Identifying risk loci for FTD and shared genetic component with ALS: A large-scale multitrait association analysis

FTD


Introduction
Frontotemporal dementia (FTD), characterized by early-onset relative memory retention, progressive behavioral abnormalities, personality changes, and language barriers (Boeve et al., 2022), is one of the most common neurodegenerative diseases worldwide.It is estimated that, among more than 55 million people living with dementia, approximately 20% of presenile dementia (age of onset ≤65 years) patients are diagnosed with FTD (Han et al., 2021).What's more, FTD is irreversible and leads to a remarkable reduction in patients' life expectancy (Deleon and Miller, 2018), generating a serious economic burden which is 2-3 times greater than that of dementia (Khoury et al., 2021;Tsai and Boxer, 2016).Further, since FTD is highly heritable, with one-third to one-half of FTD cases exhibiting genetic predisposition (Moore et al., 2020;Rohrer and Warren, 2011), it is of great significance to probe into the complex genetic architecture of FTD, thus to better elaborate its underlying pathogenic mechanisms and to investigate potential intervention targets.
FTD so far (Broce et al., 2018;Ferrari et al., 2014); however, the sample sizes of those studies were small, in which the hitherto largest FTD GWAS included only 3526 cases and 9402 neurologically healthy individuals (Ferrari et al., 2014).In addition to collecting sufficient FTD cases in future genome research, from the perspective of methodology, leveraging advanced statistical methods to alleviate the problem of small sample size and improve the power is another efficient strategy to explore genetic mechanisms of FTD in deeper detail.Among many tools proposed recently (Baselmans et al., 2019;Zeng et al., 2018), multitrait analysis based on summary statistics has become an effective statistical method.By borrowing pleiotropic information from genetically related traits without accessing individual-level data, the efficacy of single trait GWAS can be significantly improved in multitrait analysis (Turley et al., 2018).
Alzheimer's disease (AD), Parkinson's disease (PD), and amyotrophic lateral sclerosis (ALS) share extensive similarities with FTD, but the genetic overlap between ALS and FTD is more pronounced, as demonstrated in previous studies (Ferrari et al., 2017;Qiao et al., 2023).Clinically, although ALS and FTD primarily affect the motor and frontotemporal areas of the brain, respectively, some clinical manifestations of ALS (e.g., cognitive impairment) overlap with that of FTD (Turley et al., 2018).Approximately 50% of ALS patients also develop impairments in at least 1 cognitive domain, including executive function, social behavior, or language indicative of extra motor neurodegeneration in the frontal and temporal lobes, and about 10% of these patients occur multidomain impairments consistent with FTD (Montuschi et al., 2015;Ringholz et al., 2005).Genetically, ALS and FTD commonly feature trans-active response DNA-binding protein 43 kDa (TDP-43) pathology (Ma et al., 2022;Neumann et al., 2006) and can both be caused by pathogenic C9orf72 repeat expansions (DeJesus-Hernandez et al., 2011;Renton et al., 2011).The significant genetic overlap between ALS and FTD has been also supported by other studies (Diekstra et al., 2014;Karch et al., 2018).Consequently, ALS and FTD are sometimes thought to represent a continuous disease spectrum (Ferrari et al., 2011;Lomen-Hoerth, 2011).
Given the paucity of knowledge regarding the genetic architecture underlying FTD, we here seek to identify risk loci and genes relevant to FTD with powerful pleiotropy methods informed by ALS and intend to discover common genetic backgrounds shared between them through conducting a comprehensive multitrait association analysis with only summary-level data (Baselmans et al., 2019;Turley, 2018;Zeng et al., 2018).The current sample size of the ALS GWAS is much larger than that of FTD; we thus expect there exists substantial genetic similarity which can be integrated into the FTD GWAS using multitrait analysis methods.Multitrait analysis of the 2 diseases could be more powerful not only in deeply exploring FTD-associated genetic variants by utilizing information shared with ALS but also in providing pleiotropic loci related to FTD and ALS simultaneously.The shared genetic loci could point to commonly biological mechanisms and thus be targeted for intervention of the 2 diseases.

Methods
We here offered an overview of all statistical methods and tools used and showed more descriptions in the Additional Notes.Technical details of these methods could be found in respective original papers.Our analysis flowchart is illustrated in Fig. 1.We obtained the largest GWAS summary statistics of FTD (GWAS FTD ) (3526 cases and 9402 controls) (Ferrari et al., 2014) and ALS (27,205 cases and 110,881 controls) (van Rheenen et al., 2021) to date and reserved 3,535,999 shared SNPs (Supplementary Table S1).We noted the presence of FTD overlapping with motor neuron disease (FTD-MND) samples (9.28%) in the FTD cohort (Ferrari et al., 2014); although MND includes different neurological disorders, Ferrari et al. (2014) only included ALS.We also obtained the largest GWAS summary statistics of PD (17,996 cases and 16, 502 controls) (Blauwendraat et al., 2019) and AD (71,880 cases and 383, 378 controls) (Schwartzentruber et al., 2021).
We first performed linkage disequilibrium score regression (LDSC) to assess cross-trait genetic correlation (ρ g ) between 2 diseases of interest (Bulik-Sullivan et al., 2015).Second, we applied multitrait analysis of GWAS (MTAG) (Turley et al., 2018) to analyze FTD and ALS jointly and obtained the MTAG summary statistics of FTD (MTAG FTD ).We further conducted association analysis based on subsets (ASSET) (Bhattacharjee et al., 2012) to validate the findings available from MTAG.
Third, we applied functional mapping and annotation of genetic association studies (FUMA) (Watanabe et al., 2017) to annotate genome-wide significant SNPs of MTAG FTD and identify significant risk loci.Based on the FUMA results, we employed conditional and joint association analysis (GCTA-COJO) to detect independent association signals in discovered risk loci (Yang et al., 2011(Yang et al., , 2012)), followed by Bayesian fine-mapping to obtain credible sets of candidate causal SNPs as well as colocalization analysis to pinpoint shared causal variants (Giambartolomei et al., 2014).
Fourth, to discover significant functional categories of various tissues, we implemented functional enrichment analysis using GWAS analysis of regulatory or functional information enrichment with LD correction (GARFIELD) (Iotchkova et al., 2019).
Finally, besides SNP-level analyses above, we further conducted gene-level analyses, including fast set-based association analysis (GCTA-fastBAT) (Bakshi et al., 2016;Yang et al., 2011) and multitissue transcriptome-wide association studies (TWAS) via joint-tissue imputation (Zhou et al., 2020) to prioritize FTD-associated genes.Differential expression analysis of genes was carried out to the reveal potential genetic mechanism of FTD-relevant loci (Watanabe et al., 2017).

Estimated genetic correlation
The liability-scale SNP-based heritability was 4.53% (standard error [se] = 0.64%) for ALS and 4.36% (se = 3.40%) for FTD, and there existed a positive genetic correlation between FTD and ALS ( ρg = 0.637, P = 0.032).We also estimated the genetic correlations between FTD and AD as well as between FTD and PD (Fig. 2); however, the 2 estimates were not statistically significant and much weaker than the genetic correlation between FTD and ALS, which is consistent with what has been uncovered in the previous study (Qiao et al., 2023) and is also our primary reason of selecting ALS in the present study.Because of the substantial genetic similarity between ALS and FTD, MTAG was expected to generate a great gain in statistical power and little inflation of false discovery when jointly analyzing the 2 diseases.

FTD-associated loci discovered by MTAG
Although it was not our primary objective, we considered the scenario where the genetic similarity information was integrated from FTD into ALS as a control analysis.Briefly, 302 ALS-associated SNPs were discovered in GWAS ALS (P GWAS < 5 × 10 − 8 ) (Fig. 3A); 296 ALSassociated SNPs were detected in MTAG ALS (P MTAG < 5 × 10 − 8 ) (Fig. 3B), among which 253 had P GWAS < 5 × 10 − 8 , while the rest had P GWAS ranging from 5.02 × 10 − 8 to 3.08 × 10 − 7 (Fig. 4).This result suggested that the statistical power of ALS association analysis could be slightly improved when leveraging the pleiotropy shared with FTD, although the sample size of FTD was much smaller than that of ALS.
After using MTAG, we observed the genomic factor was nearly unchanged in MTAG FTD (λ = 1.01) compared to that in GWAS FTD (λ = 1.01); the absence of inflation in test statistics suggested the homogeneous assumption of MTAG likely held in our analysis.A total of 261 FTD-associated SNPs were detected in MTAG FTD (P MTAG < 5 × 10 − 8 ) (Supplementary Table S2 and Fig. 5A), leading to 6 independent risk loci (Table 1), whereas no genome-wide significant SNPs Fig. 1.Flowchart of various statistical analyses for the present work.Genome-wide genetic correlation analysis among frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS) was first performed by linkage disequilibrium score regression, followed by multitrait meta-analysis of ALS and FTD genome-wide association studies via multitrait analysis of genome-wide association study and association analysis based on subsets; then, single nucleotide polymorphism-level analysis and gene-level analysis were further implemented.Finally, differential expression analysis was performed for discovered genes and loci to reveal the genetic architecture of FTD.Abbreviations: ALS, amyotrophic lateral sclerosis; ASSET, association analysis based on subsets; COJO, conditional and joint association analysis; FTD, frontotemporal dementia; FUMA, functional mapping and annotation of genetic association studies; GARFIELD, functional enrichment analysis using GWAS analysis of regulatory or functional information enrichment with LD correction; fastBAT; fast set-based association analysis; GWAS, genome-wide association study; LDSC, linkage disequilibrium score regression; SNP, single nucleotide polymorphism; TWAS, transcriptome-wide association analysis.
were discovered in the original GWAS FTD , largely due to insufficient sample size (Fig. 5B).

Validated association signals with ASSET
In the 1-sided ASSET analysis (Supplementary Table S3), 70 SNPs (all located at 14q12, including rs229243, one of the 6 risk loci identified above) discovered by MTAG were included only in the subset {ALS}, and the remaining 191 significant SNPs were verified in the subset {ALS, FTD} (P MTAG < 5 × 10 − 8 and P ASSET < 5 × 10 − 8 ).No SNPs were only included in the subset {FTD}.The 2-sided ASSET analysis showed that 12 SNPs were positively associated with FTD but negatively related to ALS, and 11 SNPs had opposite effect direction (Supplementary Table S4).All the 23 SNPs were included in the subset {ALS} in the 1side ASSET analysis and had consistent effect directions with the effect sizes in GWAS ALS and GWAS FTD .
Five risk loci were confirmed in the subset {ALS, FTD}, indicating that the detected significant loci in MTAG FTD were potentially pleiotropic SNPs shared with ALS.We thus focused on the 5 loci (with 190 SNPs) in our subsequent analyses.The top SNPs in these loci were mapped to MOBP, ERGIC1, MOB3B (with IFNK and C9orf72), UNC13A, and RNF114 (with SLC9A8, SPATA2, and SNAI1), respectively (Supplementary Table S5).We performed GCTA-COJO for each risk locus but did not identify additional independent SNPs conditioning on the top one (Supplementary Table S6).

Functional annotations
The variant annotations through FUMA for the 190 SNPs are summarized in Supplementary Table S7.Most of the discovered SNPs (94.24%) were in noncoding regions such as intronic and intergenic regions; only a few were exonic variants, and all were noncoding RNAs (5.76%).The most significant exonic variant of noncoding RNA was rs700791 (P MTAG = 2.69 × 10 − 39 , mapped to C9orf72) in 9p21.2.
RegulomeDB scores showed the 2 variants with the top 5 combined annotation-dependent depletion (CADD) scores, rs3736319 (P MTAG = 1.87 × 10 − 13 , CADD score: 18.5) in 9p21.2, which was an exonic variant of MOB3B, and rs700795 (P MTAG = 1.08 × 10 − 36 , CADD score: 16.7) in 9p21.2 locating in MOB3B, likely affected binding and linked to expression of a gene target.Both rs3736319 and rs700795 were verified to be in the optimal subset {ALS, FTD}, which further highlighted the role of MOB3B underlying the shared genetic etiology of FTD and ALS.

SNP credible sets within loci
A total of 73 SNPs were identified in the 90% credible sets of the 5 risk loci (Supplementary Table S8).Among these, the 90% credible sets of 19p13.11contained only the top SNP rs12608932 (PP = 0.999, mapped to UNC13A), while 6 SNPs were identified in the 90% credible set of 5q35.1 with the top SNP rs517339 (PP = 0.175, mapped to ERGIC1) included.Particularly, only 10 SNPs were detected in the 90% credible set of 9p21.2, which contained 112 significant SNPs, but the top SNP rs12554036 in this locus still had a high posterior probability (PP = 0.139, mapped to MOB3B).For the other 2 loci, there were multiple SNPs in their 90% credible sets, containing 15 SNPs in 20q13.13 and 31 SNPs in 3p22.1 (Table 2).On average, the potential candidate causal SNPs identified by Bayesian fine-mapping analysis improved accuracy by 47.9% compared to the initially discovered SNPs.

Colocalization analysis
Among the 5 risk loci, the colocalization analysis identified 2 loci (i.e., 9p21.2 and 19p13.11)with PP 4 larger than 0.75 (PP 4 = 0.764 and 0.962, respectively) (Table 3).In line with the result from ASSET which produced the optimal subset {ALS, FTD} for them, the colocalization analysis found that ALS and FTD had causal variants in 9p21.2 (PP 4 = 0.764) and 19p13.11(PP 4 = 0.962).Specially, the top SNP of 19p13.11was identified to be a potential causal variant (rs12608932, mapped to UNC13A) shared between ALS and FTD.The PP 4 of each SNP in every locus is given in Supplementary Table S9.

Functional enrichment analysis
The enrichment results of functional enrichment analysis using GARFIELD are summarized in Supplementary Table S10.Although no significant enrichments were observed for these detected SNPs in MTAG FTD (P < 4.98 × 10 − 5 ), the SNP enrichments in different chromatin states regions revealed that the transcribed regions were highly enriched in the fetal brain (odds ratio = 13.8,P = 0.023) and the human glioma cells of brain tissue (odds ratio = 10.1,P = 0.041), which were still informative for us to understand the function of discovered FTDassociated SNPs.

Associated genes discovered by gene-level analysis
According to GCTA-fastBAT, 26 candidate genes were identified to be related to FTD (Supplementary Table S11); in addition, 55 unique FTD-associated genes were detected by TWAS (29 of them in 13 different brain tissues) (Supplementary Table S12).Among these genes, 12 genes were simultaneously discovered by GCTA-fastBAT and TWAS (Table S13).

Differential expression analysis
We performed differential expression analysis for the 15 candidate  FTD-related genes using FUMA.We discovered the differentially expressed genes were significantly enriched in heart left ventricular, kidney cortex, and some brain regions (e.g., anterior cingulate cortex BA24, cortex, caudate basal ganglia, and amygdala) in terms of gene expression level across distinct tissues of the genotype-tissue expression project.The top 10 enriched tissues are shown in Fig. 6.

Genetic correlation and novel risk loci for FTD
Using the hitherto largest GWAS summary statistics data of FTD and ALS (Ferrari et al., 2014;van Rheenen et al., 2021), we have carried out a comprehensive large-scale genome-wide multitrait association analysis, followed by SNP-level and gene-level analyses, to deeply investigate the genetic architecture of FTD as well as the shared genetic etiology of ALS and FTD.We identified a significant genetic correlation between the 2 diseases, in contrast to the previously high but insignificant estimate ( ρg = 0.59, P = 0.15) (van Rheenen et al., 2021).We again calculated the genetic correlation with only HapMap3 SNPs and found the recalculated genetic correlation between ALS and FTD was high but not statistically significant ( ρg = 0.56, P = 0.18), which was in line with the previous result (van Rheenen et al., 2021).This implies that the use of only HapMap3 SNPs in genetic correlation calculation is suboptimal as it ignored some informative variants.Although the presence of the FTD overlapping with motor neuron disease samples in the FTD cohort may have affected our results, it was a relatively small proportion (9.28%), and the linkage disequilibrium score regression was robust to sample overlap.Nevertheless, the genetic correlation between FTD and ALS was likely overestimated.

Credible sets and shared causal variants in risk loci
The colocalization analysis demonstrated that ALS and FTD had causal variants in 9p21.2 (PP 4 = 0.764) and 19p13.11(PP 4 = 0.962), and that the top SNPs of 19p13.11were detected to potentially causal variants.In addition, the 90% confidence set for 19p13.11identified only the top SNP rs12608932, with a posteriori probability of 0.999, as the only causal association signal.Although this top SNP was located in the noncoding region (intronic), it may be included in the UNC13A transcript when the TDP-43 levels are lowered, affecting the UNC13A protein expression and thus increasing the risk of developing ALS and FTD (Ma et al., 2022).Existing studies have linked rs12608932 to ALS (van Es et al., 2009;Yang et al., 2019) and suggested that this variant may progress FTD in sporadic ALS cases (Placek et al., 2019).In addition, previous studies showed that FTD and ALS had significant genetic overlap at 9p21.2 and 19p13.11(Karch et al., 2018;Turley et al., 2018).

FTD-associated genes
Through the SNP-level and gene-level analyses, we ultimately identified 15 candidate FTD-associated genes (with C9orf72, UNC13A, SLC9A8, SPATA2, MOBP, SNAI1, and MOB3B discovered in both analyses).Among them, seven overlap with the prioritized genes of ALS identified in existing studies (van Rheenen et al., 2021), including C9orf72, SLC9A8, UNC13A, MOBP, SCFD1, ERGIC1, and SPATA2.For these discovered FTD-associated genes, previous studies identified MOBP, C9orf72, MOB3B, ERGIC1, and UNC13A as FTD-associated genes Fig. 4. Scatter plots of 296 significant amyotrophic lateral sclerosis-related single nucleotide polymorphisms (SNPs) in GWAS ALS and MTAG ALS .The plot demonstrates that 253 genome-wide significant SNPs in GWAS ALS remained significant and that there were 43 SNPs which became significant after using multitrait analysis of GWAS; the red dotted lines represent the threshold of − log 10 (5 × 10 − 8 ).There was a high correlation between the effect sizes of GWAS ALS and MTAG ALS for those SNPs, with an estimated R 2 of 0.999.Abbreviations: ALS: Amyotrophic lateral sclerosis; GWAS, genome-wide association study; MTAG, multitrait analysis of GWAS; SNP, single nucleotide polymorphism.(Ferrari et al., 2014(Ferrari et al., , 2015;;Karch et al., 2018).For example, UNC13A is one of the top hits for ALS and ALS-FTD in multiple studies (Diekstra et al., 2014;Placek et al., 2019;van Es et al., 2009).UNC13A encodes a large protein highly expressed in the nervous system, where it localizes to most synapses in the central nervous system and neuromuscular junctions and plays an essential role in the vesicle priming step prior to synaptic vesicle fusion (Bohme et al., 2016;Lipstein et al., 2017).Additionally, a study revealed UNC13A as a novel link between ALS and TDP-43-positive FTD (Diekstra et al., 2014), which further discovered synaptic defects as a common disease mechanism shared by ALS and FTD and corroborated the role of UNC13A and synaptic mechanisms in neuronal degeneration.

Biological mechanisms for genetic loci of FTD
In the functional enrichment analysis, FTD-related loci displayed a high enrichment in fetal brain and the human glioma cells of brain tissue.It was shown that, as the major component of the cytoplasmic inclusions in the central nervous system of ALS and FTD patients, aggregates of TDP43 were induced in the human glioma U251 cells (Ding et al., 2015).Moreover, the formation of cytoplasmic inclusions, which compose of misfolded proteins in neuronal and glial cells, is a key neuropathological feature of neurodegenerative diseases (Li and Haney, 2020;Parakh and Atkin, 2016).Previous studies also demonstrated that widespread loss of the silencing epigenetic marks in astrocytes and neurons is accompanied by hippocampal-dependent cognitive impairment (Jury et al., 2020).
The differential expression analysis for FTD-associated genes discovered significant enrichments in the heart left ventricular, kidney cortex, and some brain regions.It was recently found that atrophy of the anterior cingulate region in P301L MAPT mutation carriers caused FTD (Clarke et al., 2021).The anterior cingulate was thought to play a critical role in social cognition via contextual integration and evaluating the behavior of others (Apps et al., 2016), and its damage would lead to inattention and apathy (Bush et al., 2000), which are the core symptoms for the diagnosis of behavioral variant FTD (Rascovsky et al., 2011).
On the other hand, it was uncovered that patients with ALS-FTD exhibited pathologic changes in the regions of putamen, globus pallidus, hippocampi, caudate, and accumbens nuclei (Machts et al., 2015).Another study reported that nucleus accumbens and caudate nucleus were more severely atrophied in FTD than in AD and subjective complaints, and that hippocampus, amygdala, thalamus, and putamen were smaller in FTD patients compared to subjective complaints (Möller et al., 2015).
However, to our knowledge, no previous studies have directly linked the left ventricle to FTD.A case-control study suggested an effect of cardiovascular risk factors on FTD (Golimstok et al., 2014).Another study reported that bradycardia (<60 beats/min) was significantly more frequent and the average systolic blood pressure was lower in FTD patients (Bayon et al., 2014).These findings likely offered suggestive evidence for the connection between heart and FTD.
Existing studies have suggested that individuals at all stages of chronic kidney disease have higher risk of developing cognitive Note: Loci were identified by functional mapping and annotation of genetic association studies after merging the linkage disequilibrium blocks of significant single nucleotide polymorphisms (SNPs) that were closely located to each other (<250 kb); n.sig represents the number of SNPs with P MTAG < 5 × 10 − 8 in the locus; and the trait subset is the optimal trait subset of the top SNP identified by association analysis based on subsets.Key: ALS, amyotrophic lateral sclerosis; ASSET, association analysis based on subsets; FTD, frontotemporal dementia; GWAS, genome-wide association study; MTAG, multitrait analysis of GWAS; SNP, single nucleotide polymorphism.Note: PP 1 is the posteriori probability that amyotrophic lateral sclerosis is associated with this locus; PP 2 is the posteriori probability that frontotemporal dementia is associated with this locus; PP 3 is the posteriori probability that both diseases are associated but with distinct causal variants; PP 4 is the posteriori probability that both diseases are associated and share a single causal variant; best causal denotes the single nucleotide polymorphism with the highest PP 4 to be the causal variant in the risk loci.Key: PP, posterior probability.a represents that the potential causal single nucleotide polymorphism is the top single nucleotide polymorphism in that locus.b represents PP 4 > 0.75.
disorders and dementia and have explored the pathological mechanisms of the kidney-brain axis (Bugnicourt et al., 2013;Xie et al., 2022).Our study provides supportive evidence for exploring the pathophysiological interactions between renal impairment and brain function.

Cross-trait meta-analysis
We also conducted a cross-trait meta-analysis for ALS and FTD via the inverse-variance-weighted method (Additional Notes).We identified 11 independent loci, which duplicated 6 of the 15 FTD-associated genes we discovered (i.e., C9orf72, UNC13A, G2E3, SLC9A8, MOB3B, and ERGIC1).However, we did not include this finding as our primary result as ALS accounted for only 9.28% of participants in the FTD cohort.There were certainly many studies (Adewuyi et al., 2022;Guo, 2020) that implemented similar cross-trait meta-analyses, which we believed to be useful and deemed as an effective complementarity.

Important statistical and scientific implications of this work
Our work is of statistical and scientific importance.First, we confirmed the existence of substantial genetic foundation shared by FTD and ALS, which could not be observed within a single trait study.A recent study showed that no genetic overlap with other neurodegenerative diseases was found for FTD due to its limited sample size (Koretsky et al., 2023); however, we identified common genetic components between ALS and FTD, which implies that our used multitrait association analysis is cost-effective and powerful to mine the data of neurodegenerative diseases.
Second, the FTD-specific summary statistics from MTAG were enriched and aided by the genetic overlap of ALS, which could facilitate the discovery of common genetic background of the 2 diseases.For example, among the 5 FTD-associated loci, the colocalization analysis identified 2 loci sharing the causal variants between FTD and ALS, with MOB3B playing an important role in the common genetic foundation of FTD and ALS.Therefore, our work demonstrates that it is feasible to utilize powerful multitrait analysis approach to probe into shared genetic foundation underlying diseases to a great extent.

Limitations of the present work
Our study is not without limitations.First, we only focused on European ancestry because the large-scale GWASs of FTD and ALS are currently very scarce for non-European populations (Mills and Rahal, 2019).Therefore, it is not easy to guarantee that the findings regarding the genetic architecture of FTD and the common genetic component underlying FTD and ALS are broadly portable.More investigation is needed to explore the genetic architecture of FTD in other populations.
Second, the genetic associations for rare variants of ALS and FTD were unable to be evaluated since SNPs with minor allele frequency (MAF) <0.01 were automatically filtered in the MTAG analysis (Turley et al., 2018) or removed by our quality control.Third, although our results confirm the common genetic components between ALS and FTD, the related biological mechanisms need to be further studied.

Conclusions
We provide strong evidence of genetic correlations between FTD and ALS and identified novel FTD-associated loci and genes by leveraging such common genetic overlap.Our findings not only advance the understanding of genetic determinants of FTD but also provide novel insights into the shared genetic etiology of FTD from functional and biological pathway levels.Shared biological mechanisms could provide novel insight to simultaneously prevent and treat FTD and ALS.

Ethics approval and consent to participate
Ethical approval and consent were not available for this study since the summary-level data of frontotemporal dementia and amyotrophic lateral sclerosis were downloaded from public portals.

Fig. 2 .
Fig. 2. Genetic correlation of 4 neurodegenerative diseases with LDSC.The color and the first line of numbers indicate the magnitude of genetic correlation; the second line of numbers in the parentheses is the corresponding unadjusted P value.Abbreviations: AD, Alzheimer's disease; ALS, amyotrophic lateral sclerosis; FTD, frontotemporal dementia; PD, Parkinson's disease.

Fig. 5 .
Fig. 5. Manhattan plots for (A) MTAG FTD and (B) GWAS FTD .The x-axis denotes the chromosomal position, and the y-axis shows the − log 10 (P) value; the reference line stands for the threshold of P = 5 × 10 − 8 ; the chromosome regions with risk loci are shown.

Fig. 6 .
Fig. 6.Differential expression analysis for 15 candidate FTD-related genes.Bonferroni′s method-adjusted P values are shown in the x-axis with a scale of − log 10 ; the red dotted line represents the Bonferroni′s method-corrected threshold for multiple comparisons.

Table 1
Identified risk loci and corresponding top variants with P MTAG < 5 × 10 − 8 in MTAG FTD

Table 2
Summary of fine-mapping results in the 5 risk loci MTAG < 5 × 10 − 8 in the locus; n.set represents the number of SNPs in the 90% credible sets; the top SNP is the variant with the highest posterior probability and is also the top lead SNP in the locus identified by functional mapping and annotation of genetic association studies; PP indicates a posteriori probability of each SNP providing causality for each locus; accuracy gain = (n.sig− n. set)/n.sig.Key: PP, posteriori probability; SNP, single nucleotide polymorphism.

Table 3
Summary of colocalization results in the discovered 5 risk loci