Glial dysregulation in human brain in Fragile X-related disorders

While large trinucleotide repeat expansions at the FMR1 locus cause Fragile X Syndrome (FXS), smaller “premutations” are associated with the late-onset condition Fragile X-associated tremor/ataxia syndrome (FXTAS), which shows very different clinical and pathological features, with no clear molecular explanation for these marked differences. One prevailing theory posits that the premutation uniquely causes neurotoxic increases in FMR1 mRNA (i.e., 4-8-fold increases), but evidence to support this hypothesis is largely derived from analysis of peripheral blood. We applied single- nucleus RNA-sequencing to post-mortem frontal cortex and cerebellum from 9 individuals with Fragile X mutations as well as age and sex matched controls (n=6) to assess cell-type specific molecular neuropathology. We found robust reduction of FMR1 mRNA in FXS as expected, with modest but significant upregulation (∼1.3 fold) of FMR1 in glial clusters associated with premutation expansions. In premutation cases we identified alterations in glia number in cortex and cerebellum. Differential expression analysis demonstrated altered cortical oligodendrocyte development, while gene ontology analysis revealed alterations in neuroregulatory roles of glia, such as glial modulation of neurotransmission and synaptic structure. We identified significant enrichment of known FMR1 protein target genes in differentially expressed gene lists in FXS as well as the premutation, suggesting FMR1 protein target pathways may represent a shared source of dysfunction in both conditions despite opposite FMR1 mRNA changes. These findings challenge existing dogma regarding FXTAS and implicate glial dysregulation as a critical facet of premutation pathophysiology, representing novel therapeutic targets directly derived from the human condition.

FXS as well as the premutation, suggesting FMR1 protein target pathways may 48 represent a shared source of dysfunction in both conditions despite opposite FMR1 49 mRNA changes. These findings challenge existing dogma regarding FXTAS and 50 implicate glial dysregulation as a critical facet of premutation pathophysiology, 51 representing novel therapeutic targets directly derived from the human condition. 52 5 Despite these gaps in knowledge, there has been no cell-type specific analysis of 99 transcriptional changes related to Fragile X in human brain to date. the PM, we also included two known cases of FXS, to assess whether well-known 123 effects on FMR1 expression were present in our dataset. One case of FXS due to a 124 deletion of FMR1 was included given the known shared molecular consequences of 125 FMR1 deletion and trinucleotide expansion (Gedeon et al., 1992). Neither FXS case 126 had neuropathological abnormalities noted, consistent with expectations. The majority 127 of PM cases had either clinical and/or neuropathological evidence of FXTAS. We 128 identified one case that in the past was mistakenly categorized as FXS, but whose 129 clinical records and genetic testing revealed it to be a PM (see Table 1). We validated 130 functional effects of Fragile X disruption with western blotting of FMRP directly on frontal 131 cortex tissue ( Figure 1A) and obtained expected results: absent FMRP in FXS in both 132 the FM and gene deletion, and variably reduced FMRP in PM cases. 133 Following unsupervised clustering of single nuclei, and filtering, we obtained over 134 120,000 high quality nuclei for further analysis across samples, including nuclei from 6 135 age-and sex-matched controls ( Figure 1B, Table 2). We applied known cell type-136 specific markers to assess the specificity and accuracy of unsupervised clustering (Fig  137   2). For both prefrontal cortex and cerebellar hemisphere we identified specific transcriptional profile of OLI and OLII to oligodendrocyte lineage clusters identified in 150 mouse (Marques et al., 2016), and found that OLI gene expression resembled mouse 151 committed oligodendrocyte progenitors (COPs) and OLII resembled immature, newly 152 formed, non-myelinating oligodendrocytes. On the other hand, in the cerebellum, 153 although granule cells accounted for the majority of nuclei captured, as expected, we 154 also identified a cerebellar specific Bergmann glia cluster. OLI and OLII clusters were 155 not identified in cerebellar samples. 156 Comparison of average FMR1 mRNA across all nuclei in frontal cortex and 157 cerebellum revealed regional differences in relative expression between neurons and 158 glia. In the frontal cortex, expression of FMR1 was higher in excitatory and inhibitory 159 neurons compared to glia, as expected ( Figure 2 figure supplement 2, Figure 3). 160 However, in the cerebellum, FMR1 expression was expressed in more non-neuronal 161 subtypes at baseline. We used an independent harmonized single cell transcriptomic 162 resource to confirm these findings and identified similar region-specific patterns of 163 expression with higher relative glial to neuron FMR1 mRNA expression in cerebellum. 164  Violin plots demonstrate reduced FMR1 mRNA expression in multiple clusters in FXS, while FMR1 mRNA is variably increased in PM clusters, primarily in glia. Cluster abbreviations as in Fig. 2. Orange *: reduced FMR1 p-value in FXS vs CON padj< .05, teal *: increased FMR1 in PM vs. CON padj < .05. Right panel shows permutation plot for PM cluster proportions demonstrating glial cell number alterations in cortex and cerebellum. Red: FDR < .05 and abs(log2FC) > 0.58. # indicates significance of this cluster is not robust to outlier sample removal.
( Figure 3). Although the absence of significant FMR1 upregulation in a small number of 179 clusters in PM cases may be due to inadequate power, most clusters included more 180 than enough nuclei (including frontal cortex excitatory and inhibitory neurons, as well as 181 cerebellar granule cells and interneurons) to rule this explanation out (see Methods). 182 Thus, in general, the lack of significant upregulation of FMR1 mRNA in neuronal 183 subclusters in the PM cases is not due to a lack of power. Rather, it suggests that 184 overall, the increase in FMR1 expression in brain caused by the PM is far more modest 185 than the 4-8 fold increase observed in blood, and shows a preferential impact on glia, in 186 the regions assessed here. 187 In the cerebellum, we identified changes in nuclei number in PM cases that    Thus, we identified unexpected novel alterations in glial number in frontal cortex in PM 212

cases. 213
Although limited by small sample size, we also assessed cluster proportions in FXS 214 and found that alterations in cellular composition was markedly distinct from PM cases.

Discussion: 282
We present the first cell type specific analysis of gene expression of Fragile X 283 related disorders in human brain. We identified changes in FMR1 mRNA expression, 284 cellular proportion, and cell-type specific gene expression that sheds light on molecular 285 perturbations associated with FMR1 and specifically highlights an important role for glial 286 molecular dysregulation in PM pathology. 287 Our data suggest that FMR1 mRNA expression in PM cases, at least in the brain 288 regions analyzed here, is more modestly affected than has been observed in peripheral 289 blood cells and furthermore that it preferentially effects glial cells more than neurons. preferentially lost with time, our cellular proportion analysis (see below) does not 300 support this interpretation, as one would expect clusters that are disproportionately lost 301 to have relatively higher increases in FMR1 mRNA. We also include in our analyses one 302 21 year-old PM case, whose data is very similar to the other aged PM cases, which 303 further argues against age-related loss. Finally, changes in FMR1 expression were 304 comparable between clusters known to be vulnerable to PM associated intranuclear 305 inclusions (neurons, astrocytes) and those known to be spared (oligodendrocytes), 306 arguing against inclusion presence as being a confounding factor in FMR1 mRNA 307 measurement. Although our work challenges the causal role of extremely elevated 308 FMR1 mRNA in human brain, it is possible that more modest increases of CGG 309 containing FMR1 RNA still lead to cellular dysfunction, through previously posited 310 mechanisms including trinucleotide repeat toxicity. Thus, our work provides an 311 important foundation to understanding FMR1 mRNA levels that are relevant to 312 neurological pathophysiology in animal and human model systems and broadens the 313 scope of cellular subtypes that warrant further investigation within this context. 314

Cellular proportion 315
Alterations in cell number in PM cases in both cerebellum and cortex also 316 implicated glial dysregulation. Our finding of reduced Purkinje cells and Bergmann cell 317 increases in the cerebellum in PM cases parallels well-described neuropathological 318 findings (Greco et al., 2006), and reinforces the validity of our approach, and suggests 319 that loss of Purkinje cells may contribute to FXTAS signs and symptoms. In PM cases, 320 we also identified a proportional decrease in endothelial cells and astrocytes, with In conclusion, we provide compelling evidence from human brain regarding cell type-378 specific molecular neuropathology that helps contextualize the clinical heterogeneity 379 associated with genetic variation at the FMR1 locus in neurodevelopment and 380 neurodegeneration and specifically implicates glial dysregulation in PM pathology. 381

Materials and Methods 382
Samples 383

Post-mortem tissue was obtained from either the University of Maryland 384
Neurobiobank and for one control case, from Autism Brain Net. All tissue was from 385 deceased individuals and as such is not considered human subjects research. Fragile X 386 mutation status/repeat size was determined through direct review of de-identified clinical 387 records, and cross-referenced with prior published validation of the same cases (Table  388 1). Most of the PM cases had clinical symptomatology or neuropathological evidence of 389 FXTAS (Table 1). Samples were matched for age, sex, and PMI but no cut-offs were 390 utilized to exclude any cases (see Appendix Figure 1 for further details.) 391

Western Blotting 392
~25-50 mg frozen frontal cortex was homogenized in RIPA buffer + protease 393 inhibitors and centrifuged, total protein content was then quantified. Laemelli sample 394 buffer was added to the protein supernantant and boiled for 5 minutes. Equal amounts 395 of protein (10 ug) were loaded onto precast SDS-Page gels with molecular weight 396 ladders. Samples were transferred to membranes, blocked with Licor block (Lincoln, 397 NE), and incubated in primary antibody overnight diluted in block at four degrees.

Appendix (Supplemental Methods) 490
We found no association between: PMI & RIN, age and PMI, and age and RIN, 491 as expected (Appendix Figure 1). There was no difference in average RIN, PMI, or age 492 between PM and controls. We did observe a reduction in RIN in the FXS samples as 493 compared to controls but not PM cases. 494 Nuclear staining and sorting was conducted to select against dying cells, debris, 495 and doublets (Appendix Figure 2).