Distinct contributions of three GABAergic interneuron populations to a mouse model of Rett Syndrome

Background Rett Syndrome is a devastating neurodevelopmental disorder resulting from mutations in the gene MeCP2. MeCP2 is a transcriptional regulator active in many cell types throughout the brain. However, mutations of MeCP2 restricted to GABAergic cell types largely replicate the behavioral phenotypes associated with mouse models of Rett Syndrome, suggesting a key role for inhibitory interneurons in the pathophysiology underlying this disorder. Methods We generated conditional deletions of MeCP2 from each of three major classes of GABAergic interneurons, the parvalbumin (PV), somatostatin (SOM), and vasoactive intestinal peptide (VIP)-expressing cells, along with a pan-interneuron deletion from all three GABAergic populations. We examined seizure incidence, mortality, and performance on several key behavioral assays. Results We find that each interneuron class makes a contribution to the seizure phenotype associated with Rett Syndrome. PV, SOM, and VIP interneurons made partially overlapping contributions to deficits in motor behaviors. We find little evidence for elevated anxiety associated with any of the conditional deletions. However, MeCP2 deletion from VIP interneurons causes a unique deficit in marble burying. Furthermore, VIP interneurons make a distinct contribution to deficits in social behavior. Conclusions We find an unanticipated contribution of VIP interneuron dysfunction to the MeCP2 loss-of-function model of Rett Syndrome. Together, our findings suggest a complex interaction between GABAergic dysfunction and behavioral phenotypes in this neurodevelopmental disorder.


Introduction
Rett Syndrome is a severe neurodevelopmental disorder caused by mutations in the gene encoding methyl-CpG binding protein 2 (MeCP2). Children affected by Rett develop normally during the initial year of postnatal life but rapidly regress thereafter, showing loss of motor skills and language and developing cognitive impairments, ataxia, respiratory problems, and stereotyped hand movements (1). Extensive previous work has shown that many Rett-associated phenotypes are replicated in MeCP2 loss-offunction mouse models (2)(3)(4)(5)(6). Rett Syndrome is strongly associated with seizure (1,7,8), suggesting a possible role for GABAergic dysregulation in the pathophysiology underlying these symptoms. Indeed, previous work in mice found that conditional mutations of MeCP2 restricted to GABAergic neurons recapitulate most of the observed phenotypes (2,5) whereas rescue of MeCP2 solely in GABAergic neurons ameliorates many phenotypes (9), suggesting a key role for GABAergic dysregulation in Rett Syndrome.
One major challenge in exploring GABAergic dysfunction in Rett Syndrome is the diversity of inhibitory interneurons, which can be subdivided into distinct classes with different physiology, synaptic targets, and molecular markers. The functions of three major classes of GABAergic cells have recently been explored in depth: fast-spiking basket cells that co-express the calcium-binding protein parvalbumin (PV) and target the cell bodies of excitatory neurons, low-threshold spiking cells that co-express the peptide somatostatin (SOM) and target the distal dendrites of excitatory neurons, and sparse interneurons that co-express vasoactive intestinal peptide (VIP) and mainly target other inhibitory interneurons. These three classes are thought to play distinct roles in regulating local circuit activity throughout the brain (10) and may thus make different contributions to the pathophysiology underlying Rett Syndrome.
Recent work has suggested that MeCP2 loss of function mutations in PV and SOM interneurons may have distinct consequences. Conditional deletion of MeCP2 from PV interneurons caused motor, sensory, cognitive, and social deficits, whereas deletion from SOM interneurons resulted in increased behavioral stereotypy and late-onset seizures (5). However, Cre-dependent MeCP2 deletion in PV and SOM cells is not well matched developmentally, as the SOM promoter is active during early embryonic development but the PV promoter is not active until after the first two postnatal weeks (11). Furthermore, nothing is known about the relative contribution of VIP interneurons to this disorder.
Here we directly compared the impact of MeCP2 loss-of-function in PV, SOM, and VIP interneurons. We generated conditional mutations of MeCP2 in each interneuron class and compared these with a conditional pan-interneuron mutation that includes all three interneuron classes and drives embryonic deletion. To identify the distinct contributions of each interneuron class, we assayed seizure incidence and mortality, locomotor and anxiety phenotypes, and social behavior. We find that PV, SOM, and VIP interneuron deficits may each contribute to seizure incidence following MeCP2 mutation. All three interneuron classes likewise contribute to general locomotor deficits. However, MeCP2 deletion from VIP, but not PV or SOM, interneurons caused deficits revealed by marble burying.
Furthermore, MeCP2 deletion from VIP interneurons made a unique contribution to social interaction impairments. Overall, our findings suggest an unanticipated role for VIP interneuron dysfunction in the MeCP2 model of Rett Syndrome. However, partially overlapping contributions from the three major interneuron classes suggest that these GABAergic populations may play intersecting roles in neurodevelopmental pathophysiology.
LFP data were recorded with a bandpass 0.1-9000Hz filter.

Seizure detection
All mice were handled for at least 10 min each day throughout the study. During the daily handling regime, the mice were assessed for seizures. If a seizure did occur, the mouse was immediately returned to its home cage. Seizures were defined as events reaching Racine scale levels 4 or 5, with animals exhibiting rearing and forelimb clonus or rearing, forelimb clonus, and falling.

Morbidity analysis
After weaning at P21, all mice were monitored every day throughout the study. All deaths were noted, and animals were tracked daily until P500.

Behavioral analysis
The open field, elevated plus maze, marble-burying, and sociability assays were performed under low-level (20-25 lux) red illumination. In each assay, mice were given 15 minutes to acclimate to the behavioral assay room. In all cases, the researcher was blind to the genotypes of the mice until after all behavioral data were scored.

Open field:
The open field assay was performed in a 30cm square box divided into nine quadrants. Custom software (Labview) was used to control a camera recording the mouse's path in the box. At the start of the session, the mouse was placed in the center quadrant of the box and allowed move freely for twenty minutes. After the time elapsed, the mouse was returned to the home cage and the box was cleaned for the next mouse.
ImageJ software was used to analyze the total distance traversed by the mouse and the number of times the mouse crossed the center quadrant during the twenty-minute period. The mouse was deemed to have entered the center quadrant when all four feet were inside the quadrant and to have left the quadrant when all four feet were outside.
Elevated Plus Maze: Custom Labview software was used to control a camera recording the mouse's locomotion on the maze. At the beginning of the session, the mouse was placed in the center of the maze and allowed to freely move on either arm for five minutes. At the end of the session, the mouse was returned to the home cage and the maze was cleaned for the next mouse. Video recordings of mouse behavior were handscored to determine the amount of time spent in the open and closed arms of the maze.
Marble Burying: 12 marbles were evenly placed in a cage with 1 inch of clean bedding.
Custom Labview software was used to control a camera recording the mouse's activity in the cage. The mouse was placed in the center of the cage with the marbles and allowed to explore the cage for 20 minutes. At the end of the session, the mouse was returned to the home cage and the marbles were cleaned with a 10% bleach solution.
The proportion of marbles buried was analyzed in ImageJ using the "analyze particles" function to compare the initial and final exposed surface area of the marbles.
Sociability: The sociability apparatus was divided into three equal areas with Plexiglas dividers, each with an opening allowing access to neighboring chambers. Custom Labview software was used to control a camera recording the mouse's activity in the chamber. An unfamiliar age-and sex-matched conspecific mouse (reared in separate cages, C57BL/6 genotype) was placed into a small cylindrical holding cage in one side of the chamber and an identical empty holding cage was placed in the other side. The location of the conspecific was randomly varied across trials. At the beginning of the session, the test mouse was placed in the central chamber and habituated to the central chamber for ten minutes. The dividers were then removed to allow the mouse to freely move among all the chambers for ten additional minutes. At the end of the session, the mice were returned to their home cages and the apparatus was cleaned. Video recordings of the mouse's behavior were hand-scored to determine the amount of time spent in each of the three partitions and the number of approaches the test mouse made to the conspecific and the empty holding cage. An approach was defined as the test mouse coming within a 5-centimeter radius of a cage or making contact with a cage.

Statistical analysis
Paired and unpaired non-parametric tests generated in GraphPad Prism (version 7 for Mac; San Diego CA) were used throughout the study due to non-normal data distributions. Exact p values are reported for all tests.

MeCP2 expression in PV, SOM, and VIP interneurons
To confirm that MeCP2 is expressed in the three major populations of GABAergic interneurons, we co-stained sections of primary visual cortex from adult mice with antibodies for interneuron markers and MeCP2 (Fig. 1). As reported previously (5), nearly all PV and SOM interneurons expressed MeCP2. In addition, ~80% of VIP interneurons expressed MeCP2, suggesting a previously unappreciated role for this signaling pathway in VIP interneuron function.
To determine the relative contributions of each interneuron class to behavioral phenotypes observed following MeCP2 deletion from all interneurons (2)

Seizure incidence following MeCP2 deletion
Previous work identified a characteristic seizure phenotype resulting from MeCP2 deletion in the brain (2) and found that MeCP2 deletion from SOM-expressing cells may confer a late-onset tendency to seizure (5). We therefore evaluated the incidence of seizure in each of the three interneuron-specific MeCP2 deletion lines from weaning (P21) through late adulthood (P500). We compared the impact of MeCP2 deletion from PV, SOM, or VIP interneuron populations with simultaneous deletion from all three interneuron classes using the Dlx5/6-Cre line. We further compared the interneuron deletion mice with MeCP2 f/y littermate controls. To identify the relative impact of MeCP2 deletion from GABAergic cells compared to loss of MeCP2 in all cells, we also compared seizure incidence in the interneuron-specific deletion mice with that in male mice carrying a complete knockout of the MeCP2 gene (MeCP2 -/y ).
In a subset of SOM mutants, we performed electrophysiological recordings in the cortex and hippocampus of awake animals and confirmed spontaneous seizure activity in each animal at ~P200 (n = 4 mutants; Fig. 2C). These results suggest that MeCP2 deletions from PV, SOM, and VIP populations each contribute a degree of seizure pathophysiology.
In addition to assaying the onset of seizure activity, we compared the incidence of death associated with seizure in each mouse line (Fig. 2D-E Fig. 2E), suggesting that GABAergic interneuron dysfunction does not entirely account for the early seizure phenotype in the MeCP2 deletion model. These results suggest that although MeCP2 deletion specifically from PV, SOM, and VIP interneurons is associated with elevated seizure incidence in mature animals, only seizure activity following SOMspecific mutations is accompanied by increased mortality rates.

Behavioral phenotypes associated with interneuron-specific MeCP2 deletion
We next characterized the contributions of MeCP2 deletions in each interneuron population to behavioral deficits. We found that locomotor activity in the open field test was impaired in pan-interneuron Dlx5/6 mutants (4988 ± 648.8 cm; n = 7; p = 0.011) and  Together, these data suggest that VIP interneurons, and potentially SOM interneurons, may contribute to the deficits in sociability behavior caused by global MeCP2 deletion.

Discussion
Our results reveal a complex pattern of GABAeric contributions to behavioral phenotypes in the mouse model of Rett Syndrome and an unanticipated role for MeCP2 function in VIP interneurons. Based on previous characterizations of the MeCP2 deletion model (1,2,5,17,22), we examined interneuron contributions to three major categories of neural dysregulation: seizure and mortality, motor and anxiety behaviors, and social behaviors. We find evidence for overlapping contributions of all three major interneuron classes to seizure and motor dysfunction, but a unique role for MeCP2 deletion from VIP interneurons in social deficits.
Patients with Rett Syndrome exhibit respiratory impairments and seizure (1), and these phenotypes are replicated in mouse models following deletion of MeCP2 from all GABAergic cells (2,5). We compared seizure and death incidence across interneuron cell type-specific MeCP2 loss of function mutations and a pan-interneuron loss of function. We found that MeCP2 deletion from each interneuron class increased the risk of seizure to some degree. In agreement with previous work (5), we found that MeCP2 loss of function in SOM interneurons conferred a late seizure phenotype beginning around postnatal day 200. However, conditional mutations of MeCP2 in the PV and VIP interneuron populations also caused modest increases in late seizure incidence.
Seizure incidence was not associated with elevated mortality rates in the PV and VIP mutants, suggesting that the seizures were less severe than those in the SOM mutants.
We likewise observed a low rate of late-onset seizure in the MeCP2 f/y control animals, indicating that the decrease in MeCP2 levels associated with the conditional allele (17) may be associated with epileptogenic consequences in addition to some mild behavioral phenotypes.
Distinct interneuron populations had different impacts on seizure incidence and overall mortality. Deletion of MeCP2 in the pan-interneuron Dlx5/6 mutants led to significant advances in seizure onset and mortality as compared to the SOM-specific mutants, suggesting a possible contribution from other interneuron populations such as PV cells. Because the PV promoter does not become active until ~2 weeks after birth, MeCP2 deletion from cells in the PV mutants likely did not occur until mid-adolescence (22). However, because the Dlx5/6 promoter turns on during early embryonic development in cells arising in both the medial and caudal ganglionic eminences (10,11,15), it is possible that early deletion of MeCP2 from PV interneurons in the Dlx5/6 animals promotes early seizure onset, as suggested for other developmental dysregulations of PV interneurons (15,23,24). Alternatively, deletion of MeCP2 from all three interneuron classes may have a supralinear impact on the early development of pathological activity patterns and respiratory impairments. A previous study did not observe seizure or early mortality phenotypes in the Dlx5/6 mutants, possibly due to methodological differences (2). MeCP2 knockout animals had significantly earlier seizure onset and mortality than the Dlx5/6 mutants, in agreement with previous work suggesting that loss of MeCP2 in excitatory neurons also contributes to seizure (25) and that respiratory impairments in the MeCP2 knockouts may increase early mortality (3,4).
We further compared locomotor and anxiety phenotypes in each line and found that MeCP2 mutation in all three interneuron classes resulted in deficits in general motor function in the open field task, but only MeCP2 mutations in VIP interneurons led to deficits in marble burying, a task that is susceptible to altered anxiety and OCD-like behaviors as well as changes in fine motor function. None of the MeCP2 loss of function mutations resulted in anxiety-related phenotypes in the elevated plus maze, in agreement with previous work (2,5). These data suggest that VIP interneuron dysfunction contributes to the motor deficits observed following broader deletion of MeCP2.
Abnormal or reduced social behavior is a hallmark of many autism spectrum disorder models, and has previously been shown in mice lacking MeCP2 in all GABAergic populations (2,5). We found that pan-interneuron MeCP2 deletion mice exhibited a reversal of normal social preferences in the sociability assay, preferring an empty chamber to one containing a conspecific. Notably, MeCP2 deletion from VIP interneurons fully replicated this phenotype. In contrast, PV-specific deletion had no effect. Intriguingly, SOM-specific deletion led to loss of any social preference, suggesting a potential contribution of both VIP-and SOM-expressing cells to deficits in social behavior following pan-interneuron MeCP2 deletions. In agreement with results of a previous study using a partition test, which assays social preference between two conspecifics, we found that PV mutants showed elevated social interactions with novel conspecifics (5).
Our results comparing the PV and SOM interneuron contributions to seizure incidence and behavioral deficits are in agreement with some aspects of previous studies. However, we found very little impact of MeCP2 deletion in the PV-Cre mice.
The PV-Cre line used here largely expresses in PV interneurons, along with some thalamocortical projection neurons (13,26). In comparison, the PV-2A-Cre line used in some previous work (5) also expresses Cre in a subset of pyramidal neurons and additional thalamic nuclei (27). These differences may contribute to the more severe behavioral phenotypes and early mortality previously observed in the PV-2A-Cre line (5).
Other work examining MeCP2 deletion in the PV-Cre line likewise found only mild behavioral phenotypes (22). However, as noted above, our findings from the PV-Cre line do not preclude a substantial contribution of embryonic MeCP2 deletion from PV interneurons to Rett Syndrome phenotypes.
Our findings suggest a partially overlapping set of contributions of different interneuron populations to seizure and behavioral deficits following deletion of MeCP2.

E.
Mean age at death for each group shown in D. Death age was set to P500 for animals surviving beyond day 500. *p < 0.05, ***p < 0.0001