Immune-related genetic enrichment in frontotemporal dementia

Background Converging evidence suggests that immune-mediated dysfunction plays an important role in the pathogenesis of frontotemporal dementia (FTD). Although genetic studies have shown that immune-associated loci are associated with increased FTD risk, a systematic investigation of genetic overlap between immune-mediated diseases and the spectrum of FTD-related disorders has not been performed. Methods and findings Using large genome-wide association studies (GWAS) (total n = 192,886 cases and controls) and recently developed tools to quantify genetic overlap/pleiotropy, we systematically identified single nucleotide polymorphisms (SNPs) jointly associated with ‘FTD-related disorders’ namely FTD, corticobasal degeneration (CBD), progressive supranuclear palsy (PSP), and amyotrophic lateral sclerosis (ALS) – and one or more immune-mediated diseases including Crohn’s disease (CD), ulcerative colitis (UC), rheumatoid arthritis (RA), type 1 diabetes (T1D), celiac disease (CeD), and psoriasis (PSOR). We found up to 270-fold genetic enrichment between FTD and RA and comparable enrichment between FTD and UC, T1D, and CeD. In contrast, we found only modest genetic enrichment between any of the immune-mediated diseases and CBD, PSP or ALS. At a conjunction false discovery rate (FDR) < 0.05, we identified numerous FTD-immune pleiotropic SNPs within the human leukocyte antigen (HLA) region on chromosome 6. By leveraging the immune diseases, we also found novel FTD susceptibility loci within LRRK2 (Leucine Rich Repeat Kinase 2), TBKBP1 (TANK-binding kinase 1 Binding Protein 1), and PGBD5 (PiggyBac Transposable Element Derived 5). Functionally, we found that expression of FTD-immune pleiotropic genes (particularly within the HLA region) is altered in postmortem brain tissue from patients with frontotemporal dementia and is enriched in microglia compared to other central nervous system (CNS) cell types. Conclusions We show considerable immune-mediated genetic enrichment specifically in FTD, particularly within the HLA region. Our genetic results suggest that for a subset of patients, immune dysfunction may contribute to risk for FTD. These findings have potential implications for clinical trials targeting immune dysfunction in patients with FTD.


ABSTRACT Background
Converging evidence suggests that immune-mediated dysfunction plays an important role in the pathogenesis of frontotemporal dementia (FTD). Although genetic studies have shown that immune-associated loci are associated with increased FTD risk, a systematic investigation of genetic overlap between immune-mediated diseases and the spectrum of FTD-related disorders has not been performed.

Methods and findings
Using large genome-wide association studies (GWAS) (total n = 192,886 cases and controls) and recently developed tools to quantify genetic overlap/pleiotropy, we systematically identified single nucleotide polymorphisms (SNPs) jointly associated with 'FTD-related disorders' namely FTD, corticobasal degeneration (CBD), progressive supranuclear palsy (PSP), and amyotrophic lateral sclerosis (ALS) -and one or more immune-mediated diseases including Crohn's disease (CD), ulcerative colitis (UC), rheumatoid arthritis (RA), type 1 diabetes (T1D), celiac disease (CeD), and psoriasis (PSOR). We found up to 270-fold genetic enrichment between FTD and RA and comparable enrichment between FTD and UC, T1D, and CeD. In contrast, we found only modest genetic enrichment between any of the immune-mediated diseases and CBD, PSP or ALS. At a conjunction false discovery rate (FDR) < 0.05, we identified numerous FTD-immune pleiotropic SNPs within the human leukocyte antigen (HLA) region on chromosome 6. By leveraging the immune diseases, we also found novel FTD susceptibility loci within LRRK2 (Leucine Rich Repeat Kinase 2), TBKBP1 (TANK-binding kinase 1 Binding Protein 1), and PGBD5 (PiggyBac Transposable Element Derived 5). Functionally, we found that expression of FTD-immune pleiotropic genes (particularly within the HLA region) is altered in postmortem brain tissue from patients with frontotemporal dementia and is enriched in microglia compared to other central nervous system (CNS) cell types.

Conclusions
We show considerable immune-mediated genetic enrichment specifically in FTD, particularly within the HLA region. Our genetic results suggest that for a subset of patients, immune dysfunction may contribute to risk for FTD. These findings have potential implications for clinical trials targeting immune dysfunction in patients with FTD.

INTRODUCTION
Frontotemporal dementia (FTD) is a fatal neurodegenerative disorder and the leading cause of dementia among individuals younger than 65 years of age [1]. Given rapid disease progression and absence of disease modifying therapies, there is an urgent need to better understand FTD pathobiology to accelerate development of novel preventive and therapeutic strategies.
Converging molecular, cellular, genetic, and clinical evidence suggests that neuroinflammation plays a role in FTD pathogenesis. Complement factors and activated microglia, key components of inflammation, have been established as histopathologic features in brains of patients [2] and in mouse models of FTD [3,4]. Genome-wide association studies (GWAS) have shown that single nucleotide polymorphisms (SNPs) within the immune-regulating human leukocyte antigen (HLA) region on chromosome 6 are associated with elevated FTD risk [5]. Importantly, there is increased prevalence of immune-mediated diseases among patients with FTD [6,7]. Together, these findings suggest that immune-related mechanisms may contribute to and potentially drive FTD pathology.
We obtained FTD GWAS summary statistic data from phase I of the International FTD-Genomics Consortium (IFGC), which consisted of 2,154 clinical FTD cases and 4,308 controls with genotyped and imputed data at 6,026,384 SNPs (Table 1, for additional details, see [5]). The FTD dataset included multiple clinically diagnosed FTD subtypes: behavioral variant (bvFTD), semantic dementia (sdFTD), primary nonfluent progressive aphasia (pnfaFTD), and FTD overlapping with motor neuron disease (mndFTD). These FTD cases and controls were recruited from forty-four international research groups and diagnosed according to the Neary criteria [17]. We obtained CBD GWAS summary statistic data from 152 CBD cases and 3,311 controls at 533,898 SNPs (Table 1, for additional details see [18]). The CBD cases were collected from eight institutions and controls were recruited from the Children's Hospital of Philadelphia Health Care Network. CBD was neuropathologically diagnosed using the National Institute of Health Office of Rare Diseases Research criteria [19]. We obtained PSP GWAS summary statistic data (stage 1) from the NIA Genetics of Alzheimer's Disease Storage Site (NIAGADS), which consisted of 1,114 individuals with autopsy-confirmed PSP and 3,247 controls at 531,451 SNPs (Table 1, for additional details see [20]). We obtained publicly-available ALS GWAS summary statistic data from 12,577 ALS cases and 23,475 controls at 18,741,501 SNPs (Table 1, for additional details see [21]). The relevant institutional review boards or ethics committees approved the research protocol of the individual GWAS used in the current analysis, and all human participants gave written informed consent.

Genetic Enrichment Statistical Analyses
To identify specific loci jointly involved with each of the four FTD-related disorders and the six immune-mediated diseases, we computed conjunction FDRs [22,23]. Conjunction FDR, denoted by FDR trait1& trait2 is defined as the posterior probability that a SNP is null for either trait or for both simultaneously, given the pvalues for both traits are as small, or smaller, than the observed p-values. Unlike the conditional FDR which ranks disease/primary phenotype associated SNPs based on genetic 'relatedness' with secondary phenotypes [24], the conjunction FDR minimizes the possibility/likelihood of a single phenotype driving the common association signal.
Conjunction FDR therefore tends to be more conservative and specifically pinpoints pleiotropic loci between the traits/diseases of interest. We used an overall FDR threshold of < 0.05, which means 5 expected false discoveries per hundred reported. We constructed Manhattan plots based on the ranking of conjunction FDR to illustrate the genomic location of the pleiotropic loci. Detailed information on conjunction Q-Q plots, Manhattan plots, and conjunction FDR can be found in prior reports [22,23,25,26].

Functional evaluation of shared risk loci
To assess whether SNPs that are shared between FTD and immune-mediated disease modify gene expression, we identified cis-expression quantitative loci (eQTLs, defined as variants within 1 Mb of a gene's transcription start site) associated with shared FTD-immune SNPs and measured their regional brain expression in a publicly available dataset of normal control brains (UKBEC, http://braineac.org/) [27]. We also evaluated eQTLs using a blood-based dataset [28]. We applied an analysis of covariance (ANCOVA) to test for associations between genotypes and gene expression. We tested SNPs using an additive model.

Network based functional association analyses
To evaluate potential protein and genetic interactions, co-expression, colocalization, and protein domain similarity for the functionally expressed (i.e. with significant cis-eQTLs) overlapping genes, we used GeneMANIA (www.genemania.org), an online web-portal for bioinformatic assessment of gene networks [29]. In addition to visualizing the composite gene network, we also assessed the weights of individual components within the network [30].

Gene expression alterations in FTD brains
To determine whether functionally expressed (i.e. with significant cis-eQTLs) pleiotropic genes are differentially expressed in FTD brains, we analyzed gene expression of overlapping genes in a publically available dataset [31]. Specifically, we analyzed gene expression data from the frontal cortex, hippocampus, and cerebellum of controls and patients with frontotemporal dementia (FTD-U with or without progranulin (GRN) mutations, total n =28) (Gene Expression Omnibus (GEO) dataset GSE13162) [31].

Evaluation of cell classes within the brain
Using a publicly available RNA-sequencing transcriptome and splicing database [32], we ascertained whether the functionally expressed (i.e. with significant cis-eQTLs) pleiotropic genes are expressed by specific cell classes within the brain. The eight cell types surveyed are neurons, astrocytes, oligodendrocyte precursor cells, newly formed oligodendrocytes, myelinating oligodendrocytes, microglia, endothelial cells, and pericytes (for additional details, see [32]).

Shared genetic risk between FTD and immune-mediated disease
Using progressively stringent p-value thresholds for FTD SNPs (i.e. increasing values of nominal -log 10 (p)), we observed considerable genetic enrichment for FTD as a function of several immune-mediated diseases ( Figure 1A). More specifically, we found strong (up to 270-fold) genetic enrichment between FTD and RA, and comparable enrichment between FTD and UC, T1D, and CeD, with weaker enrichment for PSOR and CD.

Modest genetic enrichment between immune-mediated disease and PSP, CBD and ALS
To evaluate the specificity of the shared genetic overlap between FTD and immune-mediated disease, we also evaluated overlap between the 6 immune-mediated diseases and CBD, PSP, and ALS. For CBD and PSP a few of the immune-mediated diseases produced genetic enrichment comparable to that seen for FTD (Supplementary  Supplementary Figures 2A-C, Tables 1-3).
Beyond the HLA region, we found several overlapping loci between the immunemediated diseases and CBD, PSP and ALS ( Supplementary Figures 2A-C, Tables 1-3).

cis-eQTL expression
To investigate whether shared FTD-immune SNPs modify gene expression, we evaluated cis-eQTLs in both brain and blood tissue types. At a previously established conservative Bonferroni corrected p-value < 3.9 x 10 -5 [33], we found significant cisassociations between shared SNPs and genes in the HLA region on chromosome 6 in peripheral blood mononuclear cells (PBMC), lymphoblasts, and the human brain (see Supplementary Table 4 for gene expression associated with each SNP). We also found that rs199533 and rs17572851 on chr 17 were significantly associated with MAPT (p = 2 x 10 -12 ) expression in the brain. Beyond the HLA and MAPT regions, notable cis-eQTLs included rs10784359 and LRRK2 (p = 1.40 x 10 -7 ) and rs2192493 and TBKBP1 (p = 1.29 x 10 -6 ) (see Supplementary Table 4).

Protein-protein and co-expression networks
We found physical interaction and gene co-expression networks for the FTDimmune pleiotropic genes with significant cis-eQTLs (at a Bonferroni corrected p-value < 3.9 x 10 -5 ). We found robust co-expression between various HLA classes further suggesting that large portions of the HLA region, rather than a few individual loci, may be involved with FTD (Fig 3, Supplementary Table 5). Interestingly, we found that several non-HLA functionally expressed FTD-immune genes, namely LRRK2, PGBD5, and TBKPB1, showed co-expression with HLA associated genes ( Figure 3).

Genetic expression in FTD brains compared to controls
To investigate whether the FTD-immune pleiotropic genes with significant cis-eQTLs are differentially expressed in FTD brains, we compared gene expression in FTD-U brains to brains from neurologically healthy controls. We found significantly different levels of HLA gene expression in FTD-U brains compared to controls (Table 3). This was true of FTD-U brains regardless of progranulin gene (GRN) mutation status. In spite of the fact that the FTD GWAS used to identify these genes was based on patients with sporadic FTD (without GRN mutations), GRN mutation carriers showed the greatest differences in HLA gene expression ( Figure 4, Table 3). These findings are compatible with prior work showing microglial-mediated immune dysfunction in GRN carriers [3,35].

Microglial enrichment
For the FTD-immune pleiotropic genes with significant cis-eQTLs, across different CNS cell types, we found significantly greater expression within microglia compared to neurons or fetal astrocytes ( Figure 5A). Interestingly, HLA genes showed the greatest degree of differential expression. Four of the FTD-immune HLA associated genes, namely HLA-DRA, AOAH, HLA-A, and HLA-C, showed highest expression in microglia (ranging from 10 to 60 FPKM, see Figure 5B). In addition, MAPT was predominantly expressed in neurons, LRRK2 in microglia/macrophages, PGBD5 in neurons, and TBKBP1 in fetal astrocytes ( Figure 5B and Supplemental Figures 3A-C).

DISCUSSION
We systematically assessed genetic overlap between 4 FTD-related disorders and several immune-mediated diseases. We found extensive genetic overlap between FTD and immune-mediated disease particularly within the HLA region on chromosome 6, a region rich in genes associated with microglial function. This genetic enrichment was specific to FTD and did not extend to CBD, PSP, or ALS. Further, we found that shared FTD-immune gene variants were differentially expressed in FTD patients compared with controls, and in microglia compared with other CNS cells. Beyond the HLA region, by leveraging the immune-mediated traits, we detected novel FTD susceptibility loci within LRRK2, TBKBP1 and PGBD5. Considered together, these findings suggest that various microglia and inflammation-associated genes, particularly within the HLA region, play a critical and selective role in FTD pathogenesis. By combining GWAS from multiple studies and applying a pleiotropic approach, we identified genetic variants jointly associated with FTD-related disorders and immunemediated disease. We found that the strength of genetic overlap with immune-mediated disease varies markedly across FTD-related disorders, with the strongest pleiotropic effects associated with FTD, followed by CBD and PSP, and the weakest pleiotropic effects associated with ALS. We identified eleven FTD-immune associated loci that mapped to the HLA region, a concentration of loci that was not observed for the other disorders. Indeed, only two PSP-immune pleiotropic SNPs and no CBD-or ALS-immune pleiotropic SNPs mapped to the HLA region. Previous work has identified particular HLA genes associated with CBD, PSP, and ALS [34,36]. In contrast, our current findings implicate large portions of the HLA region in the pathogenesis of FTD. Together these results suggest that each disorder across the larger FTD spectrum has a unique relationship with the HLA region.
Our results also indicate that functionally expressed FTD-immune shared genetic variants are differentially expressed in FTD brains compared to controls and in microglia compared to other CNS cell types ( Figure 5). Microglia play a role in the pathophysiology of GRN+ FTD. Progranulin is expressed in microglia [35] and GRN haploinsufficiency is associated with abnormal microglia activation and neurodegeneration [3]. It is perhaps expected, therefore, that GRN+ brains show differential expression of FTD-immune genes, even though these genetic variants were derived from GWAS of patients with sporadic FTD (who lack GRN or other established FTD mutations). More surprising is the presence of similar enrichment in GRN-brains, suggesting that dysfunction of microglial-centered immune networks may be a common feature of FTD pathogenesis.
By leveraging statistical power from the large immune-mediated GWASs, we identified novel FTD susceptibility loci within LRRK2, TBKBP1 and PGBD5 and confirmed previously shown FTD associated signal within the MAPT region. LRRK2 mutations are a cause of Parkinson's disease [37] and Crohn's disease [38]. We previously described a potential link between FTD and the LRRK2 locus [39] and another study using a small sample showed that LRRK2 mutations may increase FTD risk [40].
Together these results suggest that the extended LRRK2 locus might influence FTD despite common genetic variants within LRRK2 not reaching genome-wide significance in large FTD-GWAS [5]. We suggest that increased expression of LRRK2 in microglia results in proinflammatory responses, possibly by modulating TNF-alpha secretion (Tumor-Necrosis Factor) [41]. TBKBP1 also modulates TNF-alpha signaling by binding to TBK1 (TANK-binding kinase 1) [42]; rare pathogenic variants in TBK1 cause FTD-ALS [43,44,45]. Importantly, elevated CSF levels of TNF-alpha are a core feature of FTD [6,46]. Building on these findings, in our bioinformatics 'network' based analysis, we found that both LRRK2 and TBKBP1 interact with genes within the HLA region ( Figure 3). Further, physical interactions between MAPT and the HLA network are compatible with research suggesting that under different conditions reactive microglia can either drive or mitigate tau pathology [47,48]. MAPT mutations, which disrupt the normal binding of tau protein to tubulin, account for a large proportion of familial FTD cases [49]. Together, these findings suggest that LRRK2, TBKBP1, and MAPT may, at least in part, influence FTD pathogenesis via HLA-related mechanisms.
This study has limitations. Particularly, in the original datasets that form the basis of our analysis, diagnosis of FTD was established clinically. Given the clinical overlap among neurodegenerative diseases, we cannot exclude the potential influence of clinical misdiagnosis. As such, our findings would benefit from confirmation in large pathologically confirmed cohorts. In addition, given the complex interconnectedness of the HLA region (see Figure 3), we were not able to define the specific gene(s) on chromosome 6 responsible for our pleiotropic signal. However, given the number of HLA loci associated with increased FTD risk, it may be the case that no single HLA variant will be clinically informative; rather, an additive combination of these microgliaassociated genetic variants may better inform FTD risk.
In conclusion, across a large cohort (total n = 192,886 cases and controls), we leveraged pleiotropy between FTD-related disorders and immune-mediated disease to identify several genes within the HLA region that are expressed within microglia and differentially expressed in the brains of patients with FTD. Building on prior work [6,7], our results suggest that immune dysfunction is central to the pathophysiology of a subset of FTD patients. These findings have important implications for future work focused on monitoring microglial activation as a marker of disease progression and on developing anti-inflammatory therapies to modify disease outcomes in patients with FTD.

FINANCIAL DISCLOSURE
No funding bodies had any role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.