Abstract Several genes associated with periodontitis have been identified through genome-wide association studies (GWAS); however, known genes only explain a minority of the estimated heritability. We aimed to explore more susceptibility genes and the underlying mechanisms of periodontitis. Firstly, a genome-wide meta-analysis of 38,532 patients and 316,185 healthy controls was performed. Then, cross- and single-tissue transcriptome-wide association studies (TWAS) were conducted based on GWAS summary statistics and the Genotype-Tissue Expression (GTEx) project. Risk genes were evaluated to determine if they were differentially expressed in periodontitis sites compared with unaffected sites using public datasets. Finally, gene co-expression network analysis was conducted to identify the functional biology of the susceptible genes. A total of eight single nucleotide polymorphisms (SNPs) within the introns of lncRNA LINC02141 approached genome-wide significance after meta-analysis. EZH1 was identified as a novel susceptibility gene for periodontitis by TWAS and was significantly upregulated in periodontitis-affected gingival tissues. EZH1 co-expression genes were greatly enriched in the cell-substrate junction, focal adhesion and other important pathways. Our findings may offer a fundamental clue for comprehending the genetic mechanisms of periodontitis. Keywords: genome-wide association study (GWAS), transcriptome-wide association study (TWAS), periodontitis 1. Introduction Periodontitis is a multifactorial disease with bacteria, genetic and environmental factors playing etiologic roles [[32]1]. It is characterized by the loss of periodontal tissue support manifested by clinical attachment loss (CAL) and radiographically assessed alveolar bone loss, presence of periodontal pocketing and gingival bleeding [[33]2,[34]3]. Periodontitis is initiated and sustained by microbial infection in subgingival dental biofilm by periodontal pathogens including Porphyromonas gingivalis, Tannerella forsythia, and so on [[35]4]. The innate and adaptive immune responses against these periodontal bacteria account for most of the periodontitis-related tissue damage [[36]5]. Genetic factors considerably contributed to the development of periodontitis. Periodontitis had approximately 50% heritability after adjusting for behavioral variables in a twin study, suggesting a substantial genetic contribution to the etiology of periodontitis [[37]6]. A systematic review revealed that twenty-nine per cent of the variance of periodontitis traits is related to genetic factors based on twin and family studies [[38]7]. Variants in at least 65 genes, especially genes involved in inflammation, have been implicated to be associated with periodontitis [[39]8,[40]9]. Besides the genetic factors, the impacts of environmental risk have been acknowledged for decades. Tobacco use is one of the most significant avoidable risk factors for the occurrence and development of periodontal disease [[41]10]. Moreover, medicines, poverty, obesity, uncontrolled diabetes and mental stress are also strongly associated with periodontitis [[42]11]. Thus, the severity of periodontitis is determined by genetic and environmental factors [[43]12]. In recent years, genome-wide association studies (GWAS) have identified several loci in the genome linked to periodontitis. According to the GWAS catalog, nineteen GWAS studies on periodontitis have been conducted [[44]13]. A total of six variants (rs1537415, GLT6D1; rs242016, EFCAB4B; rs729876, SHISA9; rs11084095, SIGLEC5; rs4284742, SIGLEC5; and rs2738058, DEFA1A3) meet the significant threshold (p < 5 ×10^−8) [[45]14,[46]15,[47]16,[48]17]. Although GWASs have achieved success in identifying associated loci, the statistical power and biological interpretation of GWAS results remain to be addressed [[49]18]. Meta-analysis, gene-gene and gene-environment interactions analysis, larger samples, wider computing and some other methods are used to improve statistical power [[50]18]. Many GWAS hits are non-coding variants residing in intergenic or intronic regions, and little is known about their biological significance. Furthermore, strong linkage disequilibrium (LD) within gene-content haplotypes makes it difficult to identify candidate variants and genes [[51]19,[52]20]. To overcome this issue, it is essential to translate GWAS findings into a better understanding of the biology underlying periodontitis risk by integrating other methods. The Genotype-Tissue Expression (GTEx) database of expression quantitative trait loci (eQTL), a tissue bank to investigate the correlation between single nucleotide polymorphisms (SNPs) and gene expression in human tissues, serves to fill the gap in variants, genes, and complex traits [[53]21]. Transcriptome-wide association study (TWAS), an integrated analysis of eQTL mapping with GWAS, can be applied to explore the most likely target genes and reveal the regulatory mechanisms underlying complex traits [[54]22]. Single-tissue TWAS methods such as MetaXcan [[55]23], PrediXcan [[56]24], and FUSION [[57]25] are widely used for diseases or traits with relevant tissues. Considering a certain number of relevant tissues were scarce and the similarity in transcription regulation across multiple tissues, researchers created a cross-tissue analysis method called UTMOST (Unified Test for MOlecular SignaTures) [[58]20,[59]26]. Cristina et al. used UTMOST to analyze the largest autism spectrum disorder (ASD) GWAS summary statistics and found that NKX2-2 and BLK were associated with ASD but not restricted to the brain tissue [[60]27]. Using UTMOST and FUSION in combination, Zhu et al. identified two novel susceptibility genes (DCAF16 and CBL) for lung cancer [[61]28]. In this study, we aimed to explore new genes associated with periodontitis and discover the biological mechanism underlying the association. A meta-analysis was performed to identify variants associated with periodontitis in Europeans using two large-scale GWAS datasets. We then conducted a TWAS analysis using both the UTMOST and FUSION methods to explore more susceptibility genes. Potential genes were validated using two gene expression datasets. Finally, downstream pathway enrichment analysis was conducted to confirm the function of susceptible genes in the etiology of periodontitis. 2. Materials and Methods 2.1. Periodontitis GWAS Data Summary statistics for periodontitis from the Gene-Lifestyle Interactions in Dental Endpoints (GLIDE) consortium ([62]https://data.bris.ac.uk/data/dataset/2j2rqgzedxlq02oqbb4vmycnc2, accessed on 16 June 2022) and FinnGen ([63]https://www.finngen.fi/en/access_results, accessed on 16 June 2022) were downloaded ([64]Table S1) [[65]29,[66]30]. GLIDE consists of systematic meta-analysis statistics based on nine studies with 17,353 periodontitis cases and 28,210 controls from Europe ranging from 17 to 93 years, the specifics of which have been described before [[67]29]. FinnGen comprises 21,179 periodontitis cases and 287,975 controls with a median age of 63 years of European descent. Additional information on recruitment, genotyping, and quality control was previously provided [[68]30]. Before analysis, LiftOver altered the FinnGen location (hg38) to match the location of GLIDE (hg19) ([69]https://genome.sph.umich.edu/wiki/LiftOver, accessed on 21 October 2022). Then, results from GLIDE and FinnGen were incorporated into a fixed effect inverse-variance weighted meta-analysis using Metal software, with a classical approach using effect size estimates and standard errors [[70]31]. The Manhattan and quantile-quantile plots were generated using the R package “qqman”. LocusZoom was used to visualize the genomic regions of significance ([71]http://locuszoom.sph.umich.edu/, accessed on 21 October 2022) [[72]32]. 2.2. Cross-Tissue TWAS Analysis Using UTMOST UTMOST integrated GWAS summary statistics and gene expression data of 44 tissues from 450 individuals in GTEx V7 to execute a cross-tissue association test [[73]20]. First, we downloaded UTMOST pre-calculated covariance matrices ([74]https://github.com/Joker-Jerome/UTMOST, accessed on 21 October 2022). Single-tissue association tests for 44 tissues were then performed by integrating GWAS periodontitis summary statistics and genetic expression weights. Finally, the relationships between 17,290 genes and periodontitis in 44 tissues were calculated by a joint GBJ test in combination with single-tissue gene-trait association results. The transcriptome-wide significance for the joint test was set as p-value < 1 × 10^–4. 2.3. Single-Tissue TWAS Analysis Using FUSION In our study, a single-tissue TWAS analysis was performed using FUSION [[75]25] to minimize false positive errors and train independent imputation models for various tissues. The precomputed predictive models for tissues in GTEx V7 were downloaded from the official FUSION website ([76]http://gusevlab.org/projects/fusion/, accessed on 21 October 2022). We then estimated the association of each gene with periodontitis, taking precomputed gene expression weights together with GWAS summary statistics. LD references were derived for Europeans from