Brain volumetric alterations accompanied with loss of striatal medium-sized spiny neurons and cortical parvalbumin expressing interneurons in Brd1+/− mice

Schizophrenia is a common and severe mental disorder arising from complex gene-environment interactions affecting brain development and functioning. While a consensus on the neuroanatomical correlates of schizophrenia is emerging, much of its fundamental pathobiology remains unknown. In this study, we explore brain morphometry in mice with genetic susceptibility and phenotypic relevance to schizophrenia (Brd1+/− mice) using postmortem 3D MR imaging coupled with histology, immunostaining and regional mRNA marker analysis. In agreement with recent large-scale schizophrenia neuroimaging studies, Brd1+/− mice displayed subcortical abnormalities, including volumetric reductions of amygdala and striatum. Interestingly, we demonstrate that structural alteration in striatum correlates with a general loss of striatal neurons, differentially impacting subpopulations of medium-sized spiny neurons and thus potentially striatal output. Akin to parvalbumin interneuron dysfunction in patients, a decline in parvalbumin expression was noted in the developing cortex of Brd1+/− mice, mainly driven by neuronal loss within or near cortical layer V, which is rich in corticostriatal projection neurons. Collectively, our study highlights the translational value of the Brd1+/− mouse as a pre-clinical tool for schizophrenia research and provides novel insight into its developmental, structural, and cellular pathology.

with genes important in neurodevelopment and mental health 8 . In accordance, BRD1 has been implicated with both brain development 9 and susceptibility to mental illness [10][11][12] , including gene-wise significant association in the currently largest SZ genome-wide association study (GWAS) mega-analysis 4 . A SZ case carrying a disruptive BRD1 mutation has furthermore been identified 13 . We have previously reported behavioral, molecular, and neurochemical changes with overall translational relevance to SZ in genetically modified mice with mono-allelic inactivation of Brd1 (Brd1 +/− mice) 14,15 . These include cognitive and social deficits, decrease in cortical parvalbumin immunoreactive interneurons, hypersensitivity to psychotomimetic drugs coupled with monoaminergic dysregulation and differential cortical and striatal expression of genes that are enriched for SZ risk 14,15 . In this research, we investigate the neuropathological impact of reduced Brd1 expression in mice using postmortem 3D MR imaging coupled with histology, immunostaining and regional mRNA marker analysis, and we relate our data to the most recent clinical findings in SZ.

Results
Brain structural changes in Brd1 +/− mice. To assess whether reduced Brd1 expression is associated with gross neuroanatomical changes, we subjected adult Brd1 +/− (n = 10) and WT (n = 9) mice to high-resolution 3D MR imaging, in combination with deformation-based morphometry analyses. No differences were seen in total grey matter, white matter or overall regional volumes (Table 1). However, Brd1 +/− mice were characterized by nominally significant volume reductions in several, particular, sub-cortical brain structures. Most significantly, a ~9% volume reduction was observed in striatum (Table 1 and Fig. 1a, p = 0.01), which was mainly driven by a ~10% reduction in the caudate putamen (CPu) ( Table 1 and Fig. 1b, p < 0.01). The adjacent globus pallidus was reduced by ~7% (p = 0.03), amygdala by ~5% (p = 0.01), and a tendency for decreased hippocampal volume was observed (p = 0.07). In support of these findings, striatal and amygdaloid volume reductions were observed in both hemispheres independently (CPu: left, p = 0.04; right, p = 0.05; amygdala: left, p = 0.03; right, p = 0.03). Further supporting volume loss, a stereological estimate in fixed tissue from an independent group of mice (7 WT and 6 Brd1 +/− ) showed a comparable ~11% total CPu volume reduction in Brd1 +/− mice compared to WT mice ( Fig. 1c, p < 0.01, Table 2). Although total brain volume did not differ between Brd1 +/− and WT mice (Table 1), a moderate to strong positive correlation was observed between total brain volume and volumes of most regions, structures and substructures excluding the ventricles (Fig. 1d). Furthermore, a positive moderate correlation was observed between ventricular volume and the volume of the hippocampus, amygdala, and CPu -in accordance with all of these structures being reduced in size in Brd1 +/− mice. On the other hand, a considerable negative correlation was apparent between ventricle-, and CPu volume and corticospinal tract volume (Fig. 1d). In the cerebral cortex, structural alterations were observed in the occipital lobe, where a significant volume reduction was noted in Brd1 +/− mice compared to WT mice (Table 1, p = 0.02). Assessment of cortical thickness further revealed nominal significant thinning in this region along with thickening of the sensorimotor cortex (Fig. 1e).
Loss of CPu neurons in Brd1 +/− mice. In order to assess a potential cellular underpinning of the structural change in CPu, we performed a stereological estimate of total CPu neurons in fixed tissue from an independent group of mice (7 WT and 6 Brd1 +/− ). In line with reduced CPu volume in Brd1 +/− mice, a significant (~20%) reduction in striatal neurons was seen in Brd1 +/− compared WT mice (Fig. 1f, p < 0.01, Table 2). To assess the molecular implications of reduced Brd1 expression on striatal cells, we utilized previously published RNAseq data 15 from the dorsal striatum of adolescent Brd1 +/− and WT mice and performed cell specific expression analysis (CSEA). Non-surprisingly, differentially expressed genes (DEGs) were strongly enriched with genes specifically expressed in striatal neurons, namely medium-sized spiny neurons (MSNs) (Fig. 1g, Dopamine receptor 2 expressing (D2+) MSNs: q = 8.54E-12, and Dopamine receptor 1 expressing (D1+) MSNs: q = 3.61E-05).
Indicating that cellular loss is present already in adolescent Brd1 +/− mice, DEGs overlapping the MSN enriched gene lists were exclusively downregulated in Brd1 +/− mice, suggesting that this subset of DEGs may not reflect altered gene transcription, but rather a change in cell composition. Only neuronal mRNA markers were overrepresented among striatal DEGs whereas glial-, and oligodendrocyte mRNA markers where not. Notably, the decrease in markers for D2+ MSNs was moderately larger than that for D1+ MSNs (Fig. 1h, p < 0.01). Although not significantly enriched in the DEG set, markers for cholinergic, but not GABAergic interneurons, were additionally decreased in Brd1 +/− mice (Fig. 1h).
Reduced number of parvalbumin expressing interneurons in Brd1 +/− mice. Although no gross structural abnormalities were identified in the frontal cortex of Brd1 +/− mice in this study, we have previously reported cortical cellular pathologies, including a significant reduction of parvalbumin immunoreactive (PV+) interneurons in the anterior cingulate cortex (aCC) of Brd1 +/− mice 15 . Altered levels of interneuron mRNA markers and/or immunoreactivity in the adolescent brain may reflect improper neurodevelopment (e.g., impaired proliferation of neuronal progenitor cells, migration defects or GABA interneuron subtype alteration), neuronal loss by degeneration and/or transcriptional regulation. PV+ interneurons are preferentially ensheathed by perineuronal nets 16,17 consisting of specialized extracellular matrix components which bind the lectin, Vicia Villosa Agglutinin (VVA). Hence, VVA staining was used as a reliable marker for visualizing the specific neuronal subpopulation of cells potentially expressing PV (Fig. 2a, top). Confirming a reduction of PV+ immunoreactive aCC interneurons, we found a ~30% decrease in PV+ neurons in adolescent Brd1 +/− compared to WT mice (Fig. 2a, p < 0.01, Table 2). Similarly, we found a ~23% reduction in VVA+ cells (Fig. 2a, p < 0.01, Table 2) and a ~30% decrease in cells expressing both PV and VVA ( Fig. 2a. p < 0.01, Table 2), indicating that the reduction in PV+ immunoreactive neurons in adolescent Brd1 +/− mice is primarily, but not exclusively, due to an absence of neurons. Interestingly, by assessing the regional number of VVA+ neurons in the cortical layers of the aCC, it became apparent that the loss of PV+ interneurons was not equally distributed across aCC tissue, but predominantly localized to the centermost region corresponding to cortical layer V (Fig. 2b, 15 were overall significantly enriched with markers for layer Va cells (Fig. 2c, q = 1.759e-07), whereas no enrichment was seen in markers for layer Vb or layer VI neurons. mRNA markers for layer Va cells were furthermore predominantly decreased in Brd1 +/− compared to WT mice (Fig. 2c). To investigate if Brd1 +/− mice show early signs of interneuron loss, we additionally assessed a selection of cortical interneuron mRNA markers in newborn (P0)-, and juvenile (P21) mice. Surprisingly, in newborn mice, we found a ~40% reduction in Cck mRNA (Fig. 2d, p < 0.05), whereas Cck mRNA levels were normal in juvenile and adolescent Brd1 +/− mice. Contrarily, both Vip and Pvalb mRNA levels were normal in newborn mice and then progressively decreased towards adolescence (Fig. 2d).

Discussion
In vivo neuroimaging has revealed structural brain alterations in SZ cases 1,18-22 and abnormalities in neuron morphology, density and number has been reported in SZ post-mortem studies 23,24 . Developmental processes linked to genetic, prenatal and environmental factors 25 as well as post-onset progression 26 may contribute to these changes. However, medication exposure, patient heterogeneity and the temporal gap between in vivo imaging and neuropathological studies complicates the interpretation of these changes in relation to disorder initiation and progression 27 . Here we assess brain morphometry and neuronal mRNA markers in an animal model with construct and face validity to SZ in order to illuminate potential structural-and neuronal pathologies of the disorder. Brd1 +/− mice are characterized by behavioral and neurochemical alterations that implicates widespread brain impairments 15 , including monoaminergic dysregulation and psychotomimetic drug sensitivity. Using an MR imaging based approach, we find volumetric changes in brain tissues that are functionally linked to reported deficits in Brd1 +/− mice, including volumetric changes in striatum and other subcortical structures. Collectively, our results compare to the findings of the presently largest SZ MRI mega-analysis (2 028 individuals with SZ and 2 540 healthy controls) 1 , which particularly identify subcortical structural abnormalities in SZ (Fig. 3). Importantly, however, whereas SZ cases are characterized by volume reduction of the ventral striatum (accumbens), the striatal volume loss in Brd1 +/− mice, is seen in the dorsal part (caudate/putamen) (Figs 1b and 3) as has previously been reported in medication free cases 19 . Dorsal striatal activity has been shown to correlate with positive symptoms such as delusions and hallucinations 28 . It serves as the primary gateway of the basal ganglia and play key roles in motor control as well as learning, memory and decision-making 29 , especially in action selection and initiation through the convergence of corticostriatal and thalamostriatal afferents 30 Table 1. 3D MR imaging-based brain volumetric analysis. 3D MR imaging coupled with deformation-based morphometry analyses of differences in absolute cerebral volumes (whole brain, brain regions, brain structures and sub-structures) of 15 weeks old Brd1 +/− mice (n = 9) compared to WT mice (n = 10). The change in volume was calculated as the percentage change between the mean volume of the WT group and the mean volume of the Brd1 +/− group. Nominally significant groupwise differences are in bold. CPu volume correlate negatively with corticospinal tract volume and thickening of the somatosensory cortex in Brd1 +/− mice. Although there is a large degree of functional homology between the rodent and primate striatum, a direct comparison is complicated by differences in striatal organization 31 , cellular compositions 32 , and efferent connectivity 33 . Striatal volume loss may reflect either neuronal loss, shrinkage of neuron cell bodies and/ or non-neural changes, such as loss of glial components. The neuronal types of striatum includes MSNs, which constitute the majority of neurons (>90% in rodents) 34,35 , cholinergic interneurons (1-3%) 36 and various types of GABAergic interneurons 37 . We find an overall decrease in striatal neurons, with decreased mRNA markers for both MSNs and cholinergic interneurons. MSNs exist as two approximately equally-sized populations based on axonal projections and expression of dopamine receptor subtypes (D2+ or D1+) which have distinctive linkages to intracellular signaling cascades and targets, leading to fundamentally different cellular responses to extracellular dopamine. Namely, activation of dopamine receptor D1 facilitates ATP to cAMP conversion and thus increased cAMP mediated signaling, whereas activation of dopamine receptor D2 inhibits this process 38 . In line with previously reported over-activation of cAMP mediated signaling in the dorsal striatum of Brd1 +/− mice 15 , we find a larger decrease in D2+ than D1+ mRNA markers. D1+ MSNs project via the direct pathway, and D2+ MSNs via the indirect pathway to the substantia nigra pars reticulate 38 where they control the activity of afferents to the thalamus, and consequently produce opposing influence on motor output systems 39 . Activation of the direct pathway promotes motor activity whereas stimulation of the indirect pathway inhibits motor activity 40 . The imbalance in striatal D2+/D1+ MSNs suggested in the present study might thus explain the increased sensitivity towards the psychomotor stimulatory effects of cocaine and phencyclidine observed in Brd1 +/− mice 15 . Striatal cholinergic interneurons, however, differentially modulate MSN excitability dependent on dopamine tone 36 , thus complicating the interpretation. Although the largest structural change was detected in the dorsal striatum, volume reductions were similarly noted in neighboring globus pallidus, and amygdala where the cellular compositions are markedly different from that of striatum (i.e. amygdala is composed of 80% glutamatergic spiny principal neurons and GABA interneurons 41 ). Assessment of cellular pathologies in these brain tissues in Brd1 +/− mice is thus warranted. Prefrontal cortex (PFC) dysfunction has been implicated in SZ pathology, across symptom domains [42][43][44][45] , and alterations of the inhibitory PFC circuitry preferentially involving PV+ interneurons, are among the most robust findings 46 . In agreement with previously reported reduction in the number of PV immunoreactive interneurons in aCC in Brd1 +/− mice, we find a similar reduction using a different staining method and a different stereological approach. Moreover, we demonstrate the source of reduction primarily lies in reduced number of neurons and not in loss of immunoreactivity. Additionally, the reduction is most prominent in the centermost part of aCC, corresponding to layer V. This is supported by assessment of RNAseq data from the aCC region in Brd1 +/− mice, showing that DEGs cluster particularly in genes expressed in layer Va. Intriguingly, cortical projections to striatum originates exclusively in cortical layer V [47][48][49] . Finally, through assessment of interneuron mRNA markers in cortical tissue from newborn and juvenile Brd1 +/− mice along with RNAseq data from aCC in adolescent mice, we suggest that PV dysfunction might not be restricted to aCC, and that the dysfunction could be progressive and thus, likely neurodevelopmental in its origin. However, further investigations are warranted to delineate the neurodevelopmental impact of reduced Brd1 expression.
Despite recent advances in neuro-genetics, functional neuro-science, and profiling of the brain structural abnormalities in SZ, there has only been limited progress in understanding the pathophysiology and the development of effective therapies. Whereas studies in patients are largely limited to in vivo neuroimaging and postmortem examination, the much needed translation of disease pathology into mouse models has proven to be challenging, primarily due to the complex genetic architecture of SZ 50 . In the present study, we assess cellular and structural changes in a well characterized genetically modified mouse model that incorporates SZ genetic risk at three levels: allelic, chromatin interactomic, and brain transcriptomic 15 . We identify volumetric changes in brain structures that are largely overlapping with structural changes reported in the presently largest SZ MR imaging mega-analysis and functionally linked to SZ symptomatology. Intriguingly, we demonstrate that the striatal volume reduction is caused by a general loss of striatal neurons, likely differentially implicating D1+ and D2+ MSNs, thus presumably resulting in overall altered striatal output. Potentially coupled to this defect, we find indication of developmental loss of cortical PV+ neurons in cortical layer V that is rich with corticostriatal projection neurons.
Collectively, our study highlights the translational value of the Brd1 +/− mouse as a pre-clinical tool for SZ research and provides novel insight into its structural and cellular pathology.
(no correlation), blue (negative correlation), and red (positive correlation). (e) Illustration of cortical surface showing the localization of groupwise differences in cortical thickness. Only uncorrected p values are presented. Top: t-map (red color marks Brd1 +/− mice thicker and blue color marks Brd1 +/− thinner). Bottom: t-map with threshold cutoff at p < 0.05. (f) Stereological estimation of total number of CPu neurons show a significant reduction in 15 weeks old Brd1 +/− (n = 6) compared to WT (n = 7) mice (t test, p = 0.012). White arrows points to neurons, whereas black arrow points to glia. (g) Cell specific expression analysis (CSEA) of DEGs identified in dorsal striatum of 8 weeks old Brd1 +/− mice. Black bars show the Fisher's Exact Benjamini-Hochberg (BH) corrected p values (−log 10 (q value)) of overlap between DEGs and 27 broad and specific cell type gene sets (at CSEA specificity threshold set to 0.0001). Red bars show the relative contribution of downregulated DEGs to that significance. Enrichments were regarded as statistically significant at q < 0.05 as indicated by the red dotted line. (h) Log2 fold change in expression of mRNA markers of various striatal cell types. Red dotted line indicates groupwise mean for markers for D1+ and D2+ MSNs, respectively. The decrease in markers for D2+ MSNs was moderately larger than that for D1+ MSNs (t test, p = 0.008). *p < 0.05; **p < 0.01; ***p < 0.001.  Image analysis. MR images were corrected for B1 inhomogeneity using N3 51 and denoised using a non-local means filter 52 . Brain masks were manually constructed using ITK-SNAP (www.itksnap.org) 53 . Then images were linearly and non-linearly 54 registered to a neuroanatomical atlas of the C57BL/6 J mouse with 62 different, labelled regions 55 . Region of interest labels were subsequently transformed and resampled to scanner native space using the calculated deformation fields and affine transformations for calculation of individual regional volumes.

SCIENTIFIC
Cortical thickness analysis. Cortical thickness was calculated as previously described 56 . Briefly, Laplace's equation, with a fixed boundary condition that differed at the inner and outer surfaces, was solved using the inner and outer surfaces of the cortex defined on the anatomical atlas and transformed to the given mouse. For each point on the cortical surface, the length of a streamline connecting the inside and outside surfaces was used to define the thickness. Cortical thickness was compared between mice on a point-by-point basis using the spatial transformations determined previously by image registration to identify corresponding surface points for each mouse.
Tissue preparation. For histological analysis of caudate putamen (CPu) tissue, an independent batch of 15 weeks old male mice (7 WT and 6 Brd1 +/− ) were used and anterior cingulate cortex (aCC) immunohistochemistry was performed on 8 weeks old mice (8 WT and 6 Brd1 +/− ). Mice were deeply anesthetized with isoflurane. Transcardial brain perfusion fixation was initiated using Cardioplex solution followed by 4% formaldehyde via the left ventricle. Brains were dissected and immersed in 4% formaldehyde for 24 hours and then cryoprotected in sucrose solution (30% w/v) for 48 hours followed by snap-freezing. Brains were molded in Tissue-Tek (Sakura, Tokyo, Japan).
Histology. Brains  Stereological estimate of CPu volume. The 2D nucleator and the Cavalieri estimator were combined for unbiased volume estimation of left CPu in the stained sections using the following equation: V(CPu) = T·∑Area(CPu). T is the distance between sections and Area(CPu) is the summed areas of CPu estimated by the 2D nucleator.
Stereological estimate of CPu neurons. The total number of CPu neurons was estimated by the optical fractionator. The step lengths in the x-and y-direction were 400 µm, and the area of the two-dimensional unbiased counting frame was 2 000 µm 2 . Disector height and location was decided by the z-axis analysis and counts were performed between −5 to −15 µm in the z-direction. CPu neurons were identified among toluine blue outlined where ssf, asf and hsf are referred to as section sampling sample fraction, area sampling fraction and height sampling fraction, respectively.
Cell-type specific expression analysis and subtype mRNA marker analysis. Cell type-specific enrichment for nominally significant differentially expressed genes (DEGs) previously reported in RNAseq studies on tissue micropunches from dorsal striatum and aCC in 8 weeks old Brd1 +/− mice 15 was performed using the cell specific expression analysis (CSEA) tool (http://genetics.wustl.edu/jdlab/csea-tool-2/). This tool contains cell type-specifically expressed genes derived from a translational profiling approach that have identified murine transcriptomes from specific, marker-defined cellular subpopulations 57 . We assessed each set of DEGs (dorsal striatum and aCC) for enrichment of 27 broad and specific cell type gene sets (at CSEA specificity threshold set to 0.0001). Enrichments were regarded as statistically significant at Benjamini-Hochberg corrected p value (q value) < 0.05.

Immunohistochemistry.
Brains were sectioned (40 μm, coronally from Bregma 1.5 mm to 0.5 mm) as Quantitative reverse transcription-PCR. Whole cortices were sampled from newborn (P0) (7 WT and 7 Brd1 +/− ) and frontal cortex from juvenile (P21) (6 WT and 6 Brd1 +/− ) male mice by free-hand dissection. An overview of RNA extraction, quality control, cDNA synthesis, and qPCR protocols can be found in Table S1. Primers were designed to span exon boundaries to avoid amplification of potential contaminating genomic DNA (Table S2). Relative gene expression was calculated by the "standard curve method" 59 using serial dilutions of a pool of cDNA form all samples as a standard. The expression stability of reference genes (n = 4) within and between experimental groups was determined by analyzing their relative mRNA levels with the Normfinder software 60 . mRNA levels were normalized to the geometric mean of the mRNA levels of the two genes selected as the best combination for normalization.
Statistical analysis. MRI, stereological and mRNA data were analyzed using the GraphPad Prism software (San Diego, USA) and Microsoft Excel 2016 (Redmond, Washington, USA9. Data were checked for normal distribution by the Shapiro-Wilk normality test. The statistical significance of group wise comparisons were determined by the student's t-test. Correlation coefficients were calculated using the Correlation data analysis tool in the Microsoft Excel Real Statistics add-on package. Differences in cortical thickness were assessed at each point on the atlas cortical surface using two-sample t-tests. Only uncorrected p values are presented. The resulting maps were thresholded at p < 0.05. For assessment of neuronal loss across cortical layers, a two-way ANOVA with genotype and layer as factors. Tukey's test was performed as post hoc test and Bonferroni's corrected p value reported. Data were presented as mean ± SEM, and p values < 0.05 were considered to be statistically significant. Details on the statistic applied in CSEA can be found in (2), where the method is described in detail. Briefly, the overlap of genes identified as differentially expressed between Brd1 +/− and WT mice and genes highly enriched in certain cell types was compared and a p value for Fisher's Exact Test is provided. Only cell type specific genes determined with the highest stringencies for enrichment (PSi = 0.0001) were used in this study and only Benjamini Hochberg corrected p-value (q value) < 0.05 were considered significant.