memory. VU595 administration significantly improved this phenotype in Mecp2 +/mice. Two-way ANOVA with Tukey post hoc analysis. F Contextual fear conditioning. Vehicle-treated Mecp2 +/mice presented with significantly decreased freezing when re-exposed to an aversive environment, suggesting an impairment in associative memory. Treatment with VU595 before the training phase of this assay normalized freezing in Mecp2 +/mice to levels comparable to Mecp2 +/+ controls. Two-way ANOVA with Tukey post hoc analysis. Ns, not significant. *p < 0.05, **p < 0.01, ***p, 0.001. #p < 0.05 within genotype comparison we administered either vehicle or VU595 and returned the mouse to the recording chamber. After trimming periods of activity from the analysis, drug effects were measured 30 min post-dose (Fig. 3A). Under this paradigm, vehicletreated Mecp2 +/mice exhibited significantly more apneas per 10k breaths than littermate controls, and administration of the M 1 PAM significantly reduced the number of apneas by 49.8% (Fig. 3B). To ensure that this result was not linked to differences in random × inactivation skewing of the mutant Mecp2 allele and/or general health differences of Mecp2 +/test groups, we compared the number of apneas in individual mice relative to their own predrug baseline. During the baseline period, both vehicle and VU595 test groups exhibited a comparable number of apneas (Fig. 3C-D); however, following compound administration, only Mecp2 +/mice treated with the M 1 Fig. 3 Administration of VU595 improves apneas by facilitating the transition from inspiration to expiration. Whole-body plethysmography (WBP). Unless noted, statistical comparisons were made relative to vehicle-treated Mecp2 +/+ mice using a two-way ANOVA with Tukey post hoc analysis. N = 15 Mecp2 +/+ /treatment group, N = 22 Mecp2 +/-/treatment group. A Representative traces of respiratory patterns. B-D Apneas were significantly increased in both groups of Mecp2 +/mice; however, VU595 treatment significantly reduced the total number of apneas relative to the vehicle-treated group. A comparison relative to the baseline of each mouse shows that both groups of Mecp2 +/mice began the experiment with comparable apnea numbers, but that only the VU595-treated group showed significant improvement. E-F Vehicle-treated Mecp2 +/mice showed a significant increase in the percentage of the total breath occupied by the pause period between both inspiration and expiration. VU595's efficacy on apneas was associated with a significant reduction in the time between inspiration and expiration, with no effect on the transition from expiration to inspiration. G Tidal volume was not impacted by treatment or genotype. H-I Breath frequency and minute volume were significantly increased in vehicle-treated Mecp2 +/mice and were normalized by VU595 administration. J-K Increased breath frequency was associated with a significant decrease in expiratory time in vehicle-treated Mecp2 +/mice, which was rescued by VU595 administration. No change in inspiratory time was observed as a function of genotype or treatment. *p < 0.05 relative to vehicle-treated Mecp2 +/+ mice, #p < 0.05 relative to vehicle-treated Mecp2 +/mice PAM showed a significant reduction in apnea numbers, indicative of a bona fide compound effect (Fig. 3D).
We next determined the location in the respiratory cycle where VU595 was impacting apneas. We examined the percent of the breath that was spent on the pause period between inspiration and expiration (TB), as well as the percent of time spent paused between expiration and inspiration (TP). In vehicle-treated Mecp2 +/mice, both transition points occupied a significantly greater percentage of the breath relative to vehicle-treated Mecp2 +/+ controls (Fig. 3E-F). Conversely, only TB (inspiration to expiration) was normalized by VU595, while the compound did not affect TP (expiration to inspiration). These data suggest that M 1 potentiation decreases apneas in Mecp2 +/mice by facilitating the transition between the inspiratory and expiratory phases of the breath cycle.
No effect of drug or genotype was recorded on tidal volume (Fig. 3G); however, we did quantify a significant increase in respiratory frequency in vehicle-treated Mecp2 +/mice that was significantly decreased by VU595 administration (Fig. 3H). As a corollary, we also observed a significant increase in the total volume respired per minute (minute volume) in vehicle-treated Mecp2 +/mice, which was also normalized by the M 1 PAM (Fig. 3I). To investigate this further, we quantified the amount of time spent in the inspiration phase relative to expiration in each breath. As shown in Fig. 3J-K, the increase in breath frequency and minute volume was associated with a significantly shorter expiratory phase of the breath in vehicle-treated Mecp2 +/mice and was normalized with VU595 treatment.

M 1 Potentiation Normalizes Global Gene Expression Patterns
MeCP2 is a highly abundant transcription factor in the brain that regulates gene expression locally in complex with proteins like CREB, as well as globally by linking methylated DNA with the larger chromatin architecture [2]. As a consequence, pathogenic mutations in MeCP2 result in a loosening of chromatin and a broad, but modest disruption of global gene expression patterns. M 1 is a postsynaptic, G q -coupled receptor whose activation is associated with an increase in intracellular calcium and activation of calcium-dependent transcription factors [32]. To determine whether VU595 normalizes Mecp2-disrupted gene expression patterns, we administered either VU595 or vehicle to Mecp2 +/and Mecp2 +/+ mice, harvested the brainstem and hippocampus at 30 min post-dose, and performed differential RNA-seq analysis. The brainstem and hippocampus were chosen because they represent regions of the brain associated with apneas and cognition, where M 1 PAM efficacy was observed. A comprehensive list of significantly disrupted genes and pathways for each brain region is provided in Supplemental Tables 2, 3 , 4, 5, 6, 7, 8, and 9. In the brainstem, vehicle-treated Mecp2 +/mice had 2,141 differentially expressed genes (DEGs) relative to vehicletreated Mecp2 +/+ samples, and 1,384 (64.7%) of those genes were no longer disrupted in mice treated with VU595 (Fig. 4A). Reactome.org (v76) analysis of the normalized genes showed enrichment in pathways associated with assembly and cell surface presentation of NMDARs, among others. We next examined genes that were significantly changed by VU595 treatment in both Mecp2 +/+ and Mecp2 +/mice, which identified 119 genes that were affected in both genotypes (Fig. 4B). Pathway analysis of these genes points to an enrichment in FOXO/P21/P53 signaling, which has previously been associated with pathogenic mutations in MeCP2 that disrupt NCoR/HDAC3 binding [33].
In total, 2,727 DEGs were quantified in the hippocampus of vehicle-treated Mecp2 +/mice, and 2,294 (84.1%) of those genes were no longer disrupted in Mecp2 +/mice treated with VU595 (Fig. 4C). Reactome.org analysis of the normalized genes showed enrichment in pathways associated with the regulation of nonsense-mediated decay and formation of the 40S ribosomal subunit, among many other pathways. The expression of 46 genes exhibited conserved disruption by VU595 in both Mecp2 +/+ and Mecp2 +/mice, which demonstrated no significant pathway enrichment (Fig. 4D).
In addition to calcium-dependent gene regulation, M 1 activation impacts the activation of calcium-independent kinases [34], many of which are integral to maintaining proper synaptic transmission. In light of the brainstem RNAseq and whole-body plethysmography data, one kinase of interest was Gsk3β, which is known to regulate NMDAR trafficking [35] and whose inhibition has been shown to reduce respiratory phenotypes in RTT model mice [36]. To determine whether M 1 potentiation was impacting Gsk3βsignaling, we treated Mecp2 +/+ and Mecp2 +/mice with VU595, isolated protein from the brainstem, and performed Western blotting to profile the inhibitory (S9, S21) and activation (Y216/Y279) phosphorylation sites on Gsk3β and its homolog, Gsk3α. Contrary to previous reports [36], we observed no difference in Gsk3β or Gsk3α phosphorylation in Mecp2 +/mice relative to vehicle-treated controls at either the inhibitory or activation sites (Fig. 5A-C). Further, neither the inhibitory nor activation sites of Gsk3α were impacted by VU595 administration. Conversely, VU595 treatment significantly increased phosphorylation of the Gsk3β S9 inhibitory site in VU595-treated Mecp2 +/mice relative to vehicle-treated Mecp2 +/+ controls, while the activation site was unaffected (Fig. 5D-E). These data suggest that M 1 potentiation promotes inhibition of Gsk3β signaling in the brainstem of RTT model mice, which may contribute to the rescue of respiratory phenotypes independent of a primary deficit in Gsk3β-signaling. Fig. 4 Global gene expression patterns are normalized by M 1 potentiation. Differential RNA-sequencing relative to vehicle-treated Mecp2 +/+ control mice followed by reactome pathway enrichment analysis. A Brainstem. A total of 2,141 genes were significantly disrupted in vehicle-treated Mecp2 +/mice. Following VU595 treatment, the expression of 1,384 (64.6%) was normalized, while 757 remained disrupted. Right: pathway analysis of rescued genes. B Brainstem. Following VU595 administration, the expression of 119 genes was significantly altered in both Mecp2 +/+ and Mecp2 +/mice. Right: pathway analysis of conserved genes. C Hippocampus. A total of 2,727 genes were significantly disrupted in vehicle-treated Mecp2 +/mice. Following VU595 treatment, the expression of 2,294 (84.1%) was normalized, while 433 remained disrupted. Right: pathway analysis of rescued genes. D Hippocampus. Following VU595 administration, the expression of 46 genes was significantly altered in both Mecp2 +/+ and Mecp2 +/mice. Right: pathway analysis of conserved genes To determine whether changes in Gsk3β S9 phosphorylation correlated with a change in NMDAR presence at synapses, we isolated PSD95 containing synaptosomes from the brainstem of mice treated with either vehicle or VU595. We then performed Western blotting and probed using a nonselective NR2 subunit antibody. As shown in Fig. 5F-H, there was no difference between vehicle-treated Mecp2 +/+ and Mecp2 +/mice; however, VU595 treatment significantly increased the presence of NMDARs at the synapse of Mecp2 +/mice relative to controls. While further work is required to establish the functional consequences of increased NMDARs specifically at respiratory nuclei, these data suggest that increased M 1 /Gsk3β/NMDAR signaling in the brainstem correlates with the decrease in apneas in RTT model mice induced by the M 1 PAM VU595.

Discussion
Recent clinical development campaigns have shown remarkable progress toward potentially viable Rett syndrome therapeutics [37]. To improve the rigor of our research, we have embraced a reverse-translational approach to target identification. These efforts identified a conserved disruption in mAChRs in RTT autopsy samples [8]. Here, we confirm those findings using a cohort of 40 temporal cortex samples from RTT patients and demonstrate that four of the five mAChRs are significantly decreased in expression. Further, we show that positive allosteric modulation of the M 1 receptor normalizes social preference, associative and spatial memory deficits, and respiratory phenotypes in Mecp2 +/mice, a model with face and construct validity for the clinical disorder [23,37].
Among our most interesting findings is that M 1 expression correlated directly with MeCP2 expression in autopsy samples, such that M 1 levels are only low when MeCP2 levels are also low. Given that not all pathogenic mutations impact MeCP2 gene or protein expression (i.e., missense mutations), these data suggest that decreased M 1 expression may be enriched in patients with truncating and/or destabilizing mutations [38]. Our sample set precludes testing this hypothesis, as it consists of patients who were clinically diagnosed with RTT, but never genotyped for MeCP2 mutations. As such, further studies with larger sample sizes are required to substantiate whether a relationship between MeCP2 mutation and M 1 expression exists.
The M 1 receptor has a complicated history as a potential therapeutic target for neurological disorders that is highlighted by the clinical efficacy of the M 1 /M 4 -preferring agonist xanomeline in Alzheimer's disease [17,18]. The overwhelming majority of subsequent drug discovery efforts were derailed preclinically by the presence of either gastrointestinal or convulsive adverse effects [39]. In recent years, both of these adverse effects have been largely mitigated through selective targeting of the M 1 receptor with positive allosteric modulators that are void of agonist activity [40][41][42] and have low cooperativity profiles [43,44] or by co-dosing xanomeline with a peripherally restricted pan-muscarinic receptor antagonist [19]; however, it bears mentioning in this context since RTT patients are prone to gastrointestinal complications and seizures [45,46]. We did not observe either phenotype in our studies with Mecp2 +/mice, although it remains possible that they might emerge with chronic dosing.
The compound used in our studies, VU595, has previously been shown to exhibit efficacy in correcting abnormal social and spatial memory phenotypes in a chronic phencyclidine (PCP) model of schizophrenia [24], and our data in Mecp2 +/mice parallels these findings. Specifically, acute administration of VU595 restored social memory and/or preference in RTT mice without impacting sociability in the three-chamber social interaction assay. Impaired spatial and associated memory phenotypes were also rescued by M 1 potentiation in the novel object recognition and contextual fear assays, respectively. These results align with M 1 's expression pattern, where enrichment is observed at the prefrontal cortical and hippocampal circuits that largely govern these behaviors.
There is a well-established role for mAChRs in the control of breathing [47]; however, this is generally associated with M 2 and M 3 receptor expression in the autonomic nervous system and not with M 1 receptors. Given M 1 's limitedexpression pattern in the brainstem, it was an unexpected result that VU595 administration normalized respiratory phenotypes in Mecp2 +/mice. While increased breath rate has previously been reported in Mecp2 +/mice [48], here we link this finding in vehicle-treated mice with a decrease in expiratory time. Equally as important as the impact of VU595 on breath rate is its capacity to decrease apneas, which our data link to facilitation of the transition from inspiration to expiration. This finding agrees with in situ physiology data showing that increased post-inspiratory duration is associated with hyperexcitation of expiratory inputs originating in the Kölliker-Fuse nucleus in Mecp2 −/y mice [49]. In agreement with these data, glutamate microinjection at these inputs can evoke apneas in ex vivo preparations from the same model [48,50]. Accordingly, our data may suggest that M 1 potentiation may decrease apneas by tempering hyperexcitation in the Kölliker-Fuse nucleus, although this will need to be tested experimentally.
Our RNA-sequencing experiments established that one of the top pathways that was rescued in the brainstem of VU595-treated Mecp2 +/animals involves the assembly and trafficking of NMDARs, and follow-up studies implicate Gsk3β inhibition in mediating the observed efficacy. NMDA-signaling has a well-developed history in RTT research, where both deletion of the NR2A subunit [28] and administration of the NMDAR antagonist ketamine rescue multiple aspects of RTT pathophysiology in mice [51][52][53]. Previous studies have also shown that Gsk3β inhibition rescues multiple symptom domains, including respiration [36], and is known to regulate NMDAR activity and trafficking, in part, via activation of phosphatidylinositol 4 kinase type II α (PI4KIIα) [35]. NMDAR activation has been shown to dephosphorylate the S9 inhibitory site of Gsk3β [54], thereby providing a mechanistic point of convergence for ketamine, VU595 (M 1 PAM), and SB216763 (Gsk3β inhibitor) to act on apneas in the brainstem of RTT model mice. One important aspect of our data quantifying Gsk3β inhibition is that it was not significantly affected in the brainstem of vehicle-treated Mecp2 +/mice, and trended towards an increased baseline, as has previously been reported in other brain regions (36). This suggests that insufficient Gsk3β inhibition is likely not the driver of respiratory phenotypes. When analyzed in concert with the SB216763 data, one potential unifying explanation is that Gsk3β inhibition is a compensatory mechanism that cannot reach the required levels to maintain proper respiratory function in the absence of Mecp2 due to decreased M 1 -expression and signaling. If true, it would suggest that VU595 and SB216763 function by enhancing an existing neuroprotective mechanism to regulate synaptic NMDAR levels; however, further work is required to determine if this is a bona fide mechanism of action.
In summary, we describe the use of a large cohort of temporal cortex autopsy samples from RTT patients to establish the rationale for the M 1 receptor as a potential therapeutic target. Drug discovery campaigns for positive modulators of mAChRs have made significant advancements in recent years, such that compounds with favorable drug-like properties now exist. Using the M 1 PAM tool compound VU595, we observed efficacy in social, cognitive, and respiratory symptom domains in Mecp2 +/mice. Phenotypic improvement was correlated with both normalization of gene expression and increased inhibition of Gsk3β. Together, these data advocate for continued development and optimization of M 1 PAMs for RTT, as well as for related autism-associated disorders with overlapping pathologies.