Fine-mapping SLE-MHC associations revealed independent contributions of HLA missense variants andC4copy number variations

Fine-mapping SLE-MHC associations revealed independent contributions of HLA missense variants and C4 copy number variations | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Fine-mapping SLE-MHC associations revealed independent contributions of HLA missense variants and C4 copy number variations View ORCID Profile Chae-Yeon Yu , Dong Mun Shin , Sung Min Kim , Yui Taek Lee , View ORCID Profile Sungwon Jeon , Sehwan Chun , So-Young Bang , Hye-Soon Lee , Xianyong Yin , Yong Cui , Xuejun Zhang , Jong Bhak , Soon Ji Yoo , View ORCID Profile Young Jin Kim , Bong-Jo Kim , View ORCID Profile Sang-Cheol Bae , View ORCID Profile Kwangwoo Kim doi: https://doi.org/10.1101/2024.11.21.24317596 Chae-Yeon Yu 1 Department of Biology, Kyung Hee University , Seoul 02447, Republic of Korea 2 Department of Biomedical and Pharmaceutical Sciences, Kyung Hee University , Seoul 02447, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Chae-Yeon Yu Dong Mun Shin 3 Division of Genome Science, Department of Precision Medicine, National Institute of Health , Cheongju 28159, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sung Min Kim 3 Division of Genome Science, Department of Precision Medicine, National Institute of Health , Cheongju 28159, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site Yui Taek Lee 1 Department of Biology, Kyung Hee University , Seoul 02447, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sungwon Jeon 4 Korean Genomics Center, Ulsan National Institute of Science and Technology , Ulsan 44919, Republic of Korea 5 AgingLab , Ulsan 44919, Republic of Korea 6 Geromics Inc. , Suwon 16226, Republic of Korea 7 Clinomics Inc. , Ulsan 44919, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sungwon Jeon Sehwan Chun 1 Department of Biology, Kyung Hee University , Seoul 02447, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site So-Young Bang 8 Department of Rheumatology, Hanyang University Hospital for Rheumatic Diseases , Seoul 04763, Republic of Korea 9 Hanyang University Institute for Rheumatology Research and Hanyang Institute of Bioscience and Biotechnology , Seoul 04763, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hye-Soon Lee 8 Department of Rheumatology, Hanyang University Hospital for Rheumatic Diseases , Seoul 04763, Republic of Korea 9 Hanyang University Institute for Rheumatology Research and Hanyang Institute of Bioscience and Biotechnology , Seoul 04763, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site Xianyong Yin 10 Department of Epidemiology, School of Public Health, Nanjing Medical University , Jiangsu, China Find this author on Google Scholar Find this author on PubMed Search for this author on this site Yong Cui 11 Department of Dermatology, China-Japan Friendship Hospital, Capital Medical University , Beijing, China 12 Department of Dermatology, China-Japan Friendship Hospital , Beijing, China Find this author on Google Scholar Find this author on PubMed Search for this author on this site Xuejun Zhang 13 Institute of Dermatology, Anhui Medical University , Hefei, Anhui, China Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jong Bhak 4 Korean Genomics Center, Ulsan National Institute of Science and Technology , Ulsan 44919, Republic of Korea 5 AgingLab , Ulsan 44919, Republic of Korea 14 Department of Biomedical Engineering, College of Information-Bio Convergence Engineering, Ulsan National Institute of Science and Technology , Ulsan 44919, Republic of Korea 15 Personal Genomics Institute, Genome Research Foundation , Osong, 28160, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site Soon Ji Yoo 1 Department of Biology, Kyung Hee University , Seoul 02447, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site Young Jin Kim 3 Division of Genome Science, Department of Precision Medicine, National Institute of Health , Cheongju 28159, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Young Jin Kim For correspondence: kkim{at}khu.ac.kr scbae{at}hanyang.ac.kr kbj6181{at}korea.kr inthistime{at}korea.kr Bong-Jo Kim 3 Division of Genome Science, Department of Precision Medicine, National Institute of Health , Cheongju 28159, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: kkim{at}khu.ac.kr scbae{at}hanyang.ac.kr kbj6181{at}korea.kr inthistime{at}korea.kr Sang-Cheol Bae 8 Department of Rheumatology, Hanyang University Hospital for Rheumatic Diseases , Seoul 04763, Republic of Korea 9 Hanyang University Institute for Rheumatology Research and Hanyang Institute of Bioscience and Biotechnology , Seoul 04763, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sang-Cheol Bae For correspondence: kkim{at}khu.ac.kr scbae{at}hanyang.ac.kr kbj6181{at}korea.kr inthistime{at}korea.kr Kwangwoo Kim 1 Department of Biology, Kyung Hee University , Seoul 02447, Republic of Korea 2 Department of Biomedical and Pharmaceutical Sciences, Kyung Hee University , Seoul 02447, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Kwangwoo Kim For correspondence: kkim{at}khu.ac.kr scbae{at}hanyang.ac.kr kbj6181{at}korea.kr inthistime{at}korea.kr Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Systemic lupus erythematosus (SLE) is a complex autoimmune disease with strong genetic associations within the major histocompatibility complex (MHC) region. Despite significant advances, precisely pinpointing the genetic variants that contribute to SLE risk within the MHC remains challenging. This study aimed to comprehensively profile SLE-driving variants using a newly developed East Asian MHC imputation reference panel, capable of simultaneously imputing diverse MHC variants, including multi-level variants of HLA genes and copy number variations (CNVs) of C4 elements, with high imputation accuracy. Applying this panel to two SLE genome-wide association study datasets, we uncovered the independent contributions from six amino acid positions altering the epitope-binding surfaces of HLA-DRB1 and HLA-C. Additionally, reduced C4A copy numbers and increased HERV copy numbers, collectively lowering C4 protein levels, were associated with increased SLE risk, independent of HLA variants. Our refined MHC-SLE association model provided superior explanations for SLE risk over previous association models. In summary, this study enhanced the understanding of HLA and C4 in SLE pathogenesis and holds promise for advancing MHC association studies for immune-mediated inflammatory disorders in East Asians using our MHC panel ( https://coda.nih.go.kr/usab/kis/intro.do ). Introduction Systemic lupus erythematosus (SLE) is a common polygenic autoimmune disorder characterized by complex dysregulations in both adaptive and innate immune systems, resulting in the breakdown of self-tolerance and impaired clearance of nuclear antigens 1 . As of 2024, genome-wide association studies (GWASs) have identified over 200 loci in the human genome associated with susceptibility to SLE 2 , 3 . Most of these associations are driven by single genetic signals, with several notable exceptions such as the major histocompatibility complex (MHC) region, which exerts particularly strong effects on SLE risk 2 . The MHC region on human chromosome 6 harbors a diverse array of immune-related genes, featuring human leukocyte antigens (HLAs), and complements, and cytokines. This region has been well documented for its extensive linkage disequilibrium (LD), exceptionally dense distribution of genetic variants, and the presence of copy number variations (CNVs). These factors contribute to the formation of various long-range haplotypes across the MHC region, including HLA classical alleles and other MHC genes. The resulting complex array of haplotypic combinations influences the functional diversity of immune system and susceptibility to various diseases. Additionally, this intricate genomic architecture has made the MHC region one of the most challenging loci to characterize in terms of individuals’ genetic variants and to dissect genetic association signals 4 . Our previous association fine-mapping studies of the MHC region have demonstrated the crucial role of structural changes in the epitope-binding groove of HLA-DR molecules in multiple Asian populations 5 – 7 . These changes are determined by HLA-DRB1 missense variants encoding amino acid positions 11, 13, 26, and 37. The haplotypes of residues at these positions accounted for the largest proportion of the MHC association effects on SLE risk, explaining both the previously reported associations of HLA-DRB1 classical alleles and the association heterogeneity between East Asians and Europeans, such as the association of HLA-DRB1*15:01 in Asians and HLA-DRB1*03:01 in Europeans 5 . In parallel, some studies have reported genetic associations of the complement component gene C4 , located near HLA-DRB1 , with SLE. C4 is crucial in the human immune system, facilitating the clearance of dead or damaged cell debris. The genetic variants of C4 are highly complex, involving various combinations of coding variants, endogenous retroviral sequence insertions, and CNVs. C4 encodes two functionally distinct isoforms, C4A and C4B , which differ by four amino acid residues within positions 1101 to 1106 and exhibit differential affinities for their molecular targets associated with different biological pathways 8 , 9 . Additionally, the insertion of human endogenous retrovirus (HERV) segments in intron 9 of C4A or C4B is common in the general population. The HERV-inserted form, often referred to as the long form, contrasts with the short form and is associated with decreased C4 expression in blood 10 . Most importantly, the total copy number of C4 varies widely among individuals, mostly ranging from two to eight copies, consisting of various copy numbers of both C4A and C4B , with and without HERV segments. Therefore, the variance in expression levels of the functionally distinct C4A and C4B isoforms in a general population is attributed mainly to the CNV of C4 , with additional modulation by the HERV insertion 9 . The complex genomic architecture of C4 requires relatively low-throughput, expensive technologies to characterize individuals’ genetic variants. As a result, most C4 -SLE association studies have been conducted with relatively small sample sizes 11 – 17 . Furthermore, due to the lack of an imputation reference panel that simultaneously infer both HLA and C4 variants along with other MHC variants, previous C4 genetic association studies have focused solely on the effects of the C4 CNV on SLE, without considering the compounding effects of HLA variants 11 – 17 . A recent study by Kamitaki et al., reported a primary MHC-SLE association signal at C4 , rather than HLA-DRB1*03:01 using data from large European and African ancestral populations 18 . Although the primary genetic effect could not be distinctly attributed to either C4 CNVs or HLA-DRB1*03:01 in European ancestry due to relatively high LD, a joint logistic regression analysis for C4 CNVs and HLA-DRB1*03:01 in African Americans revealed a sole association of C4 CNVs with SLE, while HLA- DRB1*03:01 showed no significant association. This finding challenges the long-held belief regarding HLA-DRB1 associations with SLE, suggesting that the SLE-MHC association, traditionally attributed to HLA-DRB1 , may instead be driven by CNVs of C4 alleles. However, we urge cautious interpretation of these results for several reasons. First, the study reported very poor imputation accuracy in C4 copy number calls among African populations, potentially impacting the reliability of the findings. Second, the most significant variant was not mapped to C4 in either ancestry. In an unconditional analysis, the study observed a stronger significant association of HLA-DRB1*03:01 compared to the C4 CNV in Europeans. Similarly, another recent European study identified HLA-DQB*02:01 as the most significant variant for SLE rather than C4 CNVs 19 . In Africans, HLA-DRB1*15:03 , known as the most significant classical allele in this population, also showed a stronger association with SLE compared to the C4 CNVs, in their study and others 20 , 21 . Third, adjusting solely for HLA-DRB1*03:01 may not fully capture the multi-allelic nature of HLA-DRB1 associations. Residual effects after conditioning on HLA-DRB1*03:01 are expected to remain substantial around and within HLA-DRB1 . Furthermore, in African ancestry, HLA- DRB*03:01 contributes much less to the risk of SLE. In our present study, to better understand the MHC associations in SLE through a sophisticated, comprehensive genetic analysis using large-scale case-control GWAS data, we have developed an innovative MHC imputation reference panel from whole-genome sequencing (WGS) data by analyzing various MHC variants, including C4 and HLA genes, and phasing them into haplotypes. Unlike existing MHC reference panels, our panel allows for the simultaneous imputation of copy numbers for C4A , C4B , and HERV, as well as amino acid residues and classical alleles of eight major HLA genes, along with neighboring SNPs and indels across the MHC region. The imputation performance of our reference panel has been assessed through various approaches, achieving a high level of accuracy. Applying our imputation reference panel to large SLE GWAS datasets, we uncovered multiple independent genetic contributions from HLA-DRB1 , C4 , and HLA-C . This led us to propose a novel SLE-MHC association model that, for the first time, incorporates HLA missense variants at HLA-DRB1 and HLA-C , and the copy numbers of C4A and HERV. The analysis schematic of this study is illustrated in Fig. 1 . Download figure Open in new tab Fig. 1 Overview of the study design. The schematic diagram illustrates the key analysis steps of the study. a. To construct an MHC imputation reference panel, WGS data from 1,904 samples in two independent batches were used to genotype classical alleles and amino acid residues in eight major HLA genes, CNVs of C4 -related elements, and other MHC SNPs/indels. For C4 , the WGS-based quantification of copy numbers was performed at the diploid level, and the reliability of this quantification method was further validated using ddPCR. The diploid copy numbers of C4A , C4B , and HERV were then segregated into haploid copy numbers through unsupervised haplotype clustering and by determining the most likely haploid copy number of each C4 -related element using a cyclic coordinate descent method, based on the WGS-based diploid copy numbers across the reference individuals. For instance, the example cluster “cluster1” comprises three haplotypes that are likely to possess a single copy of C4A , a single copy of C4B , and two copies of HERV (designated as A1+B1+L2). b. The MHC imputation reference panel was constructed by phasing all binary-coded MHC variants into haplotypes. Its imputation performance was evaluated based on the concordance rates between actual and imputed data from both reference and external samples. c. The newly developed imputation reference panel was applied to two independent SLE GWAS datasets: discovery and replication datasets. Plasma C4 protein levels measured in 1,284 SLE patients were analyzed for correlations with imputed C4 -related copy numbers, revealing a positive correlation with C4 copy number and a negative correlation with HERV copy number. Finally, through stepwise conditional and fine-mapping analyses, three independent signals were identified at HLA-DRB1 , C4 , and HLA-C , which included key variants contributing to the risk of SLE: amino acid positions 11, 13, 26, 37, and 86 in the epitope-binding pocket of HLA-DRB1, position 99 in the epitope-binding pocket of HLA-C, and the copy numbers of C4A and HERV in C4 . Our refined MHC association model was compared with previous models using the replication dataset, evaluating model significance to confirm that the variants identified through our fine-mapping analysis provided the most robust explanation for SLE risk. Results Development of the MHC reference panel for HLA and C4 imputation The MHC reference panel for imputing diverse MHC variants, including HLA and C4 variants, was constructed to include >3,000 haplotypes of the general Korean population, based on WGS data from two independent experimental batches. To achieve this, we genotyped classical alleles of eight major HLA genes ( HLA-A , -B , -C , -DPA1 , -DPB1 , -DQA1 , -DQB1 , and -DRB1 ) at the G-group resolution using a highly accurate graph-based genotyping method, HLA-LA 22 ( Supplementary Fig. 1 ). We further defined these alleles at one-field and two-field resolutions and retrieved their amino acid residue information. In our reference samples, we observed 102 classical alleles at one-field resolution, 267 at two-field resolution, and 289 at G-group resolution, along with 1,440 polymorphic amino acid positions. The diploid copy numbers of C4 , C4A , C4B , and HERV were estimated using Genome STRiP 23 , based on read depth and the allele-specific read ratio between C4A and C4B , as detailed in Methods ( Supplementary Fig. 2 ). We observed a wide range of diploid gene copy numbers of C4 , from 2 to 8 in the reference samples. Similarly, the diploid copy number of the allele-specific isoforms of C4 ranged from 0 to 5 for C4A , and from 0 to 6 for C4B . The HERV segment within C4 was detected with diploid copy numbers ranging from 1 to 7 copies ( Supplementary Fig. 3 ). The diverse combinations among CNVs, alleles, and HERV inserts greatly expanded the genetic variability of C4 ( Fig. 2a ). For instance, the most common combination was to carry four copies of C4 , which included two C4A copies, two C4B copies, and at least two copies of HERV. Download figure Open in new tab Fig. 2 Distribution of diploid copy numbers of C4 . a. The joint distribution of diploid copy numbers of C4A , C4B , and HERV was examined using WGS data for reference samples (n=1,537). The size of each data point is proportional to the frequency of samples with the corresponding combination of C4A and C4B copy numbers. Each data point is a pie chart illustrating the distribution of HERV copy numbers for the given combination of C4A and C4B copy numbers. b. The reliability of the C4 copy number quantification method using WGS data was evaluated by comparing WGS-based results with ddPCR-based results from 127 non-reference samples. The bi-directional histogram shows the distribution of diploid copy numbers of C4 , C4A , C4B , and HERV, as determined by ddPCR (left) and the WGS-based approach (right). Concordance rates between the two methods are displayed in the bar plot on the right end, according to each copy number of C4 genes and related elements (≥ 98.4%). The accuracy of diploid copy number calls for C4 from WGS data was validated by comparing with those derived from digital droplet polymerase chain reaction (ddPCR) technology 24 , 25 across 127 additional Korean samples ( Supplementary Fig. 4 and 5 ). We found highly reliable concordance rates between the WGS-based and ddPCR-based diploid copy numbers (99.2% for C4 , 98.4% for C4A , 99.2% for C4B , and 99.2% for HERV) ( Fig. 2b ). Additionally, the distribution of diploid copy numbers observed in this study aligns with previously reported distributions in other populations 16 , 17 , 19 . To segregate each diploid copy number of C4A , C4B , and HERV into haploid copy numbers, we first defined haplotype clusters based on neighboring SNPs and short indels using unsupervised machine learning, employing various Gaussian mixture variational autoencoder (GMVAE) models 26 ( Fig. 3a-b ). We then conducted the cyclic coordinate descent method to determine the most likely haploid copy numbers in the linkage with each haplotype cluster, aiming to best explain the observed diploid copy numbers across the entire reference sample set ( Supplementary Table 1 and Supplementary Fig. 6 ). The genotypic combination of possible haploid copy numbers in each sample was determined based on the actual diploid copy number and the likelihood of having each haplotype cluster for each homologous chromosome pair in the sample ( Fig. 3a ). The frequency distributions of determined haploid copy numbers for C4A , C4B , and HERV showed single-modal patterns, each centered around one copy, and this pattern was consistently reflected in the copy-number genotype distributions ( Fig. 3c, d ). Stochastically inferred diploid copy numbers from these haplotype frequencies matched the observed diploid copy numbers, indicating that C4 CNVs followed Hardy-Weinberg equilibrium (HWE) ( Supplementary Table 2 ). Download figure Open in new tab Fig. 3 Segregating the diploid copy numbers of C4A , C4B , and HERV into haploid copy numbers. a. Unsupervised haplotype clustering was performed using various GMVAE models based on phased SNPs and indels near C4 in reference samples, estimating the likelihood of each haplotype belonging to each cluster. For each GMVAE model, each haplotype was assigned to the cluster with the highest likelihood. The most likely haploid copy number for each cluster was then determined using a cyclic coordinate descent method, which iteratively minimized the loss function about the difference between inferred and true diploid copy numbers across reference samples. The GMVAE model with the lowest average loss for C4A , C4B , and HERV was selected for subsequent procedures. Individuals’ CNV genotypes, consisting of two haploid copy numbers for C4A , C4B , and HERV, were determined as the most likely haploid copy number combinations, evaluated by multiplying the normalized likelihoods for the two haplotypes belonging to specific clusters, and aligned with the actual diploid copy numbers. b. Haplotype clustering analysis near the C4 region (31.7-32.2Mb on chromosome 6, hg38) identified 47 haplotype clusters. The most common clusters representing 80% of all haplotypes are visualized, along with allocated copy number compositions and frequencies. Each row represents a haplotype, where black and white dots indicate major and minor alleles of bi-allelic variants upstream and downstream of C4 , respectively. The C4 composition is labeled using the initials and copy numbers of C4A (A), C4B (B), and HERV (L; long form). For instance, “A3+B1+L4” denotes the combination of three haploid copies of C4A , a single copy of C4B , and four copies of HERV (meaning all C4 copies are in long form. c. Distribution of haploid copy numbers for C4A , C4B , and HERV in reference samples are displayed on the bar plots. d. The frequencies of CNV genotypes are plotted according to diploid copy numbers of each C4 -related elements. Finally, we assembled the MHC imputation reference panel by incorporating a comprehensive set of 55,632 MHC variants, including HLA variants, C4 -related CNVs, and other SNPs and indels in 1,537 quality control (QC)-passed samples. These variants were phased into haplotypes, with each allele of multi-allelic variants (e.g., HLA classical alleles, HLA amino acid residues) encoded in a binary format to indicate its absence or presence. Similarly, each haploid copy number of C4 -related elements ( C4A , C4B , and HERV) was encoded in the same binary format to denote the presence or absence of each possible copy number. Performance of the new MHC reference panel in imputing HLA and C4 variants We conducted leave-one-out cross-validation, where each reference sample was sequentially excluded, its C4 and HLA variants were masked, and those variants were then imputed using the MHC reference panel with the remaining reference samples. For HLA genes, our panel demonstrated generally excellent concordance rates across all eight classical HLA genes, including HLA-DRB1 and HLA-B , which have previously shown relatively lower imputation accuracy compared to other HLA genes 27 . The average concordance rates were 99.2% at one-field resolution, 97.6% at two-field resolution, and 97.3% at G-group resolution ( Table 1 ). View this table: View inline View popup Download powerpoint Table 1. Concordance rates between imputed and actual data for HLA classical alleles and C4 CNVs from leave-one-out cross validation and external sample validation For C4 -related CNVs, our imputation also achieved highly reliable concordance rates for haploid copy numbers, showing 94.4% for C4A , 94.4% for C4B , and 94.1% for HERV, bringing the overall accuracy to 94.4%. As anticipated, the imputed diploid copy numbers were highly consistent with the observed diploid copy number, yielding an overall accuracy of 90.2% (r 2 =0.83) at the diploid level ( Table 1 ), which was slightly better than a previous C4 imputation approach (e.g., r 2 values of 0.78 and 0.65 for C4A in Europeans and Africans, respectively) 18 . Additionally, the vast majority (95.2%) of discordant results showed only one-copy differences between imputed and observed copy numbers ( Supplementary Tables 3 and 4 ). Application of the MHC reference panel to large-scale SLE GWAS datasets: imputation and accuracy assessment We implemented the new MHC reference panel to simultaneously impute various MHC variants, including HLA and C4 variants, across the MHC region on chromosome 6, using SNP genotypes from two Korean SLE GWAS datasets (discovery dataset: 2,023 SLE patients and 71,125 healthy controls; replication dataset: 348 patients and 4,203 controls; Supplementary Table 5 ). The average imputation quality scores, INFO, were reliable, showing 0.92 for HLA alleles with a minor allele frequency over 0.5% in the discovery dataset and 0.90 in the replication dataset. For C4 -related CNVs, both datasets achieved an average INFO score of 0.81, which surpassed the typical INFO threshold used in GWAS ( Supplementary Fig. 7 ). To assess imputation accuracy, we compared imputed results with directly genotyped results from 127 QC-passed samples, using WGS for HLA classical alleles and ddPCR for C4 -related CNVs. The concordance rates were high and consistent with those observed in leave-one-out cross-validation of the MHC reference panel ( Table 1 ). HLA alleles achieved an average concordance of 95.3% at two-field resolution, and C4 -related CNVs achieved 89.8% at the diploid copy number level. Additionally, we investigated the correlation between imputed diploid copy numbers of C4 -related elements and plasma C4 protein levels in 1,284 patents with SLE, adjusting for covariates including SLE disease activity index (SLEDAI), sex, and the top five genotypic principal components (PCs). Our correlation analysis using imputed CNVs replicated the known effects of C4 -related CNVs on C4 expression, as higher plasma C4 protein levels were strongly associated with increased copy numbers of C4 and decreased copy numbers of HERV inserts ( P =8.8×10 -69 for C4 and P =2.9×10 -23 for HERV; Fig. 4 and Supplementary Table 6 ). This finding reinforces the reliability of our C4 CNV imputation. Notably, we found no evidence that CNVs of C4A and C4B have differential regulatory effects on C4 production ( Supplementary Table 6 ). Download figure Open in new tab Fig. 4 Association of plasma C4 protein levels with imputed C4 and HERV copy numbers in SLE patients. The boxplot displays the relationship between plasma C4 protein levels and imputed C4 copy numbers, further stratified by imputed HERV copy numbers in 1,284 SLE patients. A significant positive correlation is observed between C4 copy numbers and protein levels ( P =8.8×10 -69 ), while HERV copy numbers show a negative correlation ( P =2.9×10 -23 ). One outlier with C4 protein levels over 100mg/dL from a sample with six HERV-inserted C4 copies was omitted from the plot. Primary SLE association at HLA-DRB1 tagged by five amino acids in the epitope-binding groove We performed a stepwise conditional regression analysis to unravel multiple independent SLE association signals across the MHC region (28-34Mb on chromosome 6; hg38) using the discovery GWAS dataset. The genetic associations of HLA classical alleles, HLA amino acid residues, SNPs, indels, and the diploid and haploid copy numbers of C4 -related elements with SLE risk were assessed using logistic regression adjusting for the top five principal components and sex. The multi-allelic effect of each HLA amino acid position was evaluated by a likelihood ratio test (LRT), comparing a full model that included amino acid residues with a nested model that excluded these residues. We identified the primary association signal at HLA-DRB1 , with the most significant association occurring at amino acid position 13, which is occupied by six possible residues ( P =3.16×10 -59 ; Figs. 5a , 6a and 6g ). Position 11, also including six possible residues, demonstrated strong LD with position 13 (r 2 =0.81) and showed a highly significant association ( P =1.58×10 -49 ; Fig. 5a and Fig. 6a, g ). Among the HLA classical alleles, HLA-DRB1*15:01:01G ( P =4.90×10 -50 ; Fig 5a ) exhibited the most significant association with SLE risk, followed by HLA-DQB1*06:02:01G ( P =7.11×10 -48 ; Fig 5a ), both of which are well-established as SLE-risk alleles in East Asian populations. Download figure Open in new tab Fig. 5 Independent SLE association signals at HLA-DRB1 , C4 , and HLA-C . MHC-SLE association results from stepwise conditional regression analyses are shown in mirror regional association plots. The lower section of each mirror plot represents analysis results from the discovery GWAS dataset, while the upper section shows results from combined analyses incorporating both discovery and replication GWAS datasets. Significance levels were calculated using logistic regression for binary-coded variables (e,g., presence of an HLA classical allele or a C4 - related haploid copy number allele) and for diploid copy numbers of C4 -related elements, and using LRTs for multi-allelic variants (e.g., HLA amino acid positions with multiple residues, C4 -related CNVs with multiple haploid copy number alleles). The gray horizontal line indicates the genome-wide significance threshold ( P =5.0×10 -8 ). a. Unconditional analysis identified the strongest association within HLA-DRB1 . Amino acid positions 13 exhibited the highest significance ( P =3.16×10 -59 ; P combined =5.35×10 -68 ) followed by position 11 ( P =1.58×10 -49 ; P combined =1.82×10 -56 ), while HLA- DRB1*15:01:01G was the most significantly associated allele ( P =4.90×10 -50 ; P combined =2.51×10 -52 ). b. After conditioning on all G-group classical alleles of HLA-DRB1 , a secondary signal was uncovered at C4 , with the diploid copy number ( P =1.97×10 -14 ; P combined =1.86×10 -16 ) and haploid copy number ( P =3.34×10 -12 ; P combined =5.34×10 -14 ) of the C4A allele showing the strongest associations. c. Conditioning on all G-group HLA-DRB1 classical alleles and diploid copy numbers of all C4 elements ( C4A , C4B , and HERV) revealed a tertiary signal within HLA-C in the combined analysis, with amino acid position 99 exhibiting the most significant association ( P =1.87×10 -9 ). No tertiary signal was detected in the discovery GWAS dataset alone. d. After conditioning on HLA-DRB1 , C4 , and HLA-C in the combined analysis, no significant variants were identified. Download figure Open in new tab Fig. 6 Key SLE-associated variants within HLA-DRB1 , C4 , and HLA-C . a-f. Stepwise conditional analyses at the amino acid level pinpointed independently SLE-associated positions, with HLA-DRB1 positions identified in the discovery dataset and HLA-C positions identified in the combined discovery and replication datasets. a. Position 13 and its correlated position 11 were identified in HLA-DRB1. b. Conditioning on positions 13 and 11 identified independent signals at position 37 (or 26). c. Position 86 was identified when conditioned on positions 13, 11, 37, and 26. d. No significant positions remained in HLA-DRB1 after conditioning on all five identified positions ( P <5.0×10 -8 ). e. Conditioning on HLA-DRB1 and C4 identified HLA-C amino acid position 99. f. No additional HLA-C positions were significant after conditioning on position 99 along with HLA-DRB1 and C4 . g-i. ORs (dots) with 95% CIs (error bars) for amino acid residues, observed in stepwise conditional analysis, are plotted for HLA-DRB1 positions 13 and 11 (g) , positions 37 and 26 (h) , and position 86 (i) , across datasets: discovery (n=73,148, red), replication (n=4,551, blue), and combined (n=77,699, green). j. The five identified amino acid positions are located on the epitope-binding groove of the HLA-DR beta chain. k. Per-copy ORs with 95% CIs for C4A and HERV are plotted across datasets. Effect estimates for C4A were adjusted for HLA-DRB1 and other covariates, while those for HERV were adjusted for HLA-DRB1 , C4A , and the same set of covariates. l. ORs with 95% CIs for residues at HLA-C position 99, conditioned on HLA-DRB1 and C4 , are plotted across datasets. m. Amino acid position 99 is located on the epitope-binding groove of HLA-C. n. P LRT values from the replication dataset for previously proposed and our MHC-SLE association models are plotted. To address potential overfitting, Model 9 was derived from the discovery dataset, excluding the HLA-C signal. Model 10 represents the full association model of this study, including the HLA-C signal. Our MHC-SLE association models, containing the identified signals, achieved the highest significance whether or not the HLA-C signal was included ( P Model_9 =5.50×10 -11 ; P Model_10 =1.91×10 -12 ). AA, amino acid; DF, degree of freedom. Given the complex nature of long-range haplotypes, which may include key amino acid combinations involving residues at positions 13 and 11 that tag specific HLA-DRB1 classical alleles and drive the SLE association of HLA-DRB1 , we further investigated additional independent associations within HLA-DRB1 at the amino acid level. After accounting for the effects of the tightly correlated amino acid positions 13 and 11, we identified significant associations at amino acid positions 37 ( P =1.54×10 -15 ) and 26 ( P =8.36×10 -14 ), which showed moderate residue correlation (r 2 =0.42) ( Fig. 6b, h ). These positions provided similar model fits for SLE association ( P LRT =1.59×10 -71 for positions 11-13-37 and P LRT =8.65×10 -70 for positions 11-13-26), which has already been reported in previous studies 5 , 6 . Considering the uncertainty of the secondary SLE-risk amino acid position, we accounted for both positions 26 and 37 in the next round of stepwise conditional regression. After conditioning on the effects of positions 11, 13, 26, and 37, we newly identified amino acid position 86 ( P =4.93×10 -10 ; Fig. 6c, i ). Adding or excluding position 26 in the conditional regression model did not change the statistical significance of position 86. No additional significant positions were detected after conditioning on the amino acid position 11, 13, 26, 37, and 86 ( Fig. 6d ). The associations at these positions were further examined in the replication GWAS dataset, revealing highly consistent residue-specific effect sizes across both the discovery and replication GWAS datasets ( Fig. 6g-i ) and increased significance in the combined analysis of both datasets ( Supplementary Data 1 ). Moreover, the three signals derived from the five identified amino acid positions exhibited the best SLE association model fit among all possible combinations of HLA-DRB1 amino acid positions in LRTs, accounting for the differences in the number of residues as indicated by the degree of freedom ( Supplementary Fig. 8 ). These positions are located in close spatial proximity within the epitope-binding pocket of HLA-DR molecules ( Fig. 6j ). Based on these findings, amino acid haplotypes of five key positions were tested for SLE associations ( Table 2 ). Sixteen haplotypes with frequency > 0.5% were identified, including five SLE- risk and seven SLE-protective haplotypes ( P combined < 0.05). The amino acid haplotype model effectively explained associations of various classical alleles, such as HLA-DRB1*15:01 with the PRSFV haplotype. Furthermore, HLA-DRB1*03:01 (SSYNV) and HLA-DRB1*15:03 (PRSFV), known to be specific to Europeans and Africans, respectively, were identified as having SLE-risk amino acid haplotypes, even though these classical alleles were present at low or absent frequencies in in the Korean population. Similarly, HLA-DRB1*13:02 and HLA-DRB1*14:03 , previously reported as significant SLE-protective alleles 28 , were found to have the SLE-protective haplotype SSPNG. Taken together, this suggests that amino acid haplotypes not only provide a parsimonious disease association model compared to the conventional model using classical alleles but also demonstrate a highly transferable model across different ancestries. View this table: View inline View popup Download powerpoint Table 2. Association of HLA amino acid haplotypes with susceptibility to SLE Dissecting the secondary SLE association revealed that decreased C4A and increased HERV copy numbers confer the risk of SLE, independent of HLA-DRB1 To determine if there were additional independent MHC associations beyond HLA-DRB1 , we conducted a conditional analysis that adjusted for all HLA-DRB1 G-group alleles with frequencies greater than 0.5%, thereby eliminating the influence of HLA-DRB1 on the SLE associations. Using the discovery GWAS dataset, we identified a secondary signal within the C4 and observed moderate correlation with HLA-DRB1 , which was adequately controlled for in the logistic regression model ( Fig. 5b and Supplementary Table 7 ). We found that the diploid copy numbers of C4A showed the most significant association in the conditional analysis ( P =1.97×10 -14 ), which was much stronger than the copy numbers of C4B ( P =1.09×10 -7 ) and HERV ( P =1.85×10 -2 ), as well as the total gene copy number of C4 ( P =5.37×10 -2 ) ( Fig. 5b ). The lower copy number of C4A was associated with a higher risk of SLE, with a per-copy odds ratio (OR) of 0.70 with a 95% confidence interval (CI) from 0.63 to 0.76 ( Fig. 6k ). This association pattern was consistently observed at the haploid-level, with C4A showing the strongest significance among the C4 -related CNVs ( P =3.34×10 -12 ; Fig. 5b ). Compared with one haploid copy of C4A as the reference, the deficiency of C4A increased SLE risk (OR=1.36, 95% CI=1.18—1.57, P =2.38×10 -5 ), while the gain of C4A copy showed a protective effect (OR=0.75, 95% CI=0.67—0.83, P =8.87×10 -8 ). As previously known, the haploid-level copy numbers of C4A , C4B and HERV are in LD with each other ( Supplementary Table 1 ). To understand the overall effect of C4 -related CNVs and estimate the mutually independent SLE-risk effects of each C4 -related CNV, we conducted a joint conditional regression analysis using the diploid copy number of C4A , C4B , and HERV, adjusting for HLA-DRB1 and genetic background ( Table 3 ). This joint model revealed a significant contribution of C4 -related CNVs to SLE susceptibility ( P LRT =1.80×10 -15 ), exhibiting significant, independent associations of C4A ( P =4.12×10 -11 , per-copy OR=0.69, 95% CI= 0.62 –0.77) and HERV ( P =1.13×10 -2 , per-copy OR=1.10, 95% CI= 1.02 –1.19), but not C4B ( P =0.24). In contrast to the effects of the C4A CNV, the copy number of HERV was positively associated with the risk of SLE ( Fig. 6k ). However, the known biological effects of both decreased C4A copy number and increased HERV copy number are the same, collectively resulting in decreased C4A expression, as supported by our correlation analysis between plasma C4 protein and the copy number of C4 and HERV in our study cohort ( Fig. 4c ). View this table: View inline View popup Download powerpoint Table 3. Joint effects of C4 -related CNVs on SLE risk, adjusted for HLA-DRB1 effects All C4 -related association signals involving C4A and HERV were robustly validated in the replication GWAS dataset, and demonstrated increased significance in combined analyses incorporating both discovery and replication GWAS datasets ( Fig. 5b , 6k , Table 3 , and Supplementary Data 1 ). We also investigated previously reported sex bias in C4 -SLE associations using a logistic regression model with an interaction term between sex and C4 -related CNVs, but found no evidence of sex-biased effect sizes on SLE ( Supplementary Table 8 ). Indeed, plasma C4 protein levels normalized by the C4 copy number were consistent between males and females in 1,284 SLE patients ( Supplementary Fig. 9 and Supplementary Table 6 ). Tertiary association signals at HLA-C amino acid position 99 After conditioning on HLA-DRB1 and C4 , the tertiary signal was identified within HLA-C in the combined analysis but not in the discovery dataset alone, with the most significant association at amino acid position 99 ( P =5.39 ×10 -7 ; P combined =1.87×10 -9 ; Figs. 5c , 6e , 6l and Supplementary Data 1 ). Position 99 is highly polymorphic, with four possible residues, and its effect sizes were consistent across the discovery and replication datasets ( Fig. 6l ). Located in the center of the peptide-binding groove of HLA-C ( Fig. 6m ), this position has been associated with the risk of other immune-mediated disorders 29 , 30 . Among the four residues at position 99, serine was associated with an increased risk of SLE, while cysteine was protective ( Table 2 ). Serine at position 99 primarily tags HLA-C*07:02 in Koreans, which belongs to a previously reported long-range SLE-risk haplotype in Europeans, though its independent effect has not been investigated 20 , 31 , 32 . Cysteine at position 99 is mostly found in HLA- C*01:02 , which has been associated with other diseases but not with SLE 33 – 35 . SLE-MHC association model comparison in the Korean population We then focused on evaluating the fit of our SLE-MHC association model to assess its performance relative to previously proposed SLE-MHC association models 5 , 6 , 18 , 19 . To avoid overfitting, our association model included only the variants identified from the discovery dataset, comprising amino acid positions 11, 13, 26, 37, and 86 in HLA-DRB1, along with the diploid copy numbers of C4A and HERV, while excluding HLA-C amino acid position 99 identified in a combined analysis. The model fit was subsequently evaluated in the replication GWAS dataset. Our model demonstrated a strong association with a P LRT value of 5.50×10 -11 in the replication dataset ( Fig. 6n ), outperforming all previous SLE-MHC association models, as indicated by P LRT values that accounted for differences in the number of variables (or degrees of freedom). Notably, some previous models included European-specific alleles, such as HLA-DRB1*03:01 or HLA- DQB1*02:01 . To account for this, we re-evaluated these models by substituting the European-specific alleles with the East Asian-specific SLE-risk allele HLA-DRB1*15:01 or its strong proxy allele HLA-DQB1*06:02 (r 2 =0.94 in Koreans). Although this substitution improved model performance, the association significance level remained weaker than that of our comprehensive MHC model. These findings emphasize the importance of considering the collective effects of multiple HLA-DRB1 classical alleles (tagged by missense variants) as well as the diploid copy numbers of C4A and HERV. Additionally, when we incorporated the tertiary association at amino acid position 99 of HLA-C, identified only through the combined datasets, the fit of our association model improved, as expected ( P =1.91×10 -12 ; Fig. 6n ). Discussion This study delved into the genetic associations within the MHC region using two independent Korean SLE case-control datasets, leveraging a newly developed MHC imputation reference panel. Our findings underscore the significant, independent associations of HLA-DRB1 , C4 , and HLA-C with the risk of SLE. The MHC-SLE association model we proposed outperformed other models comprising variants reported in previous MHC fine-mapping studies. Model comparisons were conducted on an independent replication dataset based on LRTs that accounted for differences in degrees of freedom, ensuring fair and robust comparisons. The associations of HLA-DRB1 and HLA-C were predominantly observed at specific amino acid positions rather than HLA classical alleles, emphasizing the disease relevance of the missense variants that shape the surface of epitope-binding grooves of HLA-DRB1 and HLA-C in SLE. Amino acid positions 11, 13, 26, and 37 of HLA-DRB1 have been consistently highlighted in previous HLA fine-mapping studies 5 , 6 , and we have now identified an additional position, 86, which collectively accounts for most of the genetic contribution of HLA-DRB1 . Position 86 has been linked to the risk of type 1 diabetes 36 . These five amino acid positions of HLA-DRB1, representing three independent association signals, provided the simplest and ancestrally transferable explanation, supporting the associations of various classical alleles. Notably, the associations of HLA-DRB1 amino acid haplotypes well align with the ancestry-specific associations of HLA-DRB1 (e.g., HLA-DRB1*03:01 in Europeans and HLA-DRB1*15:03 in Africans). Importantly, this is the first study to investigate the disease effects of C4 CNVs while adjusting for the multiallelic effects of HLA-DRB1 . There has been debate regarding whether HLA-DRB1 or C4 is the primary driver of SLE in the MHC region, and whether the associations of SLE with HLA-DRB1 and C4 , which are in LD, are independent of each other 37 . Our conditional association model, which controlled for the influence of HLA-DRB1 on SLE risk, identified the most significant association at C4 , revealing that a lower copy number of C4A and, for the first time, a higher copy number of HERV were independently associated with an increased risk of SLE. Plasma C4 analysis indicated that C4 proteins levels are predominantly influenced by the total copy number of C4 , with no evidence of allele-specific regulatory effects between C4A and C4B . Additionally, a HERV element, inserted within C4 , leads to down-regulation of plasma C4 proteins. Considering both the genetic associations with SLE and protein levels, these findings suggest that the reduced production of the C4A protein isoform, functionally distinct from the C4B isoform, is driven by both low C4A copy numbers and HERV- mediated attenuation. This reduction plays a crucial role in SLE pathogenesis, particularly through C4A -specific pathways such as the regulation of autoreactive B cells. A recent study using a mouse model showed that the C4A isoform is more efficient than C4B at binding and removing apoptotic debris, enhancing apoptotic clearance, and reducing autoantigen exposure, which limits the activation of autoreactive B cells 38 . It also found that C4A regulates germinal centers by restricting the expansion of these B cells 38 . These findings highlight the role of C4A for maintaining immune tolerance and preventing pathogenic autoimmunity in SLE. Our fine-mapping study was particularly successful by leveraging the relatively less extensive LD in the Korean population compared to Europeans, along with the use of a high-accuracy MHC imputation reference panel. Previous studies have demonstrated that European populations exhibits larger and more extended LD within the MHC region compared to African and Asian populations, likely due to population-specific factors such as lower recombination rates, resulting in slower decay of LD in Europeans 18 , 19 , 39 – 42 . Consistently, European SLE-MHC association analyses have reported long-range SLE-risk haplotypes involving multiple HLA genes in extensive LD, spanning MHC class I and II regions, such as the C*07:01 – B*08:01 – DRB1*03:01 – DQB1*02:01 haplotype 19 , 31 . In contrast, Asian and African populations have shown less extensive LD patterns, with relatively more predominant SLE associations in the MHC class II region 20 , 21 . Indeed, recent analyses by Kamitaki et al. , whose study is detailed in the Introduction section, found that the significance of C4 CNV effects was comparable to that of well-known risk classical alleles in Europeans 18 , 19 . However, the strong correlations between C4 and HLA in Europeans posed challenges in distinguishing the independent effect of C4 , especially considering the strong correlation between the SLE-risk C4 haplotype (single-copy, short-form C4B ) and SLE-risk HLA alleles, including HLA-DRB1*03:01 , HLA-DRB1*08:01 , and HLA-DQB1*02:01 (r 2 >0.7) 18 , 19 , 39 . Conversely, findings from African populations revealed weaker correlations between C4 and HLA variants (maximum r 2 between C4 haplotypes and HLA alleles = 0.31), making it easier to pinpoint the independent association signal of C4 . However, in their study, the imputation accuracy for C4 in Africans was quite poor, and their conditional model controlling only for HLA-DRB1*03:01 does not fully support their conclusion that the primary signal driving the risk of SLE is derived from C4 18 . In Asian populations, the particularly strong association between HLA-DRB1 and SLE risk has been well-established 5 , 6 . In our current study, the association of C4 was initially overshadowed by the predominant effect of HLA-DRB1 in the unconditional analysis ( P combined of C4A =1.48×10 -6 ). However, correlations between HLA-DRB1 classical alleles and C4 -related haploid copy numbers were moderated (maximum r 2 between C4 -related haploid copy numbers and HLA alleles = 0.39; Supplementary Table 7 ), our conditional analysis enabled us to pinpoint the secondary association signal at C4 . We developed the first MHC imputation reference panel capable of simultaneously imputing various HLA variants, C4 -related CNVs, and other MHC variants. The C4 gene, despite its biological relevance to autoimmune diseases like SLE, has been relatively underexplored compared to HLA genes, largely due to technically complex and low-throughput methods required to characterize its genetic structure. In the present study, we leveraged WGS data from large-scale general Korean cohorts to characterize C4 alleles and CNVs using an existing reliable computational approach 18 , 23 . One of the major challenges was accurately defining haploid-level copy numbers from WGS- derived diploid copy numbers. To address this, we employed haplotype clustering based on SNPs and short indels around the C4 gene, utilizing deep learning-based models and an iterative optimization method to determine the most likely copy numbers of C4A , C4B , and HERV for each haplotype cluster. This method not only allowed us to construct an MHC imputation reference panel with high accuracy for C4-related CNVs, but it also holds significant potential for broader applications in other studies dealing with complex CNV genotyping. Nonetheless, Our imputation panel has a limitation that it only includes copy number information for C4A , C4B , and HERV in each homologous chromosome pair, leading to ambiguity in determining how many copies of C4A and C4B are in the long form with the HERV insert. This ambiguity is particularly pronounced in haplotypes where both C4A and C4B alleles are present but the number of HERV copies is fewer than the combined total of C4A and C4B . To validate the imputation performance of our reference panel, we rigorously evaluated the imputation accuracy through various methods, including cross-validation with internal and external samples, that compares imputed data with actual data from WGS and ddPCR technologies. These evaluations consistently demonstrated superior imputation accuracy for both HLA variants and C4 - related CNVs. We have made this imputation panel publicly available via an imputation server ( https://coda.nih.go.kr/usab/kis/intro.do ), providing other researchers with a valuable tool to better understand causal variants in the MHC region, especially for inflammatory diseases in East Asian populations. In summary, we have demonstrated an enhanced SLE-MHC association model from the Korean population through the development of a novel MHC imputation reference panel. This advancement not only enriches our understanding of the biological mechanisms underlying the role of HLA and C4 in SLE pathogenesis but also holds promise for refining genetic risk assessments for SLE. Furthermore, the application of our MHC panel is poised to significantly advance the investigation of MHC associations in other immune-mediated disorders within East Asian populations, paving the way for further insights and progress in personalized medicine. Methods Cohorts for constructing the MHC imputation reference panel and fine-mapping MHC-SLE associations Two independent WGS batches from 1,904 unrelated healthy Korean individuals were utilized for constructing the MHC imputation reference panel. These batches were generated by the Ulsan National Institute of Science and Technology (UNIST) (reference batch #1; n=904; published data 43 ) and the Korea Reference Genome (KRG) project, led by the National Institute of Health, Republic of Korea (KNIH) (reference batch #2; n=1,000; unpublished data). The KRG project, which aimed at advancing Korean genome research and identifying a wide spectrum of variants in the Korean population 44 , generated WGS data from around 8,000 individuals. From this dataset, sequencing data from 1,000 randomly selected participants was used to construct the MHC imputation reference panel. All participants provided written informed consent. Genomic DNA samples from the participants was sequenced using the Illumina NovaSeq platform (average depth ∼30x). Raw sequencing data was trimmed using the Trimmomatic software (version 0.39) 45 . All analyses using the WGS data were conducted under the regulations approved by Institutional Review Board (IRBs) of the UNIST (UNISTIRB-15-19-A, UNISTIRB-16-13-C) and the Korean Disease Control and Prevention Agency (2022-02-04-P-A, KDCA-2024-02-09-R-01). To characterize MHC associations in SLE, two Korean SLE case-control datasets (discovery and replication GWAS datasets) were used ( Supplementary Table 5 ). These datasets were expanded from our recent GWAS datasets 46 by adding 146 cases 38,305 controls to the previous discovery dataset. Patients with SLE were recruited from Hanyang University Hospital for Rheumatic Diseases and fulfilled either the revised American College of Rheumatology classification criteria or the Systemic Lupus International Collaborating Clinics classification criteria for SLE 47 , 48 . Healthy controls were recruited from the Korea National Institute of Health. All participants provided written informed consent, and the study protocol was approved by the IRBs at participating institutions (HY-12-005-16, HY-16-07-20, HY-16-129-15). Calling SNPs and indels in the MHC region using WGS data Illumina high-coverage whole-genome 101 bp paired-end sequencing data, with a mean read depth of 29.0 in reference batch #1 and 37.5 in batch #2, were aligned to the human genome reference assembly (GRCh38) using BWA 49 . SNP and indels were called using the Genome Analysis Toolkit (GATK) pipeline 50 . The variants called from the WGS data were filtered for each dataset to retain biallelic variants that met the following QC thresholds: minimum depth of 15 (--minDP) and minimum genotype quality of 20 (--minGQ), a variant quality score at least 20 (--minQ), minor allele frequency over 0.5%, call rate over 90%, and conformity to HWE ( P > 1×10 -7 ). After filtering, a total of 52,675 QC’ed MHC variants (25-35 Mb on chromosome 6; hg38) were retained across both datasets. To further ensure data reliability, we removed variants that showed a batch effect in genotype frequency distribution between the two datasets using Fisher’s exact test ( P < 0.001), leaving 52,602 variants for further analysis. HLA typing at G-group resolution using WGS data We determined the classical alleles of eight major HLA genes ( HLA-A, -B, -C, -DPA1, -DPB1, -DQA1, -DQB1, -DRB1 ) at the G-group resolution using HLA-LA 22 . To ensure high-quality data in the MHC imputation reference panel, we retained reference samples that had an average coverage of over 15× for each gene, resulting in 846 samples from reference batch #1 and 1,000 samples from batch #2 ( Supplementary Fig. 1 ). We excluded 12 samples with HLA-DPA1*01:11 , which were found exclusively in batch #1. We converted these G-group HLA classical alleles into one- and two-field HLA alleles using the NomenCleaner module in HLA-TAPAS 51 . Additionally, HLA amino acid residues and exonic alleles were inferred based on the two-field HLA classical alleles. Quantifying diploid copy numbers of C4 , C4A , C4B , and HERV using WGS data We utilized a previously developed method 18 implemented in Genome STRiP 23 to estimate the diploid copy numbers of C4 , C4A , C4B and HERV by analyzing the read depth across their genic regions and the relative ratio of allele-specific read counts for C4A and C4B . To ensure that the MHC imputation reference include only highly reliable CNV data, we filtered out samples with a Phred-scale quality score below 20 for any of copy number calls of C4 -related elements ( Supplementary Fig. 2 ). As a result, 1,687 reference samples were retained for the construction of the reference panel, comprising 759 samples from reference batch #1 and 928 samples from batch #2 ( Supplementary Fig. 3 ). Determination of copy number alleles of C4A , C4B , and HERV from diploid copy numbers To accurately determine the CNV genotype (an allelic pair of haploid copy numbers of C4 -related CNVs) from a given diploid copy number, we first surveyed possible haploid-level copy numbers for C4A , C4B , and HERV. These were inferred from observed haploid copy numbers in potential genotypes that met a genotype likelihood threshold over 0.01. The genotype likelihood was estimated using the ‘GenerateHaploidCNVGenotypes’ utility in Genome STRiP 23 , which applies a frequency-based expectation-maximization algorithm to estimate the likelihood of a pair of haploid copy numbers under HWE. This analysis showed that the haploid copy numbers could range from 0 to 3 for C4A and C4B , and from 0 to 4 for HERV. We then carried out haplotype clustering by examining haplotypes of SNPs and indels around the C4 region using the HaploNet framework 26 , which uses an unsupervised neural network approach to compress haplotype data into a lower-dimensional space and indirectly cluster haplotypes with a GMVAE model. Input haplotypes were phased with Eagle v2.4.1 52 and included 601 biallelic variants in the C4 region (31.7-32.2 Mb on chromosome 6; hg38) pruned at an r 2 threshold of 0.8 using PLINK 53 . GMVAE models were configured with varying combinations of three parameters: the number of clusters (ranging from 2 to 100), the number of hidden layer nodes (64, 128, 256, 300, and 512), and the number of latent variables (16, 32, 64, and 128; always less than the given number of hidden layer nodes). Each model generated the likelihoods for each haplotype belonging to each identified cluster. For each model, we assigned the haplotypes to their most likely cluster based on the estimated cluster likelihood. Clusters that did not exhibit maximum likelihood for any haplotype were excluded from subsequent analysis steps. Next, we employed the cyclic coordinate descent method to assign linked haploid copy number of C4 -related CNVs to each cluster. We initialized a cluster-specific copy number vector, , with the most common haploid copy numbers (one for C4A and C4B , and either one or two for HERV) for all n clusters. We iteratively updated these cluster-specific copy numbers by cycling through them one at a time, minimizing a multivariable loss function f ( c ) with respect to each cluster-specific copy number while holding the other variables constant. Thus, the copy number of i -th cluster at the ( k + 1)-th iteration was determined by solving the single variable optimization problem, as follows: where f ( c ) is the 0-1 loss function that measures how many times the predicted diploid copy numbers (=sums of inferred haploid copy numbers per individual) do not match the observed diploid copy number across reference individuals. The iterative optimization continued to update each element of c until convergence was achieved, ensuring that the cluster-specific copy numbers best fit the observed diploid copy number data. This analysis was performed separately for C4A , C4B , and HERV across all GMVAE models, and the model that yielded the best performance was configured with 48 clusters, 64 latent variables, and 300 hidden layer nodes ( Supplementary Fig. 6 ). Based on the cluster-wise likelihoods estimated by this model for each haplotype, we determined the CNV genotype of C4A , C4B , and HERV for each reference individual as the most likely pair of haploid copy numbers that matched the WGS-based diploid copy numbers, and used these CNV genotypes in subsequent analyses. A schematic of the genotyping process is illustrated in Fig. 3a . Construction of a Korean MHC imputation reference panel We integrated a wide range of genetic data, including haploid copy numbers of C4A , C4B , and HERV; HLA classical alleles at one-field, two-field, and G-group resolution; HLA amino acid residues; and other SNPs and indels in the MHC region to assemble an MHC imputation reference panel. All genetic information was encoded in a binary format. Biallelic variants, such as SNPs, indels, and biallelic amino acid positions, were directly represented by their allele names. For multiallelic variants, including HLA classical alleles, multiallelic amino acid positions, and C4 -related CNVs, we used a binary encoding scheme where each allele or haploid copy number was indicated as its absence (A) or presence (T). For instance, the C4A CNV, with possible haploid copy numbers of 0, 1, 2, and 3, was represented by four distinct copy number variables named as C4A_copy0, C4A_copy1, C4A_copy2, and C4A_copy3, respectively. If a sample had 2 and 3 haploid copies, these four variables were encoded as A / A , A / A , A / T (or T/A ), and A / T (or T/A ), respectively. The merged genetic information was phased into haplotypes using Eagle v.2.4.1 52 , including 55,632 biallelic variables in 1,537 samples. Evaluation of imputation accuracy of the MHC imputation reference panel To evaluate the imputation accuracy of our MHC imputation reference panel, we employed a leave-one-out cross-validation strategy using reference samples. For each reference sample, HLA classical alleles and C4 -related CNVs were imputed based on the genetic data of SNPs and indels using IMPUTE5 54 , with a reference panel composed of all other reference samples. Concordance rates between imputed and genotyped data across all samples were calculated to assess imputation performance. Additionally, we validated the imputation accuracy using external non-reference samples, randomly selected from the discovery GWAS dataset. High-coverage WGS data were generated for 144 samples to characterize HLA variants and diploid copy numbers of C4 -related CNVs, using the same analysis and QC methods as for the reference samples. After applying the same filtering criteria, 127 samples with highly reliable calls for HLA and C4 variants remained for subsequent evaluation of our MHC imputation reference panel ( Supplementary Fig. 1 and Supplementary Fig. 4 ). The diploid copy number of C4 -related elements were further examined using ddPCR ( Supplementary Fig. 5 ), as previously described 24 . C4 and HLA variants were imputed based on GWAS-array SNPs listed in the discovery GWAS dataset using IMPUTE5, and these imputed results were compared with those experimentally confirmed by ddPCR (for C4 ) and WGS (for HLA). Additionally, the accuracy of WGS- based Genome STRiP-estimated diploid copy numbers of C4 -related elements was assessed by comparison with ddPCR results ( Fig. 2b ). SLE GWAS datasets and MHC imputation For the association analysis of SLE, we used two independent genotype datasets from Korean populations, comprising 2,023 cases and 71,125 controls generating using Axiom Korean Chip 55 (discovery GWAS dataset), and 348 cases and 4,203 controls generated using Illumina Omni BeadChip (replication GWAS dataset) after sample-level QC ( Supplementary Table 5 ). In the discovery dataset, SNPs were filtered using a genotype call rate threshold of 0.05, a minor allele frequency cutoff of 0.005, and HWE threshold of 1×10 -6 in cases and 1×10 - 10 in controls. Additionally, sample-level call rates, heterozygosity, population stratification, cryptic relatedness, and sex consistency were assessed for QC at the sample level. Detailed description of the replication dataset including the stringent QC criteria applied, have been provided elsewhere 5 . We extracted SNPs located in the extended MHC region (25-35 Mb on chromosome 6; hg38) from both datasets as for imputation inputs. The imputation was performed for each case-control dataset separately, using IMPUTE5 54 and our MHC imputation reference panel. Calculation of correlation between HLA amino acid positions As an approximate LD index, we estimated “lazy” r 2 values to represent the correlation between two multi-residue amino acid positions within a single HLA gene. These estimates were derived by averaging the highest r 2 values, where each maximum was the largest correlation between any residue at one position and all residues at the other position. Disease association analysis The SLE association of each binary-encoded MHC variable (including each HLA classical allele, amino acid residue, C4 -related haploid copy number, SNP, and indel, with an allele frequency > 0.5%) was assessed using logistic regression, incorporating imputed allelic dosage and covariates such as the top five PCs and sex. For the combined analysis using both datasets, a dataset-indicating dummy variable was additionally included as a covariate. To investigate the association of diploid copy numbers of C4 -related elements, we estimated the diploid copy numbers (D) of C4A , C4B and HERV from the imputed dosages of haploid copy number variables using the formula where . Here, d c represents the dosage of the “ T ” allele, indicating the presence of c haploid copies in an individual and is normalized by k/2 , which is half the sum of the dosages of the “ T ” alleles for all possible haploid copies (ranging from 0 copy to n copies; n =3 for C4A and C4B , and n =4 for HERV). The total imputed copy number of C4 is calculated as the sum of the diploid copy numbers of C4A and C4B . To evaluate the multi-allelic effect of each amino acid position or C4 -related elements on SLE risk, an LRT was conducted to compare a null logistic regression model including only covariates with a full model that additionally included multi-allelic variables. For each multi-residue amino acid position, the full model incorporated the dosages of amino acid residues and the same covariates, excluding the most common residue and any residues with a frequency below 0.5%. For C4 , the full model included the diploid or haploid copy numbers of multiple C4 -related elements ( C4A , C4B , and/or HERV) along with the same covariates, and also included the dosages of G-group HLA-DRB1 classical alleles, except the most common classical allele, as additional covariates to adjust for the multi-allelic effect of HLA-DRB1 . After the most significant variant in the MHC region was identified at a gene, a stepwise conditional regression analysis was performed, where each of biallelic and multiallelic variants in the MHC region was tested for the independent association with SLE risk, adjusting for the overall multi-allelic effects of the identified gene(s), as described previously 56 . This process was repeated until no additional variants reached the genome-wide significance threshold of 5×10 -8 . To account for the overall multi-allelic effects of HLA or C4 , the dosages of G-group HLA classical alleles (excluding the most common allele and any alleles with a frequency below 0.5%) or the imputed diploid copy number of C4A , C4B , and HERV were added as covariates in the conditional regression models, respectively. Correlation analysis of plasma C4 protein level and C4 copy numbers In a subset of patients with SLE from the discovery GWAS dataset (n=1,284), plasma C4 concentrations were measured using immunoturbidimetric method with a normal reference range of 10-40 mg/dL, and disease activity was assessed with the SLE Disease Activity Index 2000 (SLEDAI). Linear regression analysis was performed to investigate the linear relationship between plasma C4 levels and copy numbers of C4 -related elements, adjusting for disease activity, sex, and the top five PCs. Data Availability All data except SLE GWAS datasets produced in the present study are available upon reasonable request to the authors. This individual-level genotype and phenotype data are protected and are not available due to data privacy laws. Data availability Raw WGS data from reference batch #1 will be available as possible upon request and after approval from the Korean Genomics Center’s review board in UNIST. Raw WGS data from reference batch #2 will be available as possible upon request and after approval from KNIH. The SLE GWAS datasets were collected by Hanyang University. This individual-level genotype and phenotype data are protected and are not available due to data privacy laws. Code availability The code of all tools used for analyses in this paper is publicly available and has been presented in the Methods section. Custom codes used to analysis are available at https://github.com/KimLabKHU/MHC-_reference_panel . Authors’ contributions SCB and KK designed the study. SJ and JB contributed to the collection and processing of WGS data for reference batch #1. DMS, SMK, YJK, and BJK contributed to the collection and processing of WGS data for reference batch #2. CYY and KK conducted all computational analyses with assistances from DMS, SMK, and SC. YTL and SJY conducted ddPCR experiments to quantify diploid copy numbers of C4 elements. SYB, HSL, and SCB characterized patients with SLE and collected clinical data. SYB, HSL, YJK, BJK, SCB, and KK generated GWAS data. CYY, DMS, XY, YC, XZ, YJK, BJK, SCB, and KK interpreted the results. CYY and KK wrote the manuscript. All authors reviewed and approved the manuscript. Competing interests SJ is the CEO of both AgingLab and Geromics and an employee of Clinomics. JB is the founder of AgingLab. The other authors declare that they have no conflicts of interest. Acknowledgements We are grateful to all collaborators and study subjects for their participation in this study. This work was supported by the National Research Foundation (NRF) of Korea [NRF-2021R1A6A1A03038899, NRF-2022R1A2C2092164], the National Facility & Equipment Center [2023R1A6C101A009], the Hanyang University Institute for Rheumatology Research, the U-K BRAND Research Fund [1.200108.01] of UNIST, the Research Project Funded by Ulsan City Research Fund [1.200047.01] of UNIST, and the Ministry of SMEs and Startups (MSS, Korea) [P0016195, P0016193, P0016191; 1425156792, 1425157301, 1425157253; 2.220035.01, 2.220036.01, 2.220037.01]. Further funding was provided by the Korea Planning & Evaluation Institute of Industrial Technology with support from the Ministry of Trade, Industry and Energy in 2024 [RS-2024-00435468, Development and Dissemination of National Standard Technology]. References 1. ↵ Hoi , A. , Igel , T. , Mok , C.C. & Arnaud , L . Systemic lupus erythematosus . Lancet 403 , 2326 – 2338 ( 2024 ). OpenUrl PubMed 2. ↵ Guga , S. , Wang , Y. , Graham , D.C. & Vyse , T.J . A review of genetic risk in systemic lupus erythematosus . Expert Rev Clin Immunol 19 , 1247 – 1258 ( 2023 ). OpenUrl PubMed 3. ↵ Ha , E. , Bae , S.C. & Kim , K . Recent advances in understanding the genetic basis of systemic lupus erythematosus . Semin Immunopathol 44 , 29 – 46 ( 2022 ). OpenUrl CrossRef 4. ↵ Trowsdale , J. & Knight , J.C . Major histocompatibility complex genomics and human disease . Annu Rev Genomics Hum Genet 14 , 301 – 23 ( 2013 ). OpenUrl CrossRef PubMed Web of Science 5. ↵ Kim , K. et al. The HLA-DRbeta1 amino acid positions 11-13-26 explain the majority of SLE- MHC associations . Nat Commun 5 , 5902 ( 2014 ). OpenUrl CrossRef PubMed 6. ↵ Molineros , J.E. et al. Amino acid signatures of HLA Class-I and II molecules are strongly associated with SLE susceptibility and autoantibody production in Eastern Asians . PLoS Genet 15 , e1008092 ( 2019 ). OpenUrl PubMed 7. ↵ Sun , C. et al. High-density genotyping of immune-related loci identifies new SLE risk variants in individuals with Asian ancestry . Nat Genet 48 , 323 – 30 ( 2016 ). OpenUrl CrossRef PubMed 8. ↵ Goicoechea de Jorge , E., et al. Common and rare genetic variants of complement components in human disease . Mol Immunol 102 , 42 – 57 ( 2018 ). OpenUrl CrossRef PubMed 9. ↵ Yang , Y. et al. Diversity in intrinsic strengths of the human complement system: serum C4 protein concentrations correlate with C4 gene size and polygenic variations, hemolytic activities, and body mass index . J Immunol 171 , 2734 – 45 ( 2003 ). OpenUrl Abstract / FREE Full Text 10. ↵ Kerick , M. et al. Complement component C4 structural variation and quantitative traits contribute to sex-biased vulnerability in systemic sclerosis . NPJ Genom Med 7 , 57 ( 2022 ). OpenUrl PubMed 11. ↵ Zhou , D. et al. Low copy numbers of complement C4 and C4A deficiency are risk factors for myositis, its subgroups and autoantibodies . Ann Rheum Dis 82 , 235 – 245 ( 2023 ). OpenUrl Abstract / FREE Full Text 12. Pereira , K.M.C. et al. Impact of C4, C4A and C4B gene copy number variation in the susceptibility, phenotype and progression of systemic lupus erythematosus . Adv Rheumatol 59 , 36 ( 2019 ). OpenUrl PubMed 13. Mulvihill , E. et al. Elevated serum complement levels and higher gene copy number of complement C4B are associated with hypertension and effective response to statin therapy in childhood-onset systemic lupus erythematosus (SLE) . Lupus Sci Med 6 , e000333 ( 2019 ). OpenUrl Abstract / FREE Full Text 14. Juptner , M. et al. Low copy numbers of complement C4 and homozygous deficiency of C4A may predispose to severe disease and earlier disease onset in patients with systemic lupus erythematosus . Lupus 27 , 600 – 609 ( 2018 ). OpenUrl CrossRef PubMed 15. Lintner , K.E. et al. Gene copy-number variations (CNVs) of complement C4 and C4A deficiency in genetic risk and pathogenesis of juvenile dermatomyositis . Ann Rheum Dis 75 , 1599 – 606 ( 2016 ). OpenUrl Abstract / FREE Full Text 16. ↵ Chen , J.Y. et al. Effects of Complement C4 Gene Copy Number Variations, Size Dichotomy, and C4A Deficiency on Genetic Risk and Clinical Presentation of Systemic Lupus Erythematosus in East Asian Populations . Arthritis Rheumatol 68 , 1442 – 1453 ( 2016 ). OpenUrl PubMed 17. ↵ Lv , Y. et al. Confirmation of C4 gene copy number variation and the association with systemic lupus erythematosus in Chinese Han population . Rheumatol Int 32 , 3047 – 53 ( 2012 ). OpenUrl CrossRef PubMed 18. ↵ Kamitaki , N. et al. Complement genes contribute sex-biased vulnerability in diverse disorders . Nature 582 , 577 – 581 ( 2020 ). OpenUrl CrossRef PubMed 19. ↵ Lundtoft , C. et al. Complement C4 Copy Number Variation is Linked to SSA/Ro and SSB/La Autoantibodies in Systemic Inflammatory Autoimmune Diseases . Arthritis Rheumatol 74 , 1440 – 1450 ( 2022 ). OpenUrl PubMed 20. ↵ Hanscombe , K.B. et al. Genetic fine mapping of systemic lupus erythematosus MHC associations in Europeans and African Americans . Hum Mol Genet 27 , 3813 – 3824 ( 2018 ). OpenUrl PubMed 21. ↵ Langefeld , C.D. et al. Transancestral mapping and genetic load in systemic lupus erythematosus . Nat Commun 8 , 16021 ( 2017 ). OpenUrl PubMed 22. ↵ Dilthey , A.T. et al. HLA*LA-HLA typing from linearly projected graph alignments . Bioinformatics 35 , 4394 – 4396 ( 2019 ). OpenUrl CrossRef PubMed 23. ↵ Handsaker , R.E. et al. Large multiallelic copy number variations in humans . Nat Genet 47 , 296 – 303 ( 2015 ). OpenUrl CrossRef PubMed 24. ↵ Sekar , A. et al. Schizophrenia risk from complex variation of complement component 4 . Nature 530 , 177 – 83 ( 2016 ). OpenUrl CrossRef PubMed 25. ↵ Bell , A.D. , Usher , C.L. & McCarroll , S.A . Analyzing Copy Number Variation with Droplet Digital PCR . Methods Mol Biol 1768 , 143 – 160 ( 2018 ). OpenUrl CrossRef PubMed 26. ↵ Meisner , J. & Albrechtsen , A . Haplotype and population structure inference using neural networks in whole-genome sequencing data . Genome Res 32 , 1542 – 52 ( 2022 ). OpenUrl Abstract / FREE Full Text 27. ↵ Naito , T. & Okada , Y . HLA imputation and its application to genetic and molecular fine-mapping of the MHC region in autoimmune diseases . Semin Immunopathol 44 , 15 – 28 ( 2022 ). OpenUrl 28. ↵ Furukawa , H. et al. Human leukocyte antigens and systemic lupus erythematosus: a protective role for the HLA-DR6 alleles DRB1*13:02 and *14:03 . PLoS One 9 , e87792 ( 2014 ). OpenUrl PubMed 29. ↵ Shimizu , C. et al. HLA-C variants associated with amino acid substitutions in the peptide binding groove influence susceptibility to Kawasaki disease . Hum Immunol 80 , 731 – 738 ( 2019 ). OpenUrl PubMed 30. ↵ Pidala , J. et al. Amino acid substitution at peptide-binding pockets of HLA class I molecules increases risk of severe acute GVHD and mortality . Blood 122 , 3651 – 8 ( 2013 ). OpenUrl Abstract / FREE Full Text 31. ↵ Morris , D.L. et al. Unraveling multiple MHC gene associations with systemic lupus erythematosus: model choice indicates a role for HLA alleles and non-HLA genes in Europeans . Am J Hum Genet 91 , 778 – 93 ( 2012 ). OpenUrl CrossRef PubMed 32. ↵ Morris , D.L. et al. MHC associations with clinical and autoantibody manifestations in European SLE . Genes Immun 15 , 210 – 7 ( 2014 ). OpenUrl CrossRef PubMed 33. ↵ Wang , C.C. , Shen , C.H. , Lin , G.C. , Chen , Y.M. & Chen , I.C . Association of HLA alleles with cephalosporin allergy in the Taiwanese population . Sci Rep 14 , 17167 ( 2024 ). 34. Irish Schizophrenia Genomics , C. & the Wellcome Trust Case Control , C. Genome-wide association study implicates HLA-C*01:02 as a risk factor at the major histocompatibility complex locus in schizophrenia . Biol Psychiatry 72 , 620 – 8 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 35. ↵ Fujita , K.I. et al. Associations of HLA-C*01:02 and HLA-B*46:01 with regorafenib-induced erythema multiforme in Japanese patients with metastatic colorectal cancer . Clin Transl Sci 16 , 1741 – 1747 ( 2023 ). OpenUrl PubMed 36. ↵ Zhao , L.P. et al. The KAG motif of HLA-DRB1 (beta71, beta74, beta86) predicts seroconversion and development of type 1 diabetes . EBioMedicine 69 , 103431 ( 2021 ). OpenUrl PubMed 37. ↵ Fernando , M.M. & Vyse , T.J. Major histocompatibility complex and SLE . in Lahita’s Systemic Lupus Erythematosus 5 – 24 ( Elsevier , 2021 ). 38. ↵ Simoni , L. et al. Complement C4A Regulates Autoreactive B Cells in Murine Lupus . Cell Rep 33 , 108330 ( 2020 ). OpenUrl PubMed 39. ↵ Boteva , L. et al. Genetically determined partial complement C4 deficiency states are not independent risk factors for SLE in UK and Spanish populations . Am J Hum Genet 90 , 445 – 56 ( 2012 ). OpenUrl CrossRef PubMed 40. de Bakker , P.I. et al. A high-resolution HLA and SNP haplotype map for disease association studies in the extended human MHC . Nat Genet 38 , 1166 – 72 ( 2006 ). OpenUrl CrossRef PubMed Web of Science 41. McElroy , J.P. et al. Refining the association of MHC with multiple sclerosis in African Americans . Hum Mol Genet 19 , 3080 – 8 ( 2010 ). OpenUrl CrossRef PubMed Web of Science 42. ↵ Frisse , L. et al. Gene conversion and different population histories may explain the contrast between polymorphism and linkage disequilibrium levels . Am J Hum Genet 69 , 831 – 43 ( 2001 ). OpenUrl CrossRef PubMed Web of Science 43. ↵ Jeon , S. et al. Korean Genome Project: 1094 Korean personal genomes with clinical information . Sci Adv 6 , eaaz7835 ( 2020 ). OpenUrl FREE Full Text 44. ↵ Hwang , M.Y. , Choi , N.H. , Won , H.H. , Kim , B.J. & Kim , Y.J . Analyzing the Korean reference genome with meta-imputation increased the imputation accuracy and spectrum of rare variants in the Korean population . Front Genet 13 , 1008646 ( 2022 ). OpenUrl PubMed 45. ↵ Bolger , A.M. , Lohse , M. & Usadel , B . Trimmomatic: a flexible trimmer for Illumina sequence data . Bioinformatics 30 , 2114 – 20 ( 2014 ). OpenUrl CrossRef PubMed Web of Science 46. ↵ Yin , X. et al. Meta-analysis of 208370 East Asians identifies 113 susceptibility loci for systemic lupus erythematosus . Ann Rheum Dis 80 , 632 – 640 ( 2021 ). OpenUrl Abstract / FREE Full Text 47. ↵ Hochberg , M.C . Updating the American College of Rheumatology revised criteria for the classification of systemic lupus erythematosus . Arthritis Rheum 40 , 1725 ( 1997 ). OpenUrl CrossRef PubMed Web of Science 48. ↵ Petri , M. et al. Derivation and validation of the Systemic Lupus International Collaborating Clinics classification criteria for systemic lupus erythematosus . Arthritis Rheum 64 , 2677 – 86 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 49. ↵ Li , H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM . arXiv preprint arXiv:1303 . 3997 ( 2013 ). 50. ↵ DePristo , M.A. et al. A framework for variation discovery and genotyping using next-generation DNA sequencing data . Nat Genet 43 , 491 – 8 ( 2011 ). OpenUrl CrossRef PubMed Web of Science 51. ↵ Luo , Y. et al. A high-resolution HLA reference panel capturing global population diversity enables multi-ancestry fine-mapping in HIV host response . Nat Genet 53 , 1504 – 1516 ( 2021 ). OpenUrl CrossRef PubMed 52. ↵ Loh , P.R. , Palamara , P.F. & Price , A.L . Fast and accurate long-range phasing in a UK Biobank cohort . Nat Genet 48 , 811 – 6 ( 2016 ). OpenUrl CrossRef PubMed 53. ↵ Chang , C.C. et al. Second-generation PLINK: rising to the challenge of larger and richer datasets . Gigascience 4 , 7 ( 2015 ). OpenUrl CrossRef PubMed 54. ↵ Rubinacci , S. , Delaneau , O. & Marchini , J . Genotype imputation using the Positional Burrows Wheeler Transform . PLoS Genet 16 , e1009049 ( 2020 ). OpenUrl CrossRef PubMed 55. ↵ Moon , S. et al. The Korea Biobank Array: Design and Identification of Coding Variants Associated with Blood Biochemical Traits . Sci Rep 9 , 1382 ( 2019 ). OpenUrl CrossRef PubMed 56. ↵ Sakaue , S. et al. Tutorial: a statistical genetics guide to identifying HLA alleles driving complex disease . Nat Protoc 18 , 2625 – 2641 ( 2023 ). OpenUrl PubMed View the discussion thread. Back to top Previous Next Posted November 23, 2024. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Fine-mapping SLE-MHC associations revealed independent contributions of HLA missense variants and C4 copy number variations Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Fine-mapping SLE-MHC associations revealed independent contributions of HLA missense variants and C4 copy number variations Chae-Yeon Yu , Dong Mun Shin , Sung Min Kim , Yui Taek Lee , Sungwon Jeon , Sehwan Chun , So-Young Bang , Hye-Soon Lee , Xianyong Yin , Yong Cui , Xuejun Zhang , Jong Bhak , Soon Ji Yoo , Young Jin Kim , Bong-Jo Kim , Sang-Cheol Bae , Kwangwoo Kim medRxiv 2024.11.21.24317596; doi: https://doi.org/10.1101/2024.11.21.24317596 Share This Article: Copy Citation Tools Fine-mapping SLE-MHC associations revealed independent contributions of HLA missense variants and C4 copy number variations Chae-Yeon Yu , Dong Mun Shin , Sung Min Kim , Yui Taek Lee , Sungwon Jeon , Sehwan Chun , So-Young Bang , Hye-Soon Lee , Xianyong Yin , Yong Cui , Xuejun Zhang , Jong Bhak , Soon Ji Yoo , Young Jin Kim , Bong-Jo Kim , Sang-Cheol Bae , Kwangwoo Kim medRxiv 2024.11.21.24317596; doi: https://doi.org/10.1101/2024.11.21.24317596 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Genetic and Genomic Medicine Subject Areas All Articles Addiction Medicine (574) Allergy and Immunology (865) Anesthesia (304) Cardiovascular Medicine (4462) Dentistry and Oral Medicine (445) Dermatology (383) Emergency Medicine (611) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1517) Epidemiology (15251) Forensic Medicine (31) Gastroenterology (1132) Genetic and Genomic Medicine (6621) Geriatric Medicine (669) Health Economics (1002) Health Informatics (4564) Health Policy (1372) Health Systems and Quality Improvement (1617) Hematology (544) HIV/AIDS (1272) Infectious Diseases (except HIV/AIDS) (15938) Intensive Care and Critical Care Medicine (1107) Medical Education (624) Medical Ethics (147) Nephrology (670) Neurology (6643) Nursing (346) Nutrition (1001) Obstetrics and Gynecology (1149) Occupational and Environmental Health (957) Oncology (3350) Ophthalmology (981) Orthopedics (369) Otolaryngology (421) Pain Medicine (436) Palliative Medicine (130) Pathology (665) Pediatrics (1698) Pharmacology and Therapeutics (694) Primary Care Research (714) Psychiatry and Clinical Psychology (5465) Public and Global Health (9259) Radiology and Imaging (2212) Rehabilitation Medicine and Physical Therapy (1372) Respiratory Medicine (1198) Rheumatology (598) Sexual and Reproductive Health (716) Sports Medicine (533) Surgery (715) Toxicology (100) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a03c64c72be452ad',t:'MTc4MDEzMDAwMw=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();