Cholinergic Stress Signals Accompany MicroRNA-Associated Stereotypic Behavior and Glutamatergic Neuromodulation in the Prefrontal Cortex

Stereotypic behavior (SB) is common in emotional stress-involved psychiatric disorders and is often attributed to glutamatergic impairments, but the underlying molecular mechanisms are unknown. Given the neuro-modulatory role of acetylcholine, we sought behavioral-transcriptomic links in SB using TgR transgenic mice with impaired cholinergic transmission due to over-expression of the stress-inducible soluble ‘readthrough’ acetylcholinesterase-R splice variant AChE-R. TgR mice showed impaired organization of behavior, performance errors in a serial maze test, escape-like locomotion, intensified reaction to pilocarpine and reduced rearing in unfamiliar situations. Small-RNA sequencing revealed 36 differentially expressed (DE) microRNAs in TgR mice hippocampi, 8 of which target more than 5 cholinergic transcripts. Moreover, compared to FVB/N mice, TgR prefrontal cortices displayed individually variable changes in over 400 DE mRNA transcripts, primarily acetylcholine and glutamate-related. Furthermore, TgR brains presented c-fos over-expression in motor behavior-regulating brain regions and immune-labeled AChE-R excess in the basal ganglia, limbic brain nuclei and the brain stem, indicating a link with the observed behavioral phenotypes. Our findings demonstrate association of stress-induced SB to previously unknown microRNA-mediated perturbations of cholinergic/glutamatergic networks and underscore new therapeutic strategies for correcting stereotypic behaviors.


Introduction
Stereotypic behavior (SB) involves patterned, repetitive, apparently purposeless movements [1]. SB occurs in patients with Tourette's syndrome and obsessive-compulsive disorder [2], schizophrenia [3], autism [4,5], oculo-cerebro-renal disorder of Lowe (OCRL) [6], fragile X syndrome [7], Rett's syndrome [8] and the stereotypic movement disorder in children [9]. The underlying mechanism(s) are largely unknown, but impaired emotional stress modulation is likely involved, particularly in Tourette's regulation of motor behavior via exchanging glutamatergic with cholinergic signaling. To test this prediction, we examined transgenic mice with enforced decline in cholinergic signaling for both SB and brain transcriptomic changes of short (seeking modified miRNAs) and long transcripts (pursuing changes in their mRNA targets).
Our mouse model involved elevated expression levels of acetylcholinesterase (AChE), known to occur under stress [33]. This overexpression could be the effect of moderating excitatory influences on stress responses and involves alternative splicing, changing AChE properties [28,29,34]. The normally rare, soluble variant AChE-R is induced by stress via alternative splicing of AChE pre-mRNA and is a preferred target of the stress-induced miRNA-132 [30] and miRNA-125b [41]. Therefore, we hypothesized that long-term AChE-R over-expression could elicit stress-related stereotypic motor behavior, by acting on neural mechanisms leading to SB. We further predicted that these could encompass various brain regions which control defensive motor responses, including the basal ganglia, limbic brain nuclei [42][43][44][45] and brainstem regions [46]; and that this might involve marked changes in miRNAs targeted at cholinergic genes.
We set out to test the association of modifications in cholinergic-targeting miRNAs in SB by focusing on transgenic TgR mice overexpressing the human (h)AChE-R constitutively, under the minimal promoter-enhancer of cytomegalovirus (CMV). At the behavioral level, we tested TgR mice for spontaneous SB and reaction to the muscarinic stimulator pilocarpine. At the molecular level, we performed short RNA-sequencing and mRNA transcriptomic analysis of brain tissues from TgR mice, seeking modified cholinergic transcripts and their targeting miRNAs. CMV-directed gene expression depends on neuronal activity and blocks the neuro-inflammatory NFkB response regulated by cholinergic signaling [47]. Therefore, we further examined the neuronal distribution and intensity of hAChE-R overexpression and assessed neuronal activity levels by labeling the immediate early gene c-fos, shown to associate both with SB-related neural activation [16] and with AChE-R over-production [28].

Materials and Methods
Animals: Control, FVB/N and TgR transgenic mice were housed in the animal colony of the Herzog hospital, with an ambient temperature of 21 ± 2 • C and with a 12:12 h light: dark cycle. Experimental procedures were approved by the ethical committee MD98.08-2 of The Hebrew University of Jerusalem.

Behavioural Tests
Measurements of open field motor activity: Spontaneous stereotypic behavior in an open field was followed from weaning (3-4 weeks of age) to adulthood (2-12 months of age). Following 30 min habituation in a holding cage, mice were placed for 5 min at the mildly stressful, brightly lit, center of a 40 × 40 cm open field with 20 cm high walls. Videotaped locomotor activity was first quantified as total time spent in locomotion. Rearing episodes were quantified to evaluate the extent of disruption of normal exploration under SB. Episodes in which the mouse-initiated locomotion ("locomotion events") were counted. An index of asymmetry was calculated as [(left turns − half the total events)/total events] * 100. Thus, perfect symmetry would yield a value of 0 and total asymmetry-50%. For measurement of circadian motor activity, see Supplementary Material, Supplementary Figure S1.
Neurological tests: Mice were examined at the peak of their motor activity during the dark phase. Deficits in vestibular function were tested by holding mice by the tail, lifting and then lowering them over a metal cage top. Then, while the mouse was still, tactile stimuli to the right or left shoulder and trunk were applied using a cotton swab, and orienting to the stimulus was noted. Efficient and symmetric head and limb placement, and symmetric head orienting to tactile stimuli were taken as evidence of normal vestibular functioning.
Assessment of context-dependent SB: Mice were observed in a brightly lit open field on test day 1, 20 min after saline injection, returned to a holding cage for 20 min and re-observed in a dimly lit room while in their home-cage. On test day 2, mice were observed in the same sequence described above, 20 min after injection.
Forced swimming: Mice were placed in a circular water container, 20 × 20 cm in diameter and depth. Following a 4 min swim, at 25 • C, mice were taken out of the water and dried under a warm lamp. Total time spent swimming, number of swim episodes adjacent to the wall, number of swims across the container; number of circles and % asymmetry in the direction of swimming calculated by the index of asymmetry (see above) were quantified from video records.
Two unit serial maze: Modular units placed in a series constitute this maze (85). To complete one "run", a 22 h water-deprived mouse must choose between turning right or left to receive, at the end of the maze, a reward in the form of 40 µL of 5% sucrose and one "run" is considered complete. Then the mouse must shuttle back to the other end of the maze where it gets the same reward. There were 5 runs per session; one session per day. Quantitative measures of performance include numbers of left/right choice errors, "retrace errors"-episodes in which a mouse moves in the wrong direction (toward the end without a reward), time to complete a session (5 rewarded runs) and number of errorless runs per session (5 errorless runs is the highest performance level).

Histology and Immunolabeling
Mice were euthanized 80-90 min after behavioral experiments by an intraperitoneal injection of 200 mg/kg sodium pentobarbital ("Pental"). Brains were fixed by trans-cardial perfusion with ice-cold 4% paraformaldehyde containing 4% sucrose (pH = 7.4). For cellular tests we labeled ARP, the C-terminal peptide unique for AChE-R, using a rabbit polyclonal antibody [48]. Cholinergic neurons were identified by goat anti-choline-acetyltransferase (Chemicon, Temecula, USA). General AChE was labeled with goat anti-AChE (Santa Cruz, CA, USA, antibody N19). In co-localization studies, we used secondary antibodies: fluorescein (FITC)-labeled donkey anti-rabbit to visualize hAChE-R, donkey anti-goat for choline-acetyltransferase, and streptavidin-Cy3 to visualize general AChE. hAChE-R accumulating neurons were counted in hippocampal sub-regions CA1-2 (for brevity, hereafter referred to as CA1), CA3, and dentate gyrus (DG). In addition, hAChE-R accumulating neurons were counted in the striatum (in sampled rectangular fields measuring 420 × 680 µm) and in the red nucleus and the hypothalamus. Two to three coronal brain sections were sampled at an estimated distance of 2.8-3.3 mm posterior to bregma. Images of each hippocampal sub-region were acquired and analyzed using the AnalySIS software (SIS, Germany). In situ hybridization was essentially as described [48].

Measurements of Open Field Motor Activity
At the age of 5 weeks, all TgR mice displayed motor activity comparable to parent strain FVB/N mice, yet by 17 weeks, they segregated into two subpopulations. About 40% of transgenics (designated TgR-L) display locomotor activity comparable to that of FVB/N controls, whereas 60% express locomotor hyperactivity (TgR-H; Table 1). The hyperactivity of TgR-H mice was manifested as locomotor asymmetry which frequently developed into stereotypic circling ( Figure 1A,B), and repeating locomotor paths whereas both FVB/N and TgR-L mice readily explored and engaged in a variable succession of movements. Oral stereotypy (e.g., chewing, biting or licking) was not evident in TgR-H mice. Rearing, grooming, and the stretch-attend posture, all of which are normal components of exploratory behavior, were rare. Unlike TgR-H mice, TgR-L mice were similar to FVB/N control mice in all features (Table 1A and Supplementary Figure S1A-D). When testing behavior, we referred to the individual differences in the behavioral patterns of individual mice as significant and relevant, and separated mice into TgR-L and TgR-H as detailed below. When the mice were single-housed in cages containing a running wheel, TgR-H mice engaged in SB at the expense of wheel running and ran the wheel considerably less than did FVB/N control mice (daily means of 2455 ± 1350 as compared to 7370 ± 1570 counts, t-test (df = 10) = 2.4, p < 0.04). Thus, SB is not due only to an increased motor drive, which could be vented by use of the running wheel. Nevertheless, the circadian locomotion cycle was essentially normal except that TgR-H mice displayed higher activity during the first half of the dark phase (Supplementary Figure S1E,F).
Spontaneous stereotypic behavior (SB) in an open field was sought from weaning (3-4 weeks of age) to adulthood (2-12 months of age). Following 30 min of habituation in a holding cage, mice were placed for 5 min at the mildly stressful, brightly lit, center of a 40 × 40 cm open field with 20 cm high walls. Videotaped locomotor activity was first quantified as total time spent in locomotion. Rearing episodes were quantified to evaluate the extent of disruption of normal exploration under SB. Episodes in which the mouse has initiated locomotion ("locomotion events") were counted. An index of asymmetry was calculated as: left turns-half the total events/total events *100. Thus, perfect symmetry would yield a value of 0 and total asymmetry-50%.

Telemetric Measurement of Motor Activity
Telemetric activity measurements were performed as detailed previously [49], following intraperitoneal (ip) injection of ketamine-xylazine mixture (4.25, 0.15 mg/mouse) as anesthesia for transmitters implantation. Recordings lasted for 48 consecutive h starting at 7 am (beginning of light phase).

Vestibular Test
Mice were examined at the peak of their motor activity during the dark phase. Deficits in vestibular function were tested by holding mice by the tail, lifting and then lowering them over a metal cage top. Then, while the mouse was still, a tactile stimulus to the right or left shoulder and trunk was applied using a cotton swab, and orienting to the stimulus was noted. Efficient and symmetric head and limb placement, and symmetric head orienting to tactile stimuli were considered as evidence for normal vestibular functioning.

Forced Swimming
The TgR-H phenotype extended to situations involving motor behaviors other than locomotion. Thus, in a forced swim test, TgR-H, but not TgR-L, mice swam along the wall significantly more than FVB/N mice, which tended to sample alternative routes (e.g., swimming across the container). TgR-H mice further displayed a rigid pattern of swimming direction and circling asymmetry, to the same direction as in the open field ( Figure 1B). Mice were placed in a circular water container, 20 × 20 cm in diameter and depth. Following a 4 min swim at 25 • C, mice were taken out of the water and dried under a warm lamp. Total time spent swimming, number of swim episodes adjacent to the wall, number of swims across the container, number of circles and % asymmetry in the direction of swimming (calculated by the index of asymmetry) were quantified from video records. SB in this test was reflected by motor hyperactivity, reduced exploratory behavior and motor asymmetry.

Small RNA-Sequencing
In principle, we expected global changes in gene expression to primarily depend on the genomic composition of the tested mice. Therefore, we have used brain tissues from all of the TgR transgenics as compared to the control (wild type) mice from the same strain for the molecular analyses (microarray tests, RT-qPCR measurements and analyses thereof). Briefly, total RNA was extracted from murine brain tissues using the miRNeasy Mini Kit (Qiagen, 217004, Venlo, The Netherlands). Four TgR and three FVB/N mice were used for the prefrontal cortex (PFC) analyses, and three TgR and three FVB/N mice were used for the hippocampal preparations. RNA quality and quantity were measured using a Nanodrop 2000 (PeqLab/Thermo Scientific, Waltham, MA, USA). RNA integrity was tested using the 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) RNA 6000 Nano LabChip Kits (Agilent Technologies) which revealed a RIN values range of 8.6-9.3. 300ng of RNA were subjected to small-RNA sequencing. Libraries were generated using an RNA Library prep kit (NEBNext E7300 Multiplex Small RNA Library Prep Set for Illumina, Foster City, CA, USA) following manufacturer's instructions. Libraries were barcoded and sequenced on a NextSeq Series Sequencing System (HUJI Center for Genomic Technologies, Jerusalem, Israel) using an Illumina flow cell (Illumina 500 NextSeq High Output v2 Kit, FC-404-2005; Illumina). Quality control was performed using FastQC [50], version 0.11.8. Short RNA was aligned to the miRBase version 21 using miRExpress2.0 [51] with default parameters. Expression analysis was performed using the Bioconductor DESeq2 [52] software via R platform [53].

Transcriptomic Analysis
Prefrontal cortices (PFCs) and Striatal Caudate-Putamen regions (CP) from groups of four FVB/N (wt) and four TgR animals were dissected on ice. Total PFC and CP RNA was extracted using the RNeasy mini kit (Qiagen, Hulden, Germany). Transcriptome profiles of two independent pools of PFC and CP RNAs from male age and strain-matched FVB/N and TgR mice (3-4 animals in each pool) were each hybridized to four standard Affymetrix mouse arrays (MgU74Av2) with 12,450 different transcripts for a total of 4 microarrays. Microarrays were used as per manufacturer's instructions and the MIAME (www.mged.org) instructions. Analysis involved fluidic station FS-400, MAS 5.0 software and GeneArray 2500 scanner (Affymetrix). Transcripts were designated as absent or present according to the Affymterix manual where marginal (M) detection levels were counted as absent. A transcript was designated as changed if there was a robust change according to the Affymetrix manual (in the same direction) for at least 3 out of 4 comparisons (2 × 2 pools from each treatment). Data mining and graphs were carried out using the Spotfire program (Spotfire, Sommerville, MA, USA) and Excel. To identify the over-represented categories among those that were changed in the TgR brain, we performed Gene Ontologies (GO) classification using the freely available online GOstat tool [54].

Quantitative PCR (qPCR)
cDNA was prepared using the Quanta qScript mRNA cDNA Synthesis Kit (Quantabio, Beverly, MA, USA) according to the manufacturer's instructions and diluted 1:10 in double-distilled water prior to qPCR plate preparation. qPCR was performed in either 384-well or 96-well plates on CFX-384/96 machines (Bio-Rad), using Perfecta Sybr Green FastMix with low or No Rox (Quantabio) at a final well volume of 5 or 15 µL, respectively. β-actin and RPL-19 were used as housekeeping genes. Expression was calculated as ∆∆Ct values using the Bio-Rad CFX Maestro 1.1 Software Version 4.1.2433.1219.

Statistical Analyses
Histological data was subjected to ANOVA with a simple between-groups design comparing wild type FVB/N and TgR mice. Behavioral data, which included multiple tests conducted on each mouse, was analyzed using one-way ANOVA with repeated measures within each group (FVB/N and TgR) and between groups (FVB/N vs. TgR). Other statistical tests are noted where relevant.

TgR Mice Show Clear SB Characteristics
At the age of 17 and 30 weeks, TgR mice displayed exacerbated motor activity compared to age-matched FVB/N controls. Although there were relatively high individual behavioral differences among TgR as compared to FVB/N mice (Tables 2 and 3 and Supplementary Figure S1), most TgR mice exhibited symptoms that are indicative of SB, including motor hyperactivity, which was evident both in the forced swim ( Figure 1A and Supplementary Movie S1) and the open-field ( Figure 1B) paradigms, and was also observed under continuous automated telemetric recording. Moreover, TgR mice presented stereotypic and rigid behavioral patterns which accompanied their less than normal exploratory behaviors as observed in matched controls. For example, TgR mice exhibited stereotypic circling and swimming along the wall under forced swim stress, unlike FVB/N mice which tended to sample alternative routes (e.g., swimming across the container) ( Figure 1A). In the open-field test, TgR mice presented locomotor hyperactivity at the expense of normal exploratory behaviors, reflected by marked decline in rearing. They also showed less grooming and stretch-attend postures, which are normal components of exploratory behavior ( Figure 1B-D). The > sign denotes a value being significantly larger than the next mentioned value, the sign < denotes 'smaller than the next value'. Interestingly, when TgR mice were single-housed in cages containing a running wheel, they engaged in SB and ran the wheel considerably less than the FVB/N control mice (daily means of 2455 ± 1350 as compared to 7370 ± 1570 counts, t-test (df = 10) = 2.4, p < 0.04). Thus, the abnormal locomotion patterns were unlikely to be due to an increased motor drive, which could be vented by use of the running wheel. Additionally, TgR mice exhibited motor asymmetry, both in the forced swim ( Figure 1A) and the open-field ( Figure 1E) tests, where they displayed circling or locomotion asymmetry, turning to the same direction repeatedly. Additionally, in spite of essentially normal circadian locomotion cycle in both strains, activity levels and body temperature were both significantly higher in 2-month-old TgR mice during the first half of the dark phase ( Figure 1F,G). The contribution of vestibular dysfunction and middle ear pathology to TgR circling behavior was excluded, as TgR mice displayed symmetric head orienting when touched with a probe at the shoulder and turned symmetrically toward an approaching edge. Oral stereotypy (e.g., chewing, biting or licking) was not evident in the TgR mice (Supplementary Movie S2). We conclude that TgR mice displayed persistent SB characteristics, evident as circadian controlled motion hyperactivity, excessive circling behavior and motor asymmetry patterns which were subject to impaired control of their circadian behavior patterns.

Muscarinic Stimulation Suppresses the Motor Hyperactivity but not the Locomotor Asymmetry Component of SB
To test if the SB patterns were due to the cholinergic impairments of TgR mice, we injected these mice with the cholinergic agonist pilocarpine. This agent, at an intravenously injected dose of 25 mg/kg, profoundly suppressed motor behavior in both FVB/N and TgR mice (Figure 2 and Supplementary Table S1). Due to the higher baseline locomotion time in TgR mice, this yielded a significant interaction of pilocarpine and transgene. However, pilocarpine showed no effect on the locomotor asymmetry component of the SB profile, with no interaction of drug and transgene in this context (Supplementary Table S1), predicting that additional neurotransmission pathway(s) might be involved. Taken together, the prevalent behavioral differences between TgR and FVB/N mice, and the fact that these are manifested in multiple test conditions, suggested that these deficits did not reflect a low-level motor impairment. Rather, these findings could better be explained by a higher systematic deficit in neurotransmission pathway(s), which is what we tested next.

Muscarinic Stimulation Suppresses the Motor Hyperactivity but not the Locomotor Asymmetry Component of SB
To test if the SB patterns were due to the cholinergic impairments of TgR mice, we injected these mice with the cholinergic agonist pilocarpine. This agent, at an intravenously injected dose of 25 mg/kg, profoundly suppressed motor behavior in both FVB/N and TgR mice (Figure 2 and Supplementary Table S1). Due to the higher baseline locomotion time in TgR mice, this yielded a significant interaction of pilocarpine and transgene. However, pilocarpine showed no effect on the locomotor asymmetry component of the SB profile, with no interaction of drug and transgene in this context (Supplementary Table S1), predicting that additional neurotransmission pathway(s) might be involved. Taken together, the prevalent behavioral differences between TgR and FVB/N mice, and the fact that these are manifested in multiple test conditions, suggested that these deficits did not reflect a low-level motor impairment. Rather, these findings could better be explained by a higher systematic deficit in neurotransmission pathway(s), which is what we tested next.

hAChE-R Excess does not Disrupt Normal Cholinergic and Glutamatergic Pathways and is Inversely Correlated with ChAT Expression
Given the cholinoceptive nature of the hippocampus and its key role in controlling navigation behavior [55,56], we have first addressed this brain region in our tests. Quantitative PCR demonstrated massive selective increases of the stress-induced AChE-R variant in the hippocampi of 10 weeks old TgR mice compared to controls, with more modest increases of both the synaptic AChE-S variant and of AChE-R in the prefrontal cortex (PFC, Figure 3A-C). Extending this analysis to the protein level, cholinergic neurons in TgR mice appeared normal in number and size and were successfully labeled for both the leading AChE-S protein and the stress-induced AChE-R variant ( Figure 3D-F). Notably, TgR mice showed normal distribution of cholinergic ChAT-expressing neurons with symmetric distribution between the right and left hemispheres. However, numerous cholinergic neurons with intense ChAT expression did not express hAChE-R in the striatum, medial septum and diagonal band, indicating differential translation levels of this transcript in cholinergic neurons ( Figure 3G,H,K). This also ruled out loss or impaired function of cholinergic neurons and hemispheric differences as potential causes for SB. Importantly, intensely hAChE-R-labeled neurons were denser in the medial striatum (128 ± 19 per sampled field of 100 µm 2 ) than in the lateral striatum (25 ± 7 per sampled field, paired t-test (df = 7) = 6.8, p < 0.0005), opposite to the gradient of cholinergic activity in the striatum [57]. The nucleus basalis magnocellularis, brainstem and spinal cord (ventral horn) motor nuclei all included subsets of cholinergic neurons with less intense ChAT but pronounced hAChE-R expression ( Figure 3I,J,L). TgR mice further showed inter-hemispheric symmetry and normal distribution of tyrosine hydroxylase-positive dopaminergic neurons, without AChE-R. Thus, the brains of TgR mice presented somewhat skewed cholinergic profiles, but with otherwise regular histopathology.

hAChE-R Excess Does Not Disrupt Normal Cholinergic and Glutamatergic Pathways and Is Inversely Correlated with ChAT Expression
Given the cholinoceptive nature of the hippocampus and its key role in controlling navigation behavior [55,56], we have first addressed this brain region in our tests. Quantitative PCR demonstrated massive selective increases of the stress-induced AChE-R variant in the hippocampi of 10 weeks old TgR mice compared to controls, with more modest increases of both the synaptic AChE-S variant and of AChE-R in the prefrontal cortex (PFC, Figure 3A-C). Extending this analysis to the protein level, cholinergic neurons in TgR mice appeared normal in number and size and were successfully labeled for both the leading AChE-S protein and the stress-induced AChE-R variant ( Figure 3D-F). Notably, TgR mice showed normal distribution of cholinergic ChAT-expressing neurons with symmetric distribution between the right and left hemispheres. However, numerous cholinergic neurons with intense ChAT expression did not express hAChE-R in the striatum, medial septum and diagonal band, indicating differential translation levels of this transcript in cholinergic neurons ( Figure 3G,H,K). This also ruled out loss or impaired function of cholinergic neurons and hemispheric differences as potential causes for SB. Importantly, intensely hAChE-R-labeled neurons were denser in the medial striatum (128 ± 19 per sampled field of 100 µm 2 ) than in the lateral striatum (25 ± 7 per sampled field, paired t-test (df = 7) = 6.8, p < 0.0005), opposite to the gradient of cholinergic activity in the striatum [57]. The nucleus basalis magnocellularis, brainstem and spinal cord (ventral horn) motor nuclei all included subsets of cholinergic neurons with less intense ChAT but pronounced hAChE-R expression ( Figure 3I,J,L). TgR mice further showed inter-hemispheric symmetry and normal distribution of tyrosine hydroxylase-positive dopaminergic neurons, without AChE-R. Thus, the brains of TgR mice presented somewhat skewed cholinergic profiles, but with otherwise regular histopathology.

AChE-R Distribution in Motor-Regulating Brain Regions
Multiple subsets of neurons in several brain regions of TgR mice showed an excess of both transgenic hAChE-R and murine mAChE-R mRNA, accompanied by enhanced immunolabeling of the (h)AChE-R protein. Typically, the neuronal cytoplasm, nucleus, and dendrite(s), but not axons, were stained ( Figure 4A,B). In general, hAChE-R accumulation overlapped with that of the host gene, occurring in cholinoceptive regions that tend to express the primary synaptic AChE-S variant (e.g., cortex, hippocampus and striatum).
All TgR mice showed extensive hAChE-R expression in the neocortex, striatum, red nucleus and cerebellum, all of which are known to contribute to motor regulation [58,59], as well as in the anterior cingulate cortex, amygdala, hypothalamus and medial striatum, reported as being involved in the emotion-motor interface [60,61]. Some TgR mice also showed increased AChE-R expression in the striatum, lateral septum and most prominently in the hippocampus, structures known to participate in behaviorally-inhibitory pathways [62,63]. Neuronal activation was assessed by immunohistochemical detection of the immediate early gene c-fos, known to be increased under numerous insults [64,65]. Notably, the medial striatum, lateral septum and anterior cingulate cortex in TgR mice included fewer c-fos positive cells than in FVB/N mice ( Figure 4C and Table 4). Thus, the observed SB patterns could not be attributed to elevated c-Fos levels.
Biomolecules 2020, 10, x FOR PEER REVIEW 12 of 23 in TgR mice included fewer c-fos positive cells than in FVB/N mice ( Figure 4C and Table 4). Thus, the observed SB patterns could not be attributed to elevated c-Fos levels.

RNA-Seq Reveals Modified CholinomiRs in TgR Hippocampi
The hippocampus coordinates numerous behavioral patterns [66], and it readily reacts to altered cholinergic cues [67] and shows corresponding changes in miRNAs [68]. To explore such changes in the TgR hippocampus, we performed short RNA-sequencing of hippocampal RNAs from TgR mice and matched controls. A total of 36 miRNAs were differentially expressed (DE) in the TgR hippocampal RNA-seq datasets compared to FVB controls, and eight of these DE miRNAs target at least 5 cholinergic transcripts each ( Figure 5A; dataset deposited as GEO GSE144022) [55]). Notably, the downregulated cholinergic-targeted miRNAs formed a multi-level network with other miRNAs, considerably more complex than the network formed by the upregulated miRNAs ( Figure 5B). Those included miR-125b which targets, among other mRNAs both AChE-R and the vesicular ACh transporter (VAChT) [41], miR-370, the targets of which include the AKT signaling pathway [69] and BMP2 [70], and miR-204 which may inhibit the cholinergic-suppressible inflammation-associated NFkB signaling pathway [71]. The modified miRNAs further formed intricate networks, suggesting that they may each affect the other modified miRNAs ( Figure 5B). Intriguingly, no miRNAs were DE in the prefrontal cortex (PFC) of TgR mice. However, given the power of such network changes to exert neuro-modulation impacts across the brain [72], we proceeded by outlining corresponding changes in mRNA transcripts of the prefrontal cortex, which communicates with the hippocampus.

The TgR PFC Shows More Variable Cholinergic Transcripts than the Caudate Putamen
To further explore the cholinergic signaling-related brain routes, we compared the levels of cholinergic-associated transcripts between the striatal caudate-putamen (CP) and the PFC of FVB/N control and TgR mice (dataset deposited as GEO GSE31458). Intriguingly, the PFC showed more inter-individual variability for cholinergic and stress-related transcripts including CHRM3, CHRNA7, CNR1, DLG3, GRIN2A and SHANK3 compared to their CP expression patterns ( Figure 6). The greater individual diversity of cholinergic-related and stress-associated genes in the PFC compared to the CP could be functionally relevant for the SB phenotype.
Biomolecules 2020, 10, x FOR PEER REVIEW 14 of 23 control and TgR mice (dataset deposited as GEO GSE31458). Intriguingly, the PFC showed more inter-individual variability for cholinergic and stress-related transcripts including CHRM3, CHRNA7, CNR1, DLG3, GRIN2A and SHANK3 compared to their CP expression patterns ( Figure  6). The greater individual diversity of cholinergic-related and stress-associated genes in the PFC compared to the CP could be functionally relevant for the SB phenotype. Higher individual variability can be seen in the expression profiles between the PFC and the CP, especially in the SB-prone TgR strain.

Suppressed Glutamatergic Neurotransmission-Related Genes in the TgR PFC
We predicted that the miRNA changes in the TgR hippocampus reflect its cholinoceptive features; to test this hypothesis, we sought corresponding differences in miRNA-targeted coding transcripts within other brain regions that modulate the hippocampus via cholinergic signaling. Given the PFC's capacity to send stimuli-related messages to the hippocampus [74], we performed PFC transcriptomic analyses. We compared the FVB/N vs. FVB/N (C1/C2), TgR vs. TgR (R1/R2) and the four possible combinations of FVB/N vs. TgR profiles ( Figure 7A). The two FVB/N preparations yielded highly similar profiles ( Figure 7B, C2 vs. C1), with merely five genes showing a change. In contrast, the two TgR microarrays displayed higher variability, with 106 transcripts designated as DE ( Figure 7B, R1 vs. R2). In comparison, PFC-expressed genes in TgR mice differed from those of FVB/N mice ( Figure 7B, R1 vs. C1, R2 vs. C2 or the reciprocal comparisons) by 132, 388, 143, and 428 transcripts with robust changes (also, see Figure 7B for correlation coefficients (CC) between the absence/presence calls of each of the individual arrays), especially in glutamate-related functions.
The transcriptomic tests of coding mRNAs, followed by Gene Ontology analysis (GO [54]) of the corresponding biological process and molecular function ontologies revealed several significantly downregulated cellular and organism level categories in the TgR PFC ( Figure 7A). Most conspicuous were changes in transcription-related processes (e.g., DNA binding transcription factors, trans-acting transcription activators/repressors, hormone receptors and chromatin remodeling). Two major affected categories included synaptic transmission (p < 2 × 10 -5 , 12 of 62 genes changed) and ion transport/channel activity-related processes (p < 9 × 10 -4 , 18 of 238 changed). Suppressed molecular function categories included ionotropic glutamate receptor activity (1 × 10 -3 , 5/13) and glutamategated ion channel activity ( Figure 7B). Within the category of synaptic transmission, we noted

Suppressed Glutamatergic Neurotransmission-Related Genes in the TgR PFC
We predicted that the miRNA changes in the TgR hippocampus reflect its cholinoceptive features; to test this hypothesis, we sought corresponding differences in miRNA-targeted coding transcripts within other brain regions that modulate the hippocampus via cholinergic signaling. Given the PFC's capacity to send stimuli-related messages to the hippocampus [74], we performed PFC transcriptomic analyses. We compared the FVB/N vs. FVB/N (C1/C2), TgR vs. TgR (R1/R2) and the four possible combinations of FVB/N vs. TgR profiles ( Figure 7A). The two FVB/N preparations yielded highly similar profiles ( Figure 7B, C2 vs. C1), with merely five genes showing a change. In contrast, the two TgR microarrays displayed higher variability, with 106 transcripts designated as DE ( Figure 7B, R1 vs. R2). In comparison, PFC-expressed genes in TgR mice differed from those of FVB/N mice ( Figure 7B, R1 vs. C1, R2 vs. C2 or the reciprocal comparisons) by 132, 388, 143, and 428 transcripts with robust changes (also, see Figure 7B for correlation coefficients (CC) between the absence/presence calls of each of the individual arrays), especially in glutamate-related functions.
The transcriptomic tests of coding mRNAs, followed by Gene Ontology analysis (GO [54]) of the corresponding biological process and molecular function ontologies revealed several significantly downregulated cellular and organism level categories in the TgR PFC ( Figure 7A). Most conspicuous were changes in transcription-related processes (e.g., DNA binding transcription factors, trans-acting transcription activators/repressors, hormone receptors and chromatin remodeling). Two major affected categories included synaptic transmission (p < 2 × 10 −5 , 12 of 62 genes changed) and ion transport/channel activity-related processes (p < 9 × 10 −4 , 18 of 238 changed). Suppressed molecular function categories included ionotropic glutamate receptor activity (1 × 10 −3 , 5/13) and glutamate-gated ion channel activity ( Figure 7B). Within the category of synaptic transmission, we noted changes in CAMK2, associated with the control over long-term potentiation [75], which is indeed modified in the TgR mouse [76]; VAMP2, involved with synaptic transmission efficacy [77] and WNT7B, regulating the WNT signaling pathway [78] were also modified. The decrease in ion channel activity involved reductions in calcium, potassium, chloride and sodium channels, GABA transporter (GABRA3, [79,80]), the ATPase ATP 6V1A [81] and glutamate decarboxylase (GAD2, [82]) in addition to the glutamate receptor NR4A2 gene [83]. These are all compatible with our findings of reduced glutamate receptor levels in cultured hippocampal neurons overexpressing AChE [84,85] and reciprocal to the changes induced by excess AChE-S in PFC neurons. The changed categories are noted in Figure 7B, and the changed transcripts are listed in Supplementary Table S2. Biomolecules 2020, 10, x FOR PEER REVIEW 15 of 23 changes in CAMK2, associated with the control over long-term potentiation [75], which is indeed modified in the TgR mouse [76]; VAMP2, involved with synaptic transmission efficacy [77] and WNT7B, regulating the WNT signaling pathway [78] were also modified. The decrease in ion channel activity involved reductions in calcium, potassium, chloride and sodium channels, GABA transporter (GABRA3, [79,80]), the ATPase ATP 6V1A [81] and glutamate decarboxylase (GAD2, [82]) in addition to the glutamate receptor NR4A2 gene [83]. These are all compatible with our findings of reduced glutamate receptor levels in cultured hippocampal neurons overexpressing AChE [84,85] and reciprocal to the changes induced by excess AChE-S in PFC neurons. The changed categories are noted in Figure 7B, and the changed transcripts are listed in Supplementary Table 2.

Validating Brain Region-Dependent Changes in Cholinergic and SB-Related Transcripts
To validate the PFC microarray findings and seek their relevance to the over-expressed AChE-R and the modified hippocampal miRNAs, we performed qPCR tests of selected relevant mRNA transcripts which are characteristic of the predictably modified pathways in hippocampi and PFC tissues from TgR and control mice. Notably, both the muscarinic receptor CHRM3 shown to be essential for REM sleep [86] and the attention-related CHRM5 [87] showed hippocampal but not PFC differences in TgR mice compared to FVB/N controls and to unmodified transcript controls ( Figure 8). Additionally, we noted massive albeit bidirectional hippocampal changes in the autism-related transcripts of Grin2b, PTEN and SHANK3 [88][89][90][91], predicting behavioral differences; but not in the inflammation-associated CHRNA7, indicating limited relevance for brain inflammation. Notably, both miR-370-3p and miR-204-5p are known controllers of brain glia, indicating their potential relevance to the observed changes in the target transcripts. We conclude that perturbation of cholinergic/glutamatergic networks may induce stress-induced miRNA-mediated SB.

Discussion
TgR mice presented locomotor SB that appeared to be a repetition of pre-potent defensive motor responses to threat. Specifically, the running component paralleled running response under threat [43,45,92], which was consistently accompanied by the TgR hyper-locomotion. This is compatible with the reported linkage between emotionality and lateralization of motor behavior [93], which may be associated in TgR mice with locomotor-asymmetry and circling. Thus, we propose that in wild type mice, escape-mode defensive pre-potent motor responses may be subjected to inhibitory control; however, in the TgR mouse, these responses are released from inhibition as revealed by c-fos reduction in executive brain regions of the TgR emotional-motor interface.
Seeking the neuronal network origin of SB and its underlying molecular mechanisms, we studied short hippocampal and long PFC RNA transcripts in SB-expressing TgR mice over-expressing the stress-inducible soluble AChE-R variant, which accelerates acetylcholine degradation and suppresses cholinergic reactions. The transcriptomic analysis was performed on a mixed group of TgR mice with predicted inter-individual variability in their SB profiles. This indicates under-estimation of the impact of this transgene on those TgR mice presenting high level SB. We found TgR hippocampi to show 36 DE miRNAs, with 8 of those targeting cholinergic-related transcripts. Moreover, the PFCs of TgR mice, which showed no DE miRNAs, displayed 428 long DE mRNA transcripts, with a conspicuous decline of glutamatergic-related pathways (p < 1 × 10 −3 ) and pronounced changes in autism-related transcripts compared to FVB/N mice. We further noted excess of C-fos at motor behavior-regulating brain regions and of immune-labeled AChE-R in SB-regulating basal ganglia, limbic brain nuclei and the brain stem. This labeling pattern accompanied the SB profiles of TgR mice. Combined with the changes we found in hippocampal miRNAs and selected glutamatergic-related PFC transcripts, these findings may indicate miRNA-mediated perturbation of the delicate balance between cholinergic/glutamatergic networks as accompanying the impaired inhibitory control over defensive motor behavior which is causally involved in SB.

Modified Hippocampal miRNAs in TgR Mice with Altered Cholinergic Activities
Over-expression of the stress-inducible soluble AChE-R variant predicts excessive ACh hydrolysis and correspondingly impaired cholinergic transmission in several motor-related brain regions. This variant is over-expressed in the mammalian brain under stress responses (reviewed in [33]). Therefore, we assumed that modified expression of cholinergic miRNAs might be involved. To test this prediction, we performed non-biased profiling of the short RNAs in TgR hippocampi compared to wild type FVB/N controls. This analysis indeed revealed 36 DE miRNAs, 8 of which affect cholinergic function ('CholinomiRs', [94]). Those included miR-125b which targets both AChE-R and the vesicular acetylcholine transporter (VAChT) [41], indicating that transgenic over-expression of AChE-R induced miRNA-mediated suppression of both the transgene itself and the cholinergic network that it interrupts in TgR brains. Other miRNAs that were affected are miR-370, which targets the AKT signaling pathway [69] and BMP2 [70]. Changes were also observed in miR-204, which inhibits the NFkB signaling pathway that is associated with inflammation and is suppressible by cholinergic inputs [71]. Altogether, the impaired miRNA signaling in the TgR hippocampi suggests a causal involvement of the AChE-R variant in both the cholinergic pathway in general, and the SB profile of these mice in particular.

Prefrontal Cortex and Striatal Transcriptomics of TgR vs FVB/N Mice
The hippocampal CholinomiR changes and the neuro-modulatory role of cholinergic neurotransmission [33,95] suggested the existence of further modifications in other brain regions. Notably, the PFC showed larger variability in cholinergic-related transcripts than the striatal caudate-putamen. Given the stress-related links of PFC functioning with the hippocampus, we proceeded by transcriptomic profiling of long RNAs in the PFC. This analysis revealed greater variability in PFC transcripts from TgR mice compared to their parent FVB/N strain. Specifically, we found the categories of both signal transduction and protein kinase C activation to be modified in the TgR PFC, which indicates that AChE-R may play an active role in modulating motor responses to stress. Additionally, the neuro-anatomical distribution of AChE-R overlapped with that of its partner kinase [96], consistent with our previous studies showing intensified, PKC-mediated LTP responses in TgR mice [76,97]. Therefore, the variable neuroanatomical distribution of transgenic AChE-R may support the gene-environment interaction(s) accounting for the TgR SB phenotype. Furthermore, the altered synaptic regulation by AChE-R occurred in specific pathways involved with stress management, consistent with the overlap between AChE-R's neuro-anatomical distribution and the distribution of stress-activated neuronal populations.

Implications for SB Studies
Several studies support glutamatergic causation of SB. This centers on anatomical substrates within the striatum [16,98], and possibly other brain regions in the motor hierarchy. Such conclusions are mainly derived from clinical studies. For example, Tourette's syndrome abnormalities involve interactions between the striatum and other brain regions [99]. However, although animal model studies of spontaneous cage stereotypies [17] associated captivity stress with SB, the anatomical substrate of SB in these models has not been elucidated yet. Our current study bridges the anatomical distribution of stress signals with the corresponding transcriptomic signals and the SB phenotype, suggesting that stress-induced changes in AChE-R distribution and in neuronal activation in the behavioral-motor interface may yield SB. Importantly, SB is not a single class of purposeless movement. Rather, we find that stress-related cholinergic signaling intensifies pre-potent escape-related locomotion while other motor behaviors such as rearing are reduced. That TgR mice display SB patterns under safety supports the notion of stress-induced restriction of their normal behavioral repertoire. Different types of stress may hence engage distinct pre-potent defensive motor responses and result in different forms of SB, with glutamate involved in some but not necessarily all aspects of SB.

Clinical Relevance to Human SB
The elevated stress signaling [100] and reduced inhibitory tone in executive brain regions of the emotional motor interface [101,102] in Tourette's syndrome parallel our observations in TgR mice. The frequency and intensity of tics (a stereotypic motor pattern) displayed by Tourette's patients increase under situations perceived as threatening [12,13]. In addition, striatal cholinergic activity is involved in both human SB and psychosis. Along with the cingulate and orbito-frontal cortex, the caudate nucleus serves as the major substrate for the effects of stress on glutamatergic pathways [103]. Both the caudate and cingulate/prefrontal cortices are involved in the TgR model. Nevertheless, the movement sequences that comprise SB differ among mouse models, caged animals or psychiatric patients. While locomotor hyperactivity and circling behavior occur both in some psychiatric disorders [104] and in TgR mice, several other SB movement sequences appear in psychiatric and developmental disorders [8] but not in TgR mice. Nevertheless, the clinical relevance of cholinergic stress signaling to SB in psychiatric disorders is supported by our finding that in human blood tests, AChE-R levels increase with state anxiety [105]. We conclude that SB reflects a multitude of pre-potent movement sequences that escape inhibitory control, that are often stress-related and that vary depending on the neural substrates involved in the given psychiatric disorder. While we are aware of the limitations of this study, which largely presents an association phenomenon rather than a conclusive mechanism of action, it may be naïve to link the complex SB phenomenon to a single splice variant of a single enzyme; nevertheless, the broad distribution of cholinergic stress signaling points to AChE-R as an important candidate to investigate in psychiatric disorders with stereotypic behaviors.