STAT3-mediated allelic imbalance of novel genetic variant Rs1047643 and B-cell-specific super-enhancer in association with systemic lupus erythematosus

  1. Yanfeng Zhang  Is a corresponding author
  2. Kenneth Day
  3. Devin M Absher  Is a corresponding author
  1. HudsonAlpha Institute for Biotechnology, United States
  2. Zymo Research Corp, United States

Abstract

Mapping of allelic imbalance (AI) at heterozygous loci has the potential to establish links between genetic risk for disease and biological function. Leveraging multi-omics data for AI analysis and functional annotation, we discovered a novel functional risk variant rs1047643 at 8p23 in association with systemic lupus erythematosus (SLE). This variant displays dynamic AI of chromatin accessibility and allelic expression on FDFT1 gene in B cells with SLE. We further found a B-cell restricted super-enhancer (SE) that physically contacts with this SNP-residing locus, an interaction that also appears specifically in B cells. Quantitative analysis of chromatin accessibility and DNA methylation profiles further demonstrated that the SE exhibits aberrant activity in B cell development with SLE. Functional studies identified that STAT3, a master factor associated with autoimmune diseases, directly regulates both the AI of risk variant and the activity of SE in cultured B cells. Our study reveals that STAT3-mediated SE activity and cis-regulatory effects of SNP rs1047643 at 8p23 locus are associated with B cell deregulation in SLE.

Editor's evaluation

Through extensive computational data analyses and functional experiments, Zhang and colleagues reported a novel functional germline variant at the risk locus 8p23 for systemic lupus erythematosus. They provided evidence that the observed risk association in this locus may be mediated through the variant regulating a nearby susceptibility gene. This study advances the understanding of the genetic susceptibility and biology underlying systemic lupus erythematosus.

https://doi.org/10.7554/eLife.72837.sa0

Introduction

Super-enhancers (SEs) are recently discovered large domains of clustered enhancers (Parker et al., 2013; Whyte et al., 2013). The extraordinary feature of SEs is the extremely high and broad enrichment of enhancer-related transcription factors (TFs), H3K4me1 and H3K27ac modifications, resulting in high capability to enhance gene expression programs (Whyte et al., 2013). A large quantity of SEs show cell/tissue specificity (Vahedi et al., 2015), thereby they have become principal determinants of cell identity (Hnisz et al., 2013). Nonetheless, disease-associated SEs, in particular those exhibiting aberrant activity in autoimmune diseases, are less characterized.

Signal transducer and activator of transcription 3 (STAT3), as one of seven STAT family members, is activated by phosphorylation at tyrosine 705 (Y705) and/or at serine 727 (S727) (Decker and Kovarik, 2000). After import to the nucleus, the phospho-STAT3 (pSTAT3) modulates gene transcription by binding its target sequence (Levy and Darnell, 2002). STAT3 has gained broad attention because it plays a key role in a variety of pathophysiological immune responses related to lymphocyte development and differentiation, and in other cellular processes of normal and tumor cells (Yu et al., 2009).

Systemic lupus erythematosus (SLE) is an autoimmune disease that is known to be associated with an array of abnormal immune cell signaling. B-cell hyperactivity in auto-antigen recognition and interaction with T-cells, which ultimately results in multi-organ damage, is central to the pathogenesis of SLE (Rahman and Isenberg, 2008). Genetic factors conferring a predisposition to the development of SLE have been widely characterized. Over 100 loci have been implicated in SLE by genome-wide association studies (GWAS) (Catalina et al., 2020; Yin et al., 2021). Among them, several genes and/or loci are potent as putative drivers of the disease. For example, genetic risk variants at the promoter of BLK at 8p23 locus alter BLK transcription activity and thus contribute to autoreactive B-cell responses (Guthridge et al., 2014). Nonetheless, the GWAS-identified genetic variants together explained approximately 30% of the heritability of SLE (Sun et al., 2016; Morris et al., 2016), suggesting a requirement of further efforts to explain the missing heritability of SLE. Meanwhile, there is growing evidence that genetic risk factors behave in a context-dependent or cell-specific manner (Guthridge et al., 2014; Gallagher and Chen-Plotkin, 2018). Thus, for SLE and other autoimmune diseases, there is a need to identify the regulatory programs in which these genetic factors impact the immune cell developmental processes.

One approach for tying genetic risk to function in the post-GWAS era (Gallagher and Chen-Plotkin, 2018), is a measurement of allelic imbalance (AI) on two alleles at a given heterozygous locus, typically at single-nucleotide polymorphism (SNP). The genes and/or loci with SNPs exhibiting AI could provide a strong foundation for implicating the genetic or epigenetic mechanisms linked to complex traits or diseases (Pastinen and Hudson, 2004; Yan et al., 2002). As a readout of AI, analyses of allele-specific chromatin accessibility and allele-specific RNA expression have accumulated a wealth of interesting findings, including functional cis-regulation (Li et al., 2013; Zhang et al., 2020b), genomic imprinting (Pollard et al., 2008), X-chromosome inactivation or escape (Zhang et al., 2020a). Therefore, tracking AI difference in a comparison between diseases and controls may enable to uncover novel functional variants associated with complex diseases. In this study, we describe one such strategy through integrative multi-omics analysis to discover known or novel functional variants associated with SLE, and report on the identification of a novel risk variant rs1047643 and B-cell-specific SE in B cells with SLE. We further demonstrate that the resultant allelic imbalanced variant and SE activity are directly controlled by STAT3, a master TF that plays a critical role in B cell development and highly associates with autoimmune diseases.

Results

Multi-omics data summary

Functional genomics sequencing data sets comprising 279 samples from eleven studies were collected (Supplementary file 1). Of eleven studies, seven are SLE case-control studies with data across three immune cell types including B cells, T cells, and Neutrophils (Supplementary file 2). Also included in the present study were SNP microarray data from a SLE GWAS study (n = 2279).

Identification of SLE-associated variant showing AI at both chromatin and RNA levels

We next designed a two-stage study (Figure 1) to identify putative SLE-associated functional variants. In stage I, also named as the discovery stage, two chromatin accessibility (ATAC-seq) data sets (Accession ID: GSE118253 and GSE71338, Supplementary file 1) comprised 49 samples were analyzed. We focused on those variants displaying difference in AI of chromatin accessibility at heterozygous SNP sites in a comparison between SLE and controls (see Materials and methods in detail). In total, 18,456 (the GSE118253 dataset) and 4319 (the GSE71338 dataset) SNPs were tested. We collected the resulting SNPs with nominal (unadjusted) p < 0.05. From the reciprocal validation between two data sets, three SNPs (rs1047643, rs246367, and rs72642993) were identified to show the significant AI (combined p < 0.01) in B cells from patients with SLE, relative to controls (Figure 2A and Figure 2—figure supplement 1). Based on annotations from the GWAS central (Beck et al., 2020) and HaploReg (Ward and Kellis, 2016) databases, we then focused on the rs1047643 because it is the most promising variant. Meanwhile, in B cells at different stages, the allelic preference of chromatin accessibility for SNP rs1047643 is alterable. For example, the T allele exhibits more preferential chromatin accessibility in activated B cells from patients, relative to the C allele. However, the direction is reversed in SLE naive B cells.

Schematic of the study design.

On the basis of the functional genomic data feature, a two-stage study was designed. Summary of data sets is available in Supplementary files 1-2.

Figure 2 with 3 supplements see all
Change of allelic chromatin accessibility and expression in B cell subtypes from SLE patients and controls.

(A) Forest plot showing AI of allelic chromatin state of SNP rs1047643 in both resting naive (rN) and activated (Non-rN) B cells in patients of SLE compared with healthy controls. The p-value per study and combined p-value (summary) are calculated based on the linear regression model and Fisher’s method, respectively. The plot in the right panel displays the 95% of confidence interval of beta-value. (B–C) Boxplots showing allelic expression of SNP rs1047643 in both rN and activated B cells in patients with SLE as compared with healthy individuals. All raw data are available in Figure 2—source data 1.

Figure 2—source data 1

Source files for presenting results in Figure 2.

This zip archive contains all source data used for the quantitative analyses shown in Figure 2.

https://cdn.elifesciences.org/articles/72837/elife-72837-fig2-data1-v2.zip

Because the rs1047643 is located in the first exon of FDFT1 gene (Figure 3E), it enables us to test the functionality of this variant at the transcriptional level. Analyzing RNA-seq data (Accession ID: GSE118254), we determined the AI of RNA transcripts for the rs1047643 by comparing the AR values (see Materials and methods in detail) between SLE and controls. In line with results shown above, we observed the dynamic AI pattern on the transcriptional level for the rs1047643 (Figure 2B). Meanwhile, this dynamic allelic expression pattern is specific during B cell development with SLE (Figure 2C).

Figure 3 with 2 supplements see all
Association analysis and functional prediction of SNP rs1047643.

(A) Association results for the SNP rs1047643 with SLE risk in single marker analyses. MAF, minor allele frequency; OR, odds ratio; CI, confidence interval. Adjusted p-trend: after adjustment for 12 GWAS index SNPs (shown in E) in a logistic regression model. (B) Haplotype analyses of the two SNPs (SNP1: GWAS indexed SNP rs17807624; SNP2: rs1047643) in relation to SLE risk. Baseline (the reference haplotype) represents the alleles associated with a reduced risk in two SNPs. (C) Barplot showing the genomic length of chromHMM-annotated enhancer state on the super-enhancer region (blue highlighted in 3 C) in 43 epigenomes. (D) Plot shows the eQTL result of SNP rs1047643 in whole blood or B cells from three databases (shown in y-axis). (E) Genomic annotations of the SNP rs1047643. The three tracks show locations of 13 GWAS index SNP, gene annotation and 15-state chromatin segments in CD19+ B cells at 8p23 locus, respectively. Vertical blue and purple lines, represents the location of super-enhancer and SNP rs1047643, respectively. (F) Long-range interaction between a super-enhancer and SNP rs1047643. The two tracks show chromatin interactions from two independent studies using whole-genome Hi-C and capture Hi-C technologies, respectively. Orange curves show the interactions between the super-enhancer and the SNP rs1047643. (G) Heatmaps showing the 3D DNA interactions at 8p23.1 locus in eight cell lines. The rectangle represents interactions between the super-enhancer and the SNP rs1047643. All raw data are available in Figure 3—source data 1.

Figure 3—source data 1

Source files for presenting results in Figure 3.

This zip archive contains all source data used for the quantitative analyses shown in Figure 3.

https://cdn.elifesciences.org/articles/72837/elife-72837-fig3-data1-v2.zip

Association with SLE risk in American Hispanic and European populations

Because SNP rs1047643 has not been reported to be associated with the susceptibility of SLE and other autoimmune diseases, we next tested the association using a dataset from an SLE GWAS case-control study. Employing the univariate analysis for SNP rs1047643 in samples from Hispanic populations, we identified an association of the rs1047643 with SLE risk (OR per C effect allele = 0.77, 95% CI 0.54–0.99, p = 0.023 after adjusting for covariates, Figure 3A), albeit not reaching the significance after adjustment for 12 GWAS index SNPs (the top track in Figure 3E, where one SNP rs2736336 is excluded due to its multivariate alleles). We also examined an association in European population (n = 19,468) from Bentham et al., 2015 study. A similar result was observed for the rs1047643 (OR per C effect allele = 0.91, 95% CI 0.84–0.98, p = 0.02, Supplementary file 3).

An analysis of linkage disequilibrium (LD) with each of 12 GWAS tag SNPs showed that there was no strong LD (r2 <0.1) between SNP rs1047643 and GWAS SNPs in European population (Supplementary file 4). This result indicates that SNP rs1047643 is a potentially SLE GWAS independent functional variant. Of the 12 index SNPs, indeed, one index SNP rs17807624 with the statistical significance with p < 1.5 × 10–3 using the univariate analysis, is the top signal to which the SNP rs1047643 is conditional. Thus, we performed haplotype analyses on these two SNPs (index SNP rs17807624 and rs1047643, Figure 3B). Compared with the reference haplotype, which carries the alleles associated with a reduced risk in two SNPs, haplotype 2, which carries the risk-associated alleles, showed a significant association (p = 0.03).

Functional annotation

An analysis of eQTL data derived from three independent cohorts indicated both proximal ( < 200 kb) and distal ( > 200 kb) regulatory potential for the SNP rs1047643 in normal B or blood cells (Figure 3D). Interestingly, besides correlated with three adjacent genes (FDFT1, CTSB and NEIL2), the rs1047643 is also an eQTL linked with an upstream BLK gene in a distance of ~300 kb, a result that is detected in two independent data sets. An analysis of RNA-seq data from two independent studies (Accession ID: GSE118254 and GSE92387, Supplementary file 1) consistently showed that expression patterns for two representative genes (BLK and FDFT1) are gradually increased in a developmental process from naive to memory B cells, in particular, the double negative memory B cell subset in patients with SLE, the pattern that is not observed in controls (Figure 2—figure supplements 2 and 3).

By searching for enhancers and other regulatory elements across 8p23 locus from a dataset of the 127 epigenomes from Roadmap, we identified a SE with a length of 7 kb in the upstream of BLK gene in CD19+ B cells (Epigenome ID: E032, Figure 3E). An analysis of enhancer elements across the 127 epigenomes showed 43 (33.9%) epigenomes had enhancers at this SE region. Comparative analysis of the enhancer length at this SE region on the 43 epigenomes further showed that this SE is specific in CD19+ B cells (Epigenome ID: E032, Figure 3C).

Analyzing Hi-C data sets from two independent studies in GM12878 cells, we observed a DNA looping between the SNP rs1047643 within FDFT1 and the SE region (Figure 3F). More importantly, in GM12878 B-lymphoblastic cells, this SE region has a wealth of long-range interactions with adjacent genes (e.g., BLK) and functional elements. In contrast, in another seven cells (Figure 3G), as well as in normal T cells (Figure 3—figure supplement 1) and nine selected tissues (Figure 3—figure supplement 2), these interactions are either much weaker or completely absent. These results indicate that the physical interaction between SNP rs1047643 and SE region, and many interactions with this SE, are specific to B-lymphocytes.

Specificity in B cells

We then hypothesized that the SE region may show aberrant activity in B cells from SLE patients. To test this hypothesis, we conducted quantitative analysis on the same ATAT-seq data (Accession ID: GSE118253 and GSE71338) used in stage I (see Materials and methods in detail). Comparison of SE activity in a quantitative manner between SLE patients and controls indicated that the SE activity is gradually increased through B cell development in SLE patients (Figure 4A–B), with a hyper-activity being observed in double negative (DN) B cells in patients, relative to controls (Figure 4B–C). It should be noted that such a pattern is not observed in the background sampling (Figure 4—figure supplement 1). Similarly, the rs1047643-containing promoter activity also shows up-regulation toward B cell development in SLE patients (Figure 4D–E). In a comparison of B cell development on activities of SE and FDFT1 promoter regions in two individuals, the chromatin accessibility on both regions in an individual with SLE is increased during B cell development, but remains relatively unchanged in the healthy individual (Figure 4F–H).

Figure 4 with 3 supplements see all
Aberration of super-enhancer and FDFT1 promoter region in B cell subtypes from SLE patients.

(A) Empirical cumulative distribution of TPM values per 50 bp window across the 7 kb SE region in B cell subsets for disease and control groups. (B) Plots showing the TPM values at the third quartile (Q3) across B cell subtypes as a comparison between SLE and controls. (C) Empirical cumulative distribution of TPM values on the SE region (same as shown in A) in a comparison between two groups across four B cell subtypes. (D) Boxplots showing the TPM values per 50 bp window at the FDFT1 promoter region in B cell subtypes for SLE and controls. The black lines and grey areas represent the linear regression results towards the B cell development from T3 to DN stages, and 95% of CI. (E) Plots showing the correlation between super-enhancer and FDFT1 promoter regions based on mean TPM values with respect to B cell subtypes in SLE and controls. (F) Wiggle plot showing the enrichment of open chromatin states at 8p23.1 locus in B cell subtypes for two individuals (a healthy individual at upper panel, and a patient with SLE at lower panel). Purple and green vertical lines represent the locations for super-enhancer and FDFT1 promoter, respectively. Quantitative comparison of chromatin accessibility states in SE (G) and FDFT1 promoter regions (H) with respect to B cell subtypes. All raw data are available in Figure 4—source data 1.

Figure 4—source data 1

This txt file contains source data used for the quantitative analyses shown in Figure 4.

https://cdn.elifesciences.org/articles/72837/elife-72837-fig4-data1-v2.txt

We also quantitatively compared open chromatin states of SE and FDFT1 promoter regions in resting naive B cells (Accession ID: GSE71338). Concordant with the results from active B cell subsets, the open chromatin states on both regions are low in non-active B cells from SLE patients, relative to healthy controls (Figure 4—figure supplement 2).

We further conducted quantitative analyses on ATAC-seq data from another two independent studies in two immune cell types, T cells, and neutrophils (Accession ID: GSE139359 and GSE110017, Supplementary file 1). The results showed that there was no marked enrichment of ATAC-seq reads on both the SE and FDFT1 promoter regions in these two immune cell types for both SLE and controls (Figure 4—figure supplement 3). Collectively, these results suggest a B cell specific, rs1047643-interacting SE whose activity is aberrant in SLE B cell development.

Hypomethylation in SLE B cells

We further analyzed DNA methylation in the SE region using RRBS data in B cell development in a comparison between SLE and controls (Accession ID: GSE118255, Supplementary file 1). Our results show that DNA methylation levels on the SE region are gradually decreased in the developmental process from resting native (rN) to memory B cells in patients with SLE (Figure 5A). In contrast, there is no such obvious change of DNA methylation pattern in the control group. In addition, an analysis of DNA methylation levels on the background sampling doesn’t show such a difference between SLE and controls (Figure 5—figure supplement 1), further suggesting that epigenetic change on the SE region is biologically meaningful in B cell development in SLE.

Figure 5 with 1 supplement see all
Hypomethylation in super-enhancer region in B cell subtypes from SLE patients.

(A) Boxplots showing the CpG methylation levels per 50 bp window in 7 kb SE region in B cell subtypes for SLE and control groups. The black and red lines represent the linear regression results towards the B cell development from rN to DN stages for SLE and controls, respectively. (B) Plots showing the correlation between TPM values (y-axis) and DNA methylation levels (x-axis) averaged over each B cell type in SLE and controls. All raw data are available in Figure 5—source data 1.

Figure 5—source data 1

This txt file contains source data used for the quantitative analyses shown in Figure 5.

https://cdn.elifesciences.org/articles/72837/elife-72837-fig5-data1-v2.txt

A correlation analysis also showed a marked negative correlation between open chromatin states (TPM values, also presented in Figure 4E) and DNA methylation levels at the SE region in the SLE group, relative to the healthy controls (Figure 5B). Together, these results reinforce the aberrant activity of SE in developmental process of B-lymphocytes in patients with SLE.

STAT3 binding on both super-enhancer and rs1047643-residing regions

TF-motif enrichment and binding analysis using the ENCODE TF ChIP-seq dataset (v3) predicted that STAT3 may bind to both the SNP rs1047643-containing promoter and SE regions (data not shown). To validate the finding, we designed two pairs of primers (SE5 and SE3, Figure 6A) to determine the STAT3 binding on SE region and its contribution to the SE activity using STAT3, H3K4me1 and H3K27ac ChIP-qPCR assays in GM11997 cells. Under normal culture conditions, we validated that pSTAT3, H3K4me1 and H3K27ac modifications are remarkably enriched on the SE region (Figure 6A, B and D), but not on the negative control region (Figure 6—figure supplement 1) in B-lymphoblastic cells, relative to IgG mock controls. We then conducted both the inhibition and activation of STAT3 DNA-binding activity using two small molecules. In B-lymphoblastic cells challenged with S3I-201, a STAT3 DNA binding inhibitor, both the DNA binding of STAT3 on SE region and the SE activity are significantly reduced (Figure 6A and B), relative to control. In GM11997 cells treated with ML115, a selective activator of STAT3 (41), both the STAT3 DNA-binding capability on SE region and the SE activity are significantly increased (Figure 6D and E), relative to controls. These results together demonstrate that STAT3 directly modulates the SE activity.

Figure 6 with 3 supplements see all
Contribution of STAT3 modulates the enhancer activity and SNP-residing locus in cultured GM11997 cells.

(A) ChIP-qPCR for H3K27ac (left lower panel), H3K4me1 (right lower panel) and pSTAT3 (B) at 8p23 super-enhancer region following 40 μM S3I-201 treatment for 24 hr. Upper panel: UCSC genome browser showing the location of two pairs of qPCR primers (SE5 and SE3) on the SE region (yellow). Two tracks shown below are the enrichment of H3K27ac and H3K4me1 across the SE region. (C) Allelic ChIP-qPCR for pSTAT3 binding on rs1047643 (T vs C alleles) following S3I-201 treatment for 24 hr. (D–E) ChIP-qPCR for H3K27ac (D), and pSTAT3 (E) at 8p23 super-enhancer region following 100 nM ML115 treatment for 6 hr. (F) Allelic ChIP-qPCR for pSTAT3 binding on rs1047643 in cells that have been challenged with ML115 for 6 hr as indicated. Note: the fold changes for the rs1047643-associated BLK and FDFT1 genes in response to small molecules compared to vehicle (0.1% DMSO) as control, which was set as one in all cases, are presented. NS, not significance; *, p < 0.05; **, p < 0.01; ***, p < 0.005.

We next tested whether the STAT3 might also regulate the rs1047643-residing regions. Using allelic qPCR assay, we confirmed that genomic DNA in the GM11997 cells carries a heterozygous variant for the SNP rs1047643 (Figure 6—figure supplement 2), enabling the AI analysis in this cell model. In GM11997 cells treated with the STAT3 inhibitor S3I-201, STAT3 binding on the risk allele T is significantly reduced, relative to the rs1049643-C allele (Figure 6C). Conversely, we observed an increase of STAT3 DNA binding at the rs1049643-T allele in cells stimulated with the STAT3 activator ML115 (Figure 6F). We further confirmed the findings in cells treated with Cucurbitacin I or IL-6 that acts an inhibitor and stimulator (Figure 6—figure supplement 3) of the Janus Kinase (JAK)/STAT3 signaling pathway (Blaskovich et al., 2003; Hunter and Jones, 2015), respectively.

Consistent with findings in the STAT3 DNA-binding study, we observed a significant change of the rs1049643-T allele at the transcriptional level after treatment with S3I-201 and ML115 at different concentrations and duration times, relative to the C allele (Figure 7A and C and Figure 7—figure supplement 1). These results suggest that the risk rs1049643-T allele is preferentially bound and regulated by STAT3 in B cells.

Figure 7 with 1 supplement see all
Expression of two alleles on SNP rs1047643 and its linked genes in cultured cells.

Left panel: allelic RT-qPCR on SNP rs1047643 (T vs C alleles) following S3I-201 (A) and ML115 (C) treatment for 24 hr, respectively. Right panel: RT-qPCR analysis showing the fold changes for the rs1047643-associated BLK and FDFT1 genes in response to different concentrations of S3I-201 (B) and ML115 (D) compared to vehicle (0.1% DMSO) as control, which was set as one in all cases, are presented. *, p < 0.05; **, p < 0.01; ***, p < 0.005.

Finally, we determined RNA expression of BLK and FDFT1, two representative genes that correlate with the risk rs1047643. The expression levels of both genes are decreased with the treatment of S3I-201 (Figure 7B), and upregulated with the STAT3 activator ML115 (Figure 7D). Together, these results suggest the STAT3-binding risk allele T is associated with increased expression of BLK and FDFT1.

Discussion

In the present study, by integrating a variety of functional genomic data, we performed AI analysis to uncover novel functional promising variants and their regulatory targets in association with SLE. Of note, the diversity of genomic data types from this comprehensive data collection for autoimmune diseases allowed us to develop an approach not used before for accessing the role of variants in SLE disease activity.

One of the most significant findings is the identification of a novel risk variant rs1047643. The association study shows that the rs1049643-T is a risk allele for SLE. Our AI analyses indicate that the rs1049643-T allele resides in more open chromatin state and has higher expression in SLE memory B cell subsets, relative to the C allele. Functional study further provides evidence that the rs1049643-T allele is preferentially bound by STAT3. The SNP rs1047643 is also an eQTL linked with both proximal and distal genes, including BLK, the gene that plays a critical role in B lymphocyte development (Saijo et al., 2003). These results demonstrate that this novel SLE-associated risk rs1047643 whose functionality is mediated by STAT3, may play a role in allele-specific control of adjacent genes at 8p23 locus in B cells. Despite no report for association with other autoimmune diseases, this SNP has been associated with multiple myeloma (Van Ness et al., 2008) and follicular lymphoma (Skibola et al., 2008), two malignant diseases whose pathogenesis is partially associated with the dysfunction of B cells. Specifically, hyperactive STAT3 has been reported to be associated with poor survival in both diseases (Huang et al., 2013; Jung et al., 2017). Therefore, our findings may provide a clue for genetic and mechanical studies on those B cell associated diseases.

Another intriguing finding in this study is the identification of an aberrant activity of a SE in lupus B cell subsets, particularly the hyperactivity in memory B cells. In contrast, there is no enhancer activity in other immune cells (T cells and neutrophils analyzed in this study) in patients with SLE. We also demonstrate that the aberrant activity of the SE can be mediated by STAT3. Some studies have consistently reported a critical role of STAT3 in the B cell maturation, differentiation, as well as the autoimmunity (Avery et al., 2010; Ding et al., 2016). These reports further support the significance of STAT3-mediated SE aberration in B cells with SLE.

Several studies have highlighted the 8p23 locus as a major SLE susceptibility region (Demirci et al., 2017). Our study further expands the significance at this locus. For example, our study and others together suggest that there are a few cis-eQTLs linked with transcriptional levels of BLK (Guthridge et al., 2014; Demirci et al., 2017). Thus, we speculate that the 8p23 locus may play functional roles in B cell development in both genetic and epigenetic fashions. Besides the SNP rs1047643 discovered in the present study, there are 13 SLE-associated GWAS leading SNPs reported in this locus. Of 13 SNPs, six SNPs (Figure 3E) directly sit in the SE region, suggesting these risk variants may play roles in a genetic interaction way, in spite of the unavailability of AI analysis, due to either low coverage (read depth <8) or homozygosity in most or all samples for the 13 SNPs. Epigenetically, the SLE-associated SE has physical interactions with adjacent genes, including BLK and FDFT1, and the risk rs1047643-residing region. This indicates a potentially complex role of the variant rs1047643 for broad regulation by physically contacting the SE. Thus, our data provide new insights into the molecular mechanisms by merging genetic susceptibility with epigenetic impacts on gene expression for autoimmune diseases.

The FDFT1 is a gene encoding for squalene synthase, the enzyme that catalyzes the early step in the cholesterol biosynthetic pathway (Tozawa et al., 1999). Previous studies have shown dyslipidemia, with elevations in total cholesterol, low-density lipoprotein, triglyceride levels in patients with lupus (Tisseverasinghe et al., 2006), especially in the active disease. Our multi-omics data indicate that the SNP rs1047643-linked FDFT1 may be aberrantly activated in B cell development in SLE patients, thereby providing an insight into the genetic implication of lipid metabolism for autoimmune diseases.

The limitations of this study include, due to the presence of six SLE GWAS tagging SNPs in SE region, we are unclear how they genetically influence the SE activity during B cell development. Second, it remains unclear how the AI pattern occurs in naive B cells with lupus. The C allele shows more open chromatin state in SLE naive B cells, this can’t be explained by STAT3 allelic DNA binding at the T allele. This implies that some other factors may also contribute to this dynamic AI pattern. Third, no functional studies on genetic manipulation at the rs1047643 prevent us draw the further conclusion about whether and how the rs1047643 impact the STAT3 binding in the present study.

In addition, it should be noted is the implementation of linear regression model in the initial step to identify the allelic difference signals between SLE and controls. Due to small sample size, there is no statistical power to analyze with more optimal statistical models, such as the logistic regression model. Meanwhile, the unavailability of other variables in regression analysis further restricts this study to remove the potential confounding factors. Together, our analysis may miss some potential AI signals and disable to evaluate the causation.

In conclusion, we identified a novel functional variant and B-cell-specific SE in association with the SLE pathogenesis, both mediated by STAT3, and influencing their gene targets. This insight into the mechanism by which manipulation of STAT3 affects the SE activity and its associated gene expression in B cells may have implications for future drug development in autoimmunity.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Chemical compound, drugML115Cayman ChemicalCayman Chemical: 15,178Madoux et al., 2010
Chemical compound, drugS3I-201Sigma-AldrichSigma-Aldrich: SML0330
Chemical compound, drugCucurbitacin ISigma-AldrichSigma-Aldrich: C4493
Chemical compound, drugRecombinant human IL-6Cell Guidance SystemsCell Guidance Systems: GFH10AF
AntibodyPhospho-STAT3 (Ser727)Thermo Fisher ScientificThermo Fisher Scientific Cat# PA5-17876; RRID:AB_10980044
AntibodyAnti-Histone H3 (acetyl K27)AbcamAbcam Cat# ab4729; RRID:AB_2118291
AntibodyH3K4me1 Recombinant Polyclonal AntibodyThermo Fisher ScientificThermo Fisher Scientific Cat# 710795; RRID:AB_2532764
Antibodynormal mouse IgGSanta Cruz BiotechnologySanta Cruz Biotechnology Cat# sc-2025; RRID:AB_737182
Antibodynormal rabbit IgGSanta Cruz BiotechnologySanta Cruz Biotechnology Cat# sc-2027; RRID:AB_737197
Cell line (H. sapiens)GM11997CoriellCoriell Cat# GM11997; RRID:CVCL_5C55
Sequence-based reagentChIP-qPCR primersThis paperSee Supplementary file 5
Sequence-based reagentRT-qPCR primersThis paperSee Supplementary file 5
Sequence-based reagentAllelic qPCR primersThis paperSee Supplementary file 5
Software, algorithmRR Foundationhttps://www.r-project.orgVersion 4.0.2
Software, algorithmHisat2Kim et al., 2019Version 2
Software, algorithmAllelic imbalance analysis and plotsThis paper (Zhang, 2021)The R code used for the
AI analysis can be
accessed via github
at https://github.com/youngorchuang/Allelic-imbalance-analysis,
(copy archived at swh:1:rev:f0db42af8fed130ebbfe0b46abf992300dadddd6)
Software, algorithmHiCUPWingett et al., 2015
Commercial assay or kitMycoplasma detection kitSigma-AldrichSigma-Aldrich:MP0025
Commercial assay or kitSuperScript III reverse transcriptaseThermo Fisher ScientificThermo Fisher Scientific:18080044
Commercial assay or kitLuna Universal qPCR Master MixNew England BiolabsNew England Biolabs:M3003X

Data collection

Request a detailed protocol

We collected a variety of functional genomics data, including ATAC-seq, RNA-seq, reduced-representation bisulfite sequencing (RRBS), Hi-C data (Supplementary file 1), from the Gene Expression Omnibus (GEO) and ArrayExpress databases. Meanwhile, we downloaded genotype and Epidemiological data from a SLE case-control study (accession: phs001025.v1) in Hispanic population (1393 cases and 886 controls) from the dbGaP database with approval (accessed 29 Sep 2020).

Analysis of RNA-Seq and ATAC-Seq data

Request a detailed protocol

RNA-seq data were analyzed as described previously with few modifications (Zhang et al., 2016). In brief, raw sequencing data were mapped to the human reference genome (hg19) using Hisat2 program (Kim et al., 2019) with the default setting. Aligned data were processed and converted into BAM files using SAMtools program (Li et al., 2009). To quantify gene expression levels, read counts were calculated using the featureCounts (version 2.0.2) program, then implemented in the edgeR package to calculate the count per million (CPM) values.

We used a similar method described previously with several modifications (Zhang et al., 2015) to process the ATAC-seq data. In brief, raw sequencing data were mapped to the human reference genome (hg19) using Bowtie2 program (Langmead and Salzberg, 2012) with the default setting. Tag per million (TPM) metric, a method commonly used for read counting normalization, was used to quantitatively present the enrichment of open chromatin states across regions of interest.

Identification of allelic imbalance sites

Request a detailed protocol

We used a similar approach described in our previous study to call variants and allelic analysis for both RNA-seq and ATAC-seq data (Zhang et al., 2020a). Briefly, the deduplicated reads in BAM format were realigned and recalibrated, and genetic variants were called in a multiple-sample joint manner implemented in the GATK toolkit (version 3.3). We next filtered out variants as follows: (Parker et al., 2013) mapping quality score <20, (Whyte et al., 2013) ≥ 3 SNPs detected within 10 bp distance, (Vahedi et al., 2015) variant confidence/quality by depth <2, (Hnisz et al., 2013) strand bias score >50, (Decker and Kovarik, 2000) genotype score <15 and (Levy and Darnell, 2002) read depth <8. Then, we extracted SNPs annotated from dbSNP (Build 150) that were called as heterozygotes for each sample. For a reasonable comparison, those heterozygous SNPs identified at least triple in both case and control samples were retained for further analysis.

For a given heterozygous SNP, we calculated allelic ratio (AR) based on read coverage onto two alleles. For RNA-seq data, the resulting AR values were used to compare the AI difference of RNA transcripts between cases and controls. For ATAC-seq data, by testing for associations between AR of each heterozygous SNP (as response variable) and SLE disease status as categorical variable (case/control comparison, control and case are coded as 0 and 1, respectively) implemented in the regression analysis (see below), we analyzed the AI difference of chromatin accessibility between cases and controls. Then, the p-value and beta coefficient were calculated to estimate the significance of the association, and the differences between cases and controls, respectively.

Genetic association analysis

Request a detailed protocol

For genotype data from a SLE case-control study in Hispanic population, all typed SNPs in chromosome eight were extracted for imputation using TOPMed Imputation Server (Das et al., 2016). To test SNP rs1047643 in association with SLE, we used a method described previously for univariate and haplotype analyses (Shi et al., 2016). In brief, the per-allele odds ratio (OR) and 95% confidence interval (CI) for the rs1047643 was estimated for SLE risk using a log-additive logistic model with covariates of five countries of the Hispanic population, sex and five principal components (PCs). We used the haplo.stats package in R for haplotype analyses with five countries of the Hispanic population, sex and five PCs as covariates. For SLE GWAS data in European population from Bentham et al., 2015 study, we downloaded summary statistical data (Accession ID: GCST003156) from GWAS catalog (Buniello et al., 2019) and extracted statistical results for the SNP rs1047643.

A dataset of GWAS leading SNPs was downloaded from the GWAS Catalog (Buniello et al., 2019). Then we extracted SLE-associated SNPs at 8p23. For each indexed SLE-associated SNP at 8p23, we tested the linkage disequilibrium (LD) score (r2) with query SNP rs1047643 from the data set of the Phase 3 of the 1,000 Genomes Project in European population using LDlink web tool (Machiela and Chanock, 2015).

Super-enhancer annotation

Request a detailed protocol

We downloaded whole-genome chromatin state segmentation data (core 15-state model) for 127 cell types from the Roadmap project. As Parker et al., 2013 defined, we consider contiguous genomic region marked by states 6–7 (enhancer states, annotated by chromHMM) with ≥3 kb as SE in a cell type. Then, we extracted and annotated super-enhancers on 8p23 locus.

Analysis of eQTL data

Request a detailed protocol

We collected eQTL data sets from three large-scale studies, the Genotype-Tissue Expression (GTEx, v8) (Laboratory, 2017), the Haploreg v4.1 dataset (Ward and Kellis, 2016) and the study by Westra, 2013. By searching for the SNP rsID or the coordinate, we extracted the linked genes with query SNPs and plotted the results based on the significance and studies.

Hi-C data analysis

Request a detailed protocol

For in situ Hi-C dataset (Accession ID: GSE63525), we downloaded the Hi-C binary file from Rao et al. study (Rao et al., 2014) and extracted the observed long-range interactions normalized with Knight-Ruiz matrix balancing (KR) method at 10 kb resolution across the 8p23.1 region (the coordinate: chr8:11260000–11740000, hg19).

For other genome-wide Hi-C (Accession ID: GSE113405) and capture Hi-C (CHi-C) datasets (Accession ID: GSE81503 and E-MTAB-6621), we used the Hi-C Pipeline (HiCUP) (Wingett et al., 2015) to truncate and align reads to the human reference genome. The deduplicated data were then processed using the Homer pipeline (Heinz et al., 2010) to call the significant chromatin interaction at 10 kb resolution with the support of ≥5 reads and p ≤ 0.001. The resulting interactions were visualized using UCSC Genome Browser or Sushi package in R environment.

DNA methylation analysis

Request a detailed protocol

We downloaded the processed RRBS dataset of DNA methylation profiles on each CpG site from Scharer et al., 2019 report, then extracted and compared CpG methylation levels on a region of interest between SLE and healthy controls.

Cell culture

Request a detailed protocol

GM11997 B lymphoblastic (purchased from Coriell Institute) cells were cultured in RPMI-1640 medium, supplemented with 10% FBS (Thermo Fisher Scientific), 2 mM L-glutamine and 1% penicillin-streptomycin at 37 °C with 5% CO2. These cells are mycoplasma-negative when tested with PCR-based mycoplasma detection kit. For perturbation of STAT3, B cells were plated in 12-well plates or 10 cm dishes one day prior to the experiment. Cells were then treated with S3I-201, ML115, Cucurbitacin I or IL-6. Cells were harvested, washed with PBS and analyzed for proper assays.

Reverse transcription qPCR

Request a detailed protocol

Total RNA was isolated from cells using TRIzol Reagent (Invitrogen) according to the manufacturer’s protocol. Oneμg of total RNA was reverse transcribed using SuperScript III reverse transcriptase and random hexamer. One-tenth of the RT reaction was used as a template for real-time PCR using Luna Universal qPCR Master Mix (New England Biolabs) on a QuantStudi six system. Relative expression was calculated with 2−ΔΔCt using the average value of housekeeping gene GAPDH.

Chromatin immunoprecipitation

Request a detailed protocol

ChIP was performed as described previously (Whyte et al., 2013). Approximately 10 × 106 suspension cells were harvested and in 10 ml PBS with 1% formaldehyde for 10 min at room temperature, followed by adding 0.125 M glycine for 5 min. Cells were washed and pelleted by centrifugation and lysed with buffer (50 mM Tris-HCl, pH 7.5, 1% IGEPAL CA-630, 1 mM EDTA, 0.1% SDS, plus 1 mM PMSF) in the presence of protease inhibitors and incubated on ice for 30 min. Cell lysate was sonicated to shear DNA to a length of 200–600 bp. The lysates were centrifuged, and supernatant transferred to new tubes. For immunoprecipitation, approximately 2 × 106 cells and 2–3 μg of antibodies or isotype matched IgG as control were used per ChIP and incubated with supernatant at 4 °C on a rotating wheel overnight. Chromatin-antibody complexes were sequentially washed with low-salt buffer, high-salt buffer, LiCl buffer, and TE buffer. Cross-links were reversed by addition of 100 μl of 1% SDS plus 100 mM NaHCO3 and by heating at 65 °C overnight. Following phenol/chloroform/isoamyl alcohol extraction, immunoprecipitated DNA was precipitated with isopropyl alcohol and resuspended in nuclease-free water. For the identification of the specific regions of interest, ~ 10 ng of purified DNA was quantified to determine the percentage of each analyzed region, as well as a negative control region from Fullwood et al., 2009, against input DNA. The PCR primers are shown in Supplementary file 5.

Additional statistical analysis

Request a detailed protocol

Data were presented as mean ± standard deviation (SD) of three replicates unless stated otherwise. Correlation analysis was performed using Pearson’s correlation coefficient. Statistical significance was considered at two-sided P-values less than 0.05.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting file; Source Data files have been provided for Figures 2-5.

The following previously published data sets were used
    1. Scharer CD
    2. Boss JM
    (2019) NCBI Gene Expression Omnibus
    ID GSE118253. Accessible chromatin profiles of B cell subsets from healthy and SLE subjects.
    1. Scharer CD
    2. Boss J
    (2016) NCBI Gene Expression Omnibus
    ID GSE71338. Effects of biobanking on chromatin accessibility.
    1. Mistry P
    2. Nakabo S
    3. O'Neil L
    4. Goel RR
    5. Jiang K
    6. Gupta S
    7. Dell'Orso S
    8. Gutierrez-Cruz G
    9. Sun HW
    10. Kaplan MJ
    (2019) NCBI Gene Expression Omnibus
    ID GSE139359. Transcriptomic, epigenetic and functional analyses implicate neutrophil diversity in the pathogenesis of systemic lupus erythematosus [ATAC-seq].
    1. Caielli S
    2. Veiga DF
    3. Domic B
    4. Murat E
    5. Banchereau R
    6. Xu Z
    7. Chandra M
    8. Athale S
    9. Chung C
    10. Walters L
    11. Baisch J
    12. Wright T
    13. Punaro M
    14. Ucar D
    15. Ueno H
    16. Zhou J
    17. Banchereau J
    18. Pascual V
    (2018) NCBI Gene Expression Omnibus
    ID GSE110017. A novel CD4 T cell population expanded in SLE blood provides B cell help through IL10 and succinate [ATAC-seq].
    1. Scharer CD
    2. Boss JM
    (2019) NCBI Gene Expression Omnibus
    ID GSE118254. Transcriptome profiles of B cell subsets from healthy and SLE subjects.
    1. Sanz I
    2. Jenks S
    3. Marigorta UM
    (2018) NCBI Gene Expression Omnibus
    ID GSE92387. Gene expresison studies of lupus and healthy B cell subsets through RNA sequencing.
    1. Scharer CD
    2. Boss JM
    (2019) NCBI Gene Expression Omnibus
    ID GSE118255. DNA methylation profiles profiles of B cell subsets from healthy and SLE subjects.
    1. Rao S
    2. Huntley M
    3. Lieberman Aiden E
    (2014) NCBI Gene Expression Omnibus
    ID GSE63525. A three-dimensional map of the human genome at kilobase resolution reveals prinicples of chromatin looping.
    1. Wells AD
    2. Chesi A
    3. Manduchi E
    4. Johnson ME
    5. Leonard ME
    6. Romberg ND
    7. Grant SFA
    8. Lu S
    (2020) ArrayExpress
    ID E-MTAB-6621. Promoter capture-C of primary human T Follicular Helper (TFH) cells and naive CD4-positive helper T cells from tonsils of healthy volunteers.
    1. Disease-Specific (Systemic lupus erythematosus, NPU, MDS, RD)
    (2017) NCBI dbGAP
    ID phs001025.v1. GWAS in an Amerindian Ancestry Population Reveals Novel Systemic Lupus Erythematosus Risk Loci and the Role of European Admixture.

References

    1. Blaskovich MA
    2. Sun J
    3. Cantor A
    4. Turkson J
    5. Jove R
    6. Sebti SM
    (2003)
    Discovery of JSI-124 (cucurbitacin I), a selective Janus kinase/signal transducer and activator of transcription 3 signaling pathway inhibitor with potent antitumor activity against human and murine cancer cells in mice
    Cancer Research 63:1270–1279.
    1. Laboratory DA
    (2017)
    Coordinating Center -Analysis Working G, Statistical Methods groups-Analysis Working Genetic effects on gene expression across human tissues
    Nature 550:204–213.
  1. Book
    1. Madoux F
    2. Koenig M
    3. Nelson E
    4. Chowdhury S
    5. Cameron M
    6. Mercer B
    (2010)
    Modulators of STAT Transcription Factors for the Targeted Therapy of Cancer (STAT3 Activators)
    National Center for Biotechnology Information (US).
    1. Shi J
    2. Zhang Y
    3. Zheng W
    4. Michailidou K
    5. Ghoussaini M
    6. Bolla MK
    7. Wang Q
    8. Dennis J
    9. Lush M
    10. Milne RL
    11. Shu X-O
    12. Beesley J
    13. Kar S
    14. Andrulis IL
    15. Anton-Culver H
    16. Arndt V
    17. Beckmann MW
    18. Zhao Z
    19. Guo X
    20. Benitez J
    21. Beeghly-Fadiel A
    22. Blot W
    23. Bogdanova NV
    24. Bojesen SE
    25. Brauch H
    26. Brenner H
    27. Brinton L
    28. Broeks A
    29. Brüning T
    30. Burwinkel B
    31. Cai H
    32. Canisius S
    33. Chang-Claude J
    34. Choi J-Y
    35. Couch FJ
    36. Cox A
    37. Cross SS
    38. Czene K
    39. Darabi H
    40. Devilee P
    41. Droit A
    42. Dork T
    43. Fasching PA
    44. Fletcher O
    45. Flyger H
    46. Fostira F
    47. Gaborieau V
    48. García-Closas M
    49. Giles GG
    50. Mervi Grip
    51. Guenel P
    52. Haiman CA
    53. Hamann U
    54. Hartman M
    55. Miao H
    56. Hollestelle A
    57. Hopper JL
    58. Hsiung C-N
    59. kConFab Investigators
    60. Ito H
    61. Jakubowska A
    62. Johnson N
    63. Torres D
    64. Kabisch M
    65. Kang D
    66. Khan S
    67. Knight JA
    68. Kosma V-M
    69. Lambrechts D
    70. Li J
    71. Lindblom A
    72. Lophatananon A
    73. Lubinski J
    74. Mannermaa A
    75. Manoukian S
    76. Le Marchand L
    77. Margolin S
    78. Marme F
    79. Matsuo K
    80. McLean C
    81. Meindl A
    82. Muir K
    83. Neuhausen SL
    84. Nevanlinna H
    85. Nord S
    86. Børresen-Dale A-L
    87. Olson JE
    88. Orr N
    89. van den Ouweland AMW
    90. Peterlongo P
    91. Putti TC
    92. Rudolph A
    93. Sangrajrang S
    94. Sawyer EJ
    95. Schmidt MK
    96. Schmutzler RK
    97. Shen C-Y
    98. Hou M-F
    99. Shrubsole MJ
    100. Southey MC
    101. Swerdlow A
    102. Teo SH
    103. Thienpont B
    104. Toland AE
    105. Tollenaar RAEM
    106. Tomlinson I
    107. Truong T
    108. Tseng C-C
    109. Wen W
    110. Winqvist R
    111. Wu AH
    112. Yip CH
    113. Zamora PM
    114. Zheng Y
    115. Floris G
    116. Cheng C-Y
    117. Hooning MJ
    118. Martens JWM
    119. Seynaeve C
    120. Kristensen VN
    121. Hall P
    122. Pharoah PDP
    123. Simard J
    124. Chenevix-Trench G
    125. Dunning AM
    126. Antoniou AC
    127. Easton DF
    128. Cai Q
    129. Long J
    (2016) Fine-scale mapping of 8q24 locus identifies multiple independent risk variants for breast cancer
    International Journal of Cancer 139:1303–1317.
    https://doi.org/10.1002/ijc.30150

Decision letter

  1. Xingyi Guo
    Reviewing Editor; Division of Epidemiology, Department of Medicine, Vanderbilt Epidemiology Center, and Vanderbilt-Ingram Cancer Center, Vanderbilt University School of Medicine, United States
  2. Betty Diamond
    Senior Editor; The Feinstein Institute for Medical Research, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "STAT3-mediated allelic imbalance of novel genetic variant rs1047643 and B cell specific super-enhancer in association with systemic lupus erythematosus" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Betty Diamond as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) The authors should explore large-scale GWAS data or provide any other statistical evidence to strengthen their discovery of the risk association signal of rs1047643 (such as summary statistics of GWAS data of European descendants).

2) To improve and clarify the statical analyses as suggested by reviewers.

Reviewer #1 (Recommendations for the authors):

This is a well conducted study aiming to identify novel risk variants for systemic lupus erythematosus risk and to further explore potential mechanisms through a comprehensive analysis of functional genomic analyses and functional assays. The authors have identified a putative novel risk variant, rs1047643, for the disease risk, and further revealed underlying disease mechanisms of a known susceptibility factor, STAT3, mediated Super enhancer activity and cis-regulatory effects of SNP rs1047643. Their discovered mechanisms are novel, providing new insights into the potential mechanism underlying other GWAS-identified variants located in the Super enhancer element to mediate the disease risk. This work could be very important to advance the understanding of the genetic susceptibility and biology of systemic lupus erythematosus risk. I hope my comments below could help authors improve their work in their revision.

1) The authors could try to explore large-scale GWAS data to strengthen their discovery of the risk association signal of rs1047643 (such as summary statistics of GWAS data of European descendant). The authors should present the information of Linkage Disequilibrium between the identified variant and previously reported GWAS-identified risk variants. The novel risk variant, rs1047643, may not be an independent risk signal (may be in LD with rs17807624).

2) Need to clarify some statistical analyses. For example, for formula presented in the line 121, it's unclear for the "disease" variable (referring to case/control status? how to link SNP information?). If referring to case/control status, it could be a response variable? The authors need to provide detailed description about how to select the risk variant rs1047643 for your focus (only this SNP satisfied your analysis a two-stage study presented in Figure 1?).

In the line 132, not sure why populations is needed to be adjusted here (only focusing on Hispanic population). The authors need to include sentence to describe about how to perform additional analysis and how to perform "AI of RNA transcripts" (the line 233/234). "Statistical analysis" section may not be necessary (including above analysis).

3) Some limitations to be included in the Discussion sections: some functional data may not be generated from the disease target cells (i.e. Bcells). The authors need to discuss some of their allelic imbalance findings about other GWAS-identified risk variants at the locus, 8p23.

4) Need to fix some typos and other minor issues. In the abstract, "SLE "- need to spell out; "Open chromatin" may change to "chromatin accessibility". A typo 'tying' (line 60); RRBS (line 87) and "SD" (line 202) need to be spelled out. The sentence about "The differences" is unclear (line 204)

Reviewer #2 (Recommendations for the authors):

Page 7, line 121. Formula looks awkward. An equivalent way to write this model may be Disease ~ \α + \β Allelic ratio +\epsilon. However, after formatting the model into this shape, it appears that a logistic regression for binary trait should be used instead.

Figure 2A: For your Allelic Imbalance analysis, how many SNPs have you tested in order to identify rs1047643? Should we conduct multiple test correction? I can't see any descriptions from line 108 to 124. If you have done this already (hence the p-value listed in the figure is the after-correction value), would you be clear? What is the meaning of "summary" in the last row?

Page 12 -13, the paragraph for association analysis indeed mentioned an "adjustment". I am a bit confused here: (1) If your analysis starts with Allelic Imbalance analysis, then the next follow-up association test doesn't need a multiple-test adjustment. (2) based on my question above, would you clarify what "adjustment" you are talking about and what is the mathematical procedure? In the event you did some rigorous multiple-test correction here, but not in the AI analysis, your flow of the discovery might be reversely presented.

Page 14, Line 276-284. The Hi-C loop looks interesting, however little details are provided here. E.g., what is the sequencing coverage evidencing this loop?

Figure 4 indeed shows the difference between SLE and controls. However, to make the discovery more convincing, would you check how significant is the difference with the background distribution? E.g., randomly select a hundred other sites and see whether there are also difference between SLE and controls? Or the background looks similar, putting the focal discovery in a significant (or at least promising) position? Similar concern may also apply to Figures 5 and 6: it may be nice to see how likely such difference will show up in randomly selected candidates.

Reviewer #3 (Recommendations for the authors):

Overall, the report by Jin Men et al. provided supportive evidence that by mining multi-omics profiles to develop and indicate a novel functional variant, the mechanism of STAT3 medication. The analytical and statistical framework is appropriate, although not fully described. In addition, additional experiments should be performed.

1. From the results presented (line 224), it is not clear the result of significance for other SNPs in the comparison. Because it has no comprehensive results, which looked like a little confusing. It's necessary to provide the list of potential significant SNPs.

2. The authors report that 'statistical significance' (line 244) about association of rs1047643 with SLE risk, and adjusted P should be 0.023 according to Figure 3. And next the choice of two SNPs for haplotype analysis, the description of the results is inadequate.

3. In Figure 6, the authors performed chip assay to confirm stat3 binding to SE and rs1047643 regions. In addition, they used the stat3 DNA binding inhibitor. However, the inhibitor didn't affect stat3 DNA binding efficiency too much. Please explain why. It's necessary to perform time-course and concentration-dependent experiments to confirm the binding specificity. Also, the authors should perform chip-PCR assay using non-specific primers. Furthermore, the authors should design the experiments to confirm whether rs1047643 affects stat3 binding.

4. The results of Figure 6b,c is confused.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "STAT3-mediated allelic imbalance of novel genetic variant rs1047643 and B cell specific super-enhancer in association with systemic lupus erythematosus" for further consideration by eLife. Your revised article has been evaluated by Betty Diamond (Senior Editor), a Reviewing Editor, and the original reviewers.

The manuscript has been improved but there are some remaining issues that need to be addressed: please provide a response to the comments of Reviewer #2.

Reviewer #1 (Recommendations for the authors):

The authors have adequately addressed all of my concerns/comments.

Reviewer #2 (Recommendations for the authors):

In the revised manuscript, most of my comments have been satisfactorily addressed, with the following exceptions:

Q1: you have not responded to my comments on why putting the disease status on the right-hand side of the model.

Q2: disagree that one doesn't have to conduct multiple test adjustment simply due to the sample size being small.

Reviewer #3 (Recommendations for the authors):

I am satisfied with the revised manuscript and suggest the manuscript to be accepted for publication.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "STAT3-mediated allelic imbalance of novel genetic variant rs1047643 and B cell specific super-enhancer in association with systemic lupus erythematosus" for further consideration by eLife. Your revised article has been evaluated by Betty Diamond (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

Please address reviewer's Q1: "you have not responded to my comments on why putting the disease status on the right-hand side of the model"

The editor agrees with the validity of the reviewer's comments and would suggest using a logistic regression with disease status as a dependent variable.

https://doi.org/10.7554/eLife.72837.sa1

Author response

Essential revisions:

1) The authors should explore large-scale GWAS data or provide any other statistical evidence to strengthen their discovery of the risk association signal of rs1047643 (such as summary statistics of GWAS data of European descendants).

2) To improve and clarify the statical analyses as suggested by reviewers.

We thank two editors for giving us an opportunity to submit the revised paper. In the revised version of manuscript, we have made a significant improvement to address these two essential comments (Page 21, and Methods section), and have now met the high standard of your journal for publication. Please see below are the point-by-point response to the reviewers’ comments.

Reviewer #1 (Recommendations for the authors):

This is a well conducted study aiming to identify novel risk variants for systemic lupus erythematosus risk and to further explore potential mechanisms through a comprehensive analysis of functional genomic analyses and functional assays. The authors have identified a putative novel risk variant, rs1047643, for the disease risk, and further revealed underlying disease mechanisms of a known susceptibility factor, STAT3, mediated Super enhancer activity and cis-regulatory effects of SNP rs1047643. Their discovered mechanisms are novel, providing new insights into the potential mechanism underlying other GWAS-identified variants located in the Super enhancer element to mediate the disease risk. This work could be very important to advance the understanding of the genetic susceptibility and biology of systemic lupus erythematosus risk. I hope my comments below could help authors improve their work in their revision.

1) The authors could try to explore large-scale GWAS data to strengthen their discovery of the risk association signal of rs1047643 (such as summary statistics of GWAS data of European descendant). The authors should present the information of Linkage Disequilibrium between the identified variant and previously reported GWAS-identified risk variants. The novel risk variant, rs1047643, may not be an independent risk signal (may be in LD with rs17807624).

We added these results in the revised manuscript (Page 21, page 7: lines 123-130).

2) Need to clarify some statistical analyses. For example, for formula presented in the line 121, it's unclear for the "disease" variable (referring to case/control status? how to link SNP information?). If referring to case/control status, it could be a response variable? The authors need to provide detailed description about how to select the risk variant rs1047643 for your focus (only this SNP satisfied your analysis a two-stage study presented in Figure 1?).

In the line 132, not sure why populations is needed to be adjusted here (only focusing on Hispanic population). The authors need to include sentence to describe about how to perform additional analysis and how to perform "AI of RNA transcripts" (the line 233/234). "Statistical analysis" section may not be necessary (including above analysis).

In the revised paper, we provide detailed description about AI analysis implemented in a linear model (Page 20-21), and re-clarify the adjustment of five countries of the Hispanic population (Page 21: lines 396-400), as well as the method section (Page 21).

3) Some limitations to be included in the Discussion sections: some functional data may not be generated from the disease target cells (i.e. Bcells). The authors need to discuss some of their allelic imbalance findings about other GWAS-identified risk variants at the locus, 8p23.

Thanks for reviewer’s suggestion. In the revised paper, we discussed the limitations (Page 16, lines 329-332) and the allelic imbalance results of 13 GWAS SNPs (Page 15, lines 307-308).

4) Need to fix some typos and other minor issues. In the abstract, "SLE "- need to spell out; "Open chromatin" may change to "chromatin accessibility". A typo 'tying' (line 60); RRBS (line 87) and "SD" (line 202) need to be spelled out. The sentence about "The differences" is unclear (line 204)

Thank reviewer for pointing out these issues. We corrected and defined these terms in the revised paper, except for ‘tying’ that is correct, not a typo.

Reviewer #2 (Recommendations for the authors):

Page 7, line 121. Formula looks awkward. An equivalent way to write this model may be Disease ~ \α + \β Allelic ratio +\epsilon. However, after formatting the model into this shape, it appears that a logistic regression for binary trait should be used instead.

In the revised paper, we rephrased the description about AI analysis implemented in a linear model (Page 20, lines 379-385)

Figure 2A: For your Allelic Imbalance analysis, how many SNPs have you tested in order to identify rs1047643? Should we conduct multiple test correction? I can't see any descriptions from line 108 to 124. If you have done this already (hence the p-value listed in the figure is the after-correction value), would you be clear? What is the meaning of "summary" in the last row?

In the revised paper, we provided more detailed description and re-clarified the statistical analyses (Page 6, lines 93-102).

Page 12 -13, the paragraph for association analysis indeed mentioned an "adjustment". I am a bit confused here: (1) If your analysis starts with Allelic Imbalance analysis, then the next follow-up association test doesn't need a multiple-test adjustment. (2) based on my question above, would you clarify what "adjustment" you are talking about and what is the mathematical procedure? In the event you did some rigorous multiple-test correction here, but not in the AI analysis, your flow of the discovery might be reversely presented.

Thanks for reviewer’s helpful comments. In the revised paper, we provided detailed description and re-clarified the statistical analyses (Page 34, lines 682-684, and Figure 3A-B).

Page 14, Line 276-284. The Hi-C loop looks interesting, however little details are provided here. E.g., what is the sequencing coverage evidencing this loop?

In the revised paper, we provided more details about read supports and statistical significance for DNA loops in the Methods section (Page 23, lines 443-444).

Figure 4 indeed shows the difference between SLE and controls. However, to make the discovery more convincing, would you check how significant is the difference with the background distribution? E.g., randomly select a hundred other sites and see whether there are also difference between SLE and controls? Or the background looks similar, putting the focal discovery in a significant (or at least promising) position? Similar concern may also apply to Figures 5 and 6: it may be nice to see how likely such difference will show up in randomly selected candidates.

In the revised paper, we presented these results (Page 9-10, lines 178-180; Page 11, lines 208-212, and Figure 4—figure supplement 1and Figure 5—figure supplement 1).

Reviewer #3 (Recommendations for the authors):

Overall, the report by Jin Men et al. provided supportive evidence that by mining multi-omics profiles to develop and indicate a novel functional variant, the mechanism of STAT3 medication. The analytical and statistical framework is appropriate, although not fully described. In addition, additional experiments should be performed.

1. From the results presented (line 224), it is not clear the result of significance for other SNPs in the comparison. Because it has no comprehensive results, which looked like a little confusing. It's necessary to provide the list of potential significant SNPs.

Because of small sample size, we just found three potential significant SNPs. In the revised paper, we reported these SNPs (Page 6, lines 93-102, and Figure 2—figure supplement 1).

2. The authors report that 'statistical significance' (line 244) about association of rs1047643 with SLE risk, and adjusted P should be 0.023 according to Figure 3. And next the choice of two SNPs for haplotype analysis, the description of the results is inadequate.

In the revised paper, we provided detailed description and re-clarified the statistical analyses (Page 7, lines 120-130).

3. In Figure 6, the authors performed chip assay to confirm stat3 binding to SE and rs1047643 regions. In addition, they used the stat3 DNA binding inhibitor. However, the inhibitor didn't affect stat3 DNA binding efficiency too much. Please explain why. It's necessary to perform time-course and concentration-dependent experiments to confirm the binding specificity. Also, the authors should perform chip-PCR assay using non-specific primers. Furthermore, the authors should design the experiments to confirm whether rs1047643 affects stat3 binding.

Thanks for reviewer’s insightful comments. Please see below are detailed description.

For the question about “the authors should perform chip-PCR assay using non-specific primers”, we designed a pair of primers as negative control which has been previously used elsewhere as ChIP-qPCR quality control (Page 25, lines 474-476). We presented these new ChIP-qPCR results in the revised manuscript (Page 12, line 228, and Figure 6—figure supplement 1).

For the question about “However, the inhibitor didn't affect stat3 DNA binding efficiency too much”, we did more experiment by increasing S3I concentration to 100 μM. The results don’t change dramatically (Figure 6C). One reason may be that the rs1047643 region is co-bound by multiple STAT family members. Because in our latest experiment where we also did STAT1 ChIP assay under S3I treatment in GM11997 cells, to our surprise is that we observed an increase of STAT1 binding on rs1047643-T allele under the S3I treatment. That is to say, as classic JAK-STAT signaling pathway shows, there may exist three different sorts of dimers (two homo-dimers, STAT1-STAT1 and STAT3-STAT3, one hetero-dimer, STAT1-STAT3) that co-regulate this locus in a complicated manner. Therefore, inhibition of STAT3 may result in a mild effect on the rs1047643 allelic binding.

For the remaining questions, we would like to logically depict our study design on this project. At the beginning, our strategy was focused on loss-of-function of STAT3 by using siRNA and shRNA methods. However, due to the extremely low transfection efficiency (<5%) in such tough suspension cells, the experiments failed. Meanwhile, we attempted to test the impact of the rs1047643 on the STAT3 binding by transfection-associated assays, but were unsuccessful, presumably due to the difficulty of transfection in lymphoblastic cells. Eventually, we changed our strategies and switched to use STAT3 drug inhibitors and activators to determine the allelic binding performance.

In addition, we have expanded this project and acquired several novel interesting findings during the last nine months. In the revised manuscript, we present the relevant results to confirm the results by controlling upstream factors in the JAK-STAT3 signaling pathway (Page 12-13, lines 244-249, and Figure 6—figure supplement 3).

4. The results of Figure 6b,c is confused.

Thanks for reviewer’s helpful comment. To make the results more readable and understandable, we re-organized the Figure 6 and split into two Figures (Figure 6 and 7) in the revised manuscript.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed: please provide a response to the comments of Reviewer #2.

Reviewer #2 (Recommendations for the authors):

In the revised manuscript, most of my comments have been satisfactorily addressed, with the following exceptions:

Q1: you have not responded to my comments on why putting the disease status on the right-hand side of the model.

To test for associations between allelic ratio and SLE disease state (case/control comparison) for each heterozygous SNP, we regarded the case/control disease status as categorical variable (control and case are coded as 0 and 1, respectively) and put it on the right-hand side implemented in regression model, the analysis has been used in other studies (1, 2). In the revised manuscript, we re-phrased the regression analysis (Page 20, lines 382-389).

Q2: disagree that one doesn't have to conduct multiple test adjustment simply due to the sample size being small.

Thanks for reviewer’s suggestion. In the revised manuscript, we re-phrased this sentence (Page 6, line 95).

1. Adoue V, Schiavi A, Light N, Almlof JC, Lundmark P, Ge B, et al. Allelic expression mapping across cellular lineages to establish impact of non-coding SNPs. Mol Syst Biol. 2014;10:754.

2. Absher DM, Li X, Waite LL, Gibson A, Roberts K, Edberg J, et al. Genome-wide DNA methylation analysis of systemic lupus erythematosus reveals persistent hypomethylation of interferon genes and compositional changes to CD4+ T-cell populations. PLoS Genet. 2013;9(8):e1003678.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Please address reviewer's Q1: "you have not responded to my comments on why putting the disease status on the right-hand side of the model"

We re-analyzed the data based on the logistic regression with disease status as a dependent variable. Due to small sample size, particularly just six samples carrying heterozygous SNP rs1047643 in one dataset (GSE71338), there is no statistical power to identify the significance when analyzing with the logistic regression model. In the revised manuscript, we discuss this limitation (Page 17, lines 334-340).

https://doi.org/10.7554/eLife.72837.sa2

Article and author information

Author details

  1. Yanfeng Zhang

    HudsonAlpha Institute for Biotechnology, Huntsville, United States
    Present address
    Department of Population and Data Sciences, University of Texas Southwestern Medical Center, Dallas, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing – original draft, Writing – review and editing
    For correspondence
    yanfengzhang1984@outlook.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3859-3839
  2. Kenneth Day

    Zymo Research Corp, Irvine, United States
    Contribution
    Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Devin M Absher

    HudsonAlpha Institute for Biotechnology, Huntsville, United States
    Contribution
    Data curation, Funding acquisition, Investigation, Resources, Supervision, Writing – review and editing
    For correspondence
    dabsher@hudsonalpha.org
    Competing interests
    No competing interests declared

Funding

HudsonAlpha Institute for biotechnology funds

  • Yanfeng Zhang
  • Devin M Absher

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Le Su for insightful suggestions, technical discussions and critical reading of the manuscript. This study was supported by the HudsonAlpha institutional funds. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Funding This study was supported by the HudsonAlpha Institute Fund. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Senior Editor

  1. Betty Diamond, The Feinstein Institute for Medical Research, United States

Reviewing Editor

  1. Xingyi Guo, Division of Epidemiology, Department of Medicine, Vanderbilt Epidemiology Center, and Vanderbilt-Ingram Cancer Center, Vanderbilt University School of Medicine, United States

Publication history

  1. Received: August 6, 2021
  2. Preprint posted: September 20, 2021 (view preprint)
  3. Accepted: February 18, 2022
  4. Accepted Manuscript published: February 21, 2022 (version 1)
  5. Version of Record published: February 28, 2022 (version 2)

Copyright

© 2022, Zhang et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 354
    Page views
  • 46
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Yanfeng Zhang
  2. Kenneth Day
  3. Devin M Absher
(2022)
STAT3-mediated allelic imbalance of novel genetic variant Rs1047643 and B-cell-specific super-enhancer in association with systemic lupus erythematosus
eLife 11:e72837.
https://doi.org/10.7554/eLife.72837
  1. Further reading

Further reading

    1. Cancer Biology
    2. Computational and Systems Biology
    Pan Cheng, Xin Zhao ... Teresa Davoli
    Research Article

    How cells control gene expression is a fundamental question. The relative contribution of protein-level and RNA-level regulation to this process remains unclear. Here, we perform a proteogenomic analysis of tumors and untransformed cells containing somatic copy number alterations (SCNAs). By revealing how cells regulate RNA and protein abundances of genes with SCNAs, we provide insights into the rules of gene regulation. Protein complex genes have a strong protein-level regulation while non-complex genes have a strong RNA-level regulation. Notable exceptions are plasma membrane protein complex genes, which show a weak protein-level regulation and a stronger RNA-level regulation. Strikingly, we find a strong negative association between the degree of RNA-level and protein-level regulation across genes and cellular pathways. Moreover, genes participating in the same pathway show a similar degree of RNA- and protein-level regulation. Pathways including translation, splicing, RNA processing, and mitochondrial function show a stronger protein-level regulation while cell adhesion and migration pathways show a stronger RNA-level regulation. These results suggest that the evolution of gene regulation is shaped by functional constraints and that many cellular pathways tend to evolve one predominant mechanism of gene regulation at the protein level or at the RNA level.

    1. Computational and Systems Biology
    2. Neuroscience
    Janus RL Kobbersmed, Manon MM Berns ... Alexander M Walter
    Research Article Updated

    Synaptic communication relies on the fusion of synaptic vesicles with the plasma membrane, which leads to neurotransmitter release. This exocytosis is triggered by brief and local elevations of intracellular Ca2+ with remarkably high sensitivity. How this is molecularly achieved is unknown. While synaptotagmins confer the Ca2+ sensitivity of neurotransmitter exocytosis, biochemical measurements reported Ca2+ affinities too low to account for synaptic function. However, synaptotagmin’s Ca2+ affinity increases upon binding the plasma membrane phospholipid PI(4,5)P2 and, vice versa, Ca2+ binding increases synaptotagmin’s PI(4,5)P2 affinity, indicating a stabilization of the Ca2+/PI(4,5)P2 dual-bound state. Here, we devise a molecular exocytosis model based on this positive allosteric stabilization and the assumptions that (1.) synaptotagmin Ca2+/PI(4,5)P2 dual binding lowers the energy barrier for vesicle fusion and that (2.) the effect of multiple synaptotagmins on the energy barrier is additive. The model, which relies on biochemically measured Ca2+/PI(4,5)P2 affinities and protein copy numbers, reproduced the steep Ca2+ dependency of neurotransmitter release. Our results indicate that each synaptotagmin engaging in Ca2+/PI(4,5)P2 dual-binding lowers the energy barrier for vesicle fusion by ~5 kBT and that allosteric stabilization of this state enables the synchronized engagement of several (typically three) synaptotagmins for fast exocytosis. Furthermore, we show that mutations altering synaptotagmin’s allosteric properties may show dominant-negative effects, even though synaptotagmins act independently on the energy barrier, and that dynamic changes of local PI(4,5)P2 (e.g. upon vesicle movement) dramatically impact synaptic responses. We conclude that allosterically stabilized Ca2+/PI(4,5)P2 dual binding enables synaptotagmins to exert their coordinated function in neurotransmission.