TMEM106B haplotypes have distinct gene expression patterns in aged brain

Given the world’s aging population, neurodegenerative diseases have become a major cause of disability and death. While the disease mechanisms differ, neurodegeneration often results from the accumulation of misfolded aggregated proteins in different areas of the aging brain, and this process yields cell death and inflammatory damage in those brain regions [1]. Recent studies have revealed that two common haplotypes in transmembrane protein 106B (TMEM106B) are associated with risk of multiple neurodegenerative diseases, most notably with frontotemporal lobar degeneration with pathological inclusions of TDP-43 (FTLD-TDP) [2‐4], progranulin (GRN)-related FTLD [3, 5], chromosome 9 open reading frame 72 (C9ORF72)-mediated FTLD [6, 7] and hippocampal sclerosis of aging [8, 9]. TMEM106B haplotypes were also shown to associate with the development of cognitive impairment in amyotrophic lateral sclerosis (ALS) [10] and with the presence of TDP-43 pathology in Alzheimer’s disease (AD) [11] and elderly individuals without FTLD [12]. The broad involvement of TMEM106B in neurodegenerative diseases makes it an important gene to characterize and a promising target for potential therapies.

As with most risk loci identified by genome-wide association studies (GWAS), the functional variant(s) in the TMEM106B locus associated with the reported associations remains elusive. Within the associated linkage disequilibrium (LD) block, rs3173615 is the only variant encoding an amino acid change from the more common, highly conserved, threonine (Thr185; risk allele) to a serine (Ser185; protective allele) at position 185. In vitro, the protective (Ser185) TMEM106B isoform was consistently expressed at lower levels than the risk (Thr185) TMEM106B isoform due to an increased rate of protein degradation, possibly resulting from changes in TMEM106B glycosylation [13]. In addition to the coding variant (rs3173615) numerous non-coding variants also differentiate the TMEM106B haplotypes and at least one of these variants (rs1990620) was suggested to affected higher-order chromatin architecture at the TMEM106B locus and changes in mRNA expression [14]. Indeed, expression studies in FTLD-TDP brains have shown increased TMEM106B mRNA levels in carriers of the risk haplotype [2]. Regardless of the identity of the specific functional variant(s), these findings suggest the presence of higher levels of TMEM106B in carriers of the risk haplotype and lower levels of TMEM106B in carriers of the protective haplotype. These findings are in line with cell biological studies which showed that increased TMEM106B levels were cytotoxic and led to an increase in lysosomal size and reduced lysosomal acidification [4]. Importantly, however, the specific mechanism by which TMEM106B haplotypes and changes in its expression contribute to neurodegeneration remains unknown.

In this study, we utilized RNA sequencing data of temporal cortex (TCX) and cerebellum (CER) samples from 312 North American Caucasian subjects to identify transcriptome signatures associated with the TMEM106B haplotypes. By comparing homozygote rs3173615 TT (risk) and SS (protective) carriers, we discovered differentially expressed genes in both TCX and CER regions, and identified shared gene co-expression networks between TCX and CER through which the TMEM106B haplotypes may contribute to brain function and brain health.

In this study, we used RNA sequencing data of two brain regions to investigate the involvement of the TMEM106B risk and protective haplotypes in brain health. We identified large transcriptional differences between the protective SS and risk TT haplotypes in TCX, and much smaller differences in CER, consistent with the higher level of tissue damage and neuronal loss in the TCX region. Importantly, even though not statistically significant in CER, our data suggested similarities in the effects of TMEM106B haplotypes on the transcriptional signatures in both brain regions. By comparing the top 500 differentially expressed genes in TCX and CER with |fold change| ≥ 1.2 between SS and TT, we found 28 overlapping genes, all with a fold-change in the same direction (25 are increased and 3 are decreased in SS carriers as compared to TT carriers). Some of the biggest increases in SS carriers were seen in GAD1 and GAD2, glutamic acid decarboxylases which are responsible for catalyzing the production of gamma-aminobutyric acid (GABA) from glutamate, and SLC32A1 (also known as VGAT) which is a transporter involved in the uptake of GABA and glycine into synaptic vesicles. The specific changes in GABA-related signaling are interesting in light of a recent study which reported the preferential elimination of inhibitory synapses (defined by VGAT+ immunoreactivity) in the ventral thalamus of Grn knock-out mouse brains [20]. This observation raises the possibility that more robust GABAergic signaling or an increase in the number of inhibitory synapses at baseline in SS haplotype carriers may contribute to the profound protection conferred by this haplotype in patients with GRN mutations. Future immunohistochemical studies in human brain samples from SS and TT TMEM106B carriers are needed to confirm this hypothesis.

Using gene co-expression network analyses we further identified significant correlations of TMEM106B haplotypes with gene expression modules enriched for genes involved in similar biological processes across both brain regions. In TCX and CER, the TMEM106B SS haplotype was correlated with gene networks involved in synaptic transmission, whereas the TT haplotype was correlated with immune response networks in both brain regions, in addition to other specific gene networks such as cell-cell adhesion found only in TCX. In a further reference to the Grn knock-out mouse model, it is of interest to note that the gene expression modules most significantly correlated with the loss of Grn in cerebral cortex, cerebellum and hippocampus in mice were annotated by GO as involved in the innate immune response with C1qa, C1qb, C1qc, and C3 among the most connected genes in the module, similar to what we observed when we compared SS to TT carriers in the CER (purple module; Fig. 3d) [20]. Finally, using 5 genes as surrogates for the 5 major brain cell types, we found that the gene networks were associated with the cell type composition in the TT and SS brains: the SS carriers showed higher neuronal gene expression as compared to the TT carriers, corresponding to an increase in networks related to synaptic transmission in SS brains; the TT carriers showed higher microglial gene expression as compared to SS carriers, in agreement with the observed enrichment for immune response networks in TT brains.

TMEM106B risk variants had previously been reported to be associated with both brain volume and connectivity. Specifically, using imaging studies, the TMEM106B risk allele was shown to significantly associate with reduced brain volume in non-demented elderly individuals, particularly in the superior temporal gyrus in the left hemisphere [21]. In addition, GRN mutation carriers with two copies of the TMEM106B risk allele demonstrated worsened brain connectivity compared to those who carried one or no risk alleles [22]. These authors found that TMEM106B haplotypes did not influence grey matter volume directly on its own, but in mutation carriers the protective TMEM106B haplotype was able to enhance the benefit of cognitive reserve on brain structure. Similarly, in our study, we found no significant differences in total brain weight between our TT and SS carriers. Instead our data suggest that the SS protective haplotype may confer higher synaptic transmission by strengthening brain connectivity in aged or diseased brain.

TMEM106B has also recently been linked to healthy neurological aging. In frontal cortex, the TMEM106B risk haplotype was associated with gene expression patterns suggestive of an older age than the individual’s true chronological age [23]. Similar to our study, the TMEM106B risk haplotype was further found to be associated with increased inflammation and reduced neuronal expression in these neuropathologically normal elderly individual. Contrary to our findings, however, they did not observe an effect of TMEM106B on gene expression in cerebellar tissues. This may be related to the fact that they focused on neurologically normal individuals whereas we included a mixture of neurodegenerative diseases and normal controls. To determine whether the significant modules we identified were mainly contributed by disease samples or controls, we performed additional WGCNA analyses with only cases in TCX (n = 74, 37 TT and 37 SS) or CER (n = 84, 42 TT and 42 SS), and using only controls in TCX (n = 42: 24 TT and 18 SS) or CER (n = 41: 25 TT and 16 SS). In both TCX and CER, analysis of the cases identified similar modules as we did in the full TCX and CER datasets (Additional file 1: Table S6). Furthermore, independent analyses of each disease group: TCX-AD, TCX-PSP, CER-AD and CER-PSP all identified synaptic transmission modules that are significantly correlated with the TMEM106B haplotype (Additional file 1: Table S7), suggesting that the effects of the TMEM106B haplotype was not specific to disease type. Analysis of only the control samples in TCX and CER identified less modules, and synaptic transmission and immune response modules were not significantly correlated with the TMEM106B haplotype in either brain region (Additional file 1: Table S8). While the smaller sample size of the control cohorts may have reduced our power to detect significant difference, these results suggest that the effects of TMEM106B haplotypes on the transcriptome are more pronounced in disease tissues than in healthy tissues.

In summary, our study demonstrates significant and partly conserved effects of the TMEM106B haplotypes on the transcriptome across multiple brain regions. It is quite remarkable that a variant which is common in the normal population (minor allele frequency = 0.40) can have such profound effects on gene expression and cell type composition, especially since our SS and TT study groups were matched for disease phenotypes and age at death. This suggests that additional genetic and/or environmental factors are likely to modulate the effect of the TMEM106B haplotypes on neurological and normal brain aging in individual human subjects. TMEM106B has only begun to be characterized since the discovery of its association to FTLD-TDP several years ago. While its function has mostly been linked to lysosome functions and trafficking, and recently to myelination [24], our results suggest that TMEM106B plays a broader role in the CNS response to pathological or age-related insults in multiple brain regions. Continued research into TMEM106Bs normal function and dysfunction in the context of neurodegenerative diseases will be critical and holds promise for future therapeutic strategies.