Association of two independent functional risk haplotypes in TNIP1 with systemic lupus erythematosus




Systemic lupus erythematosus (SLE) is an autoimmune disease characterized by autoantibody production and altered type I interferon expression. Genetic surveys and genome-wide association studies have identified >30 SLE susceptibility genes. One of these genes, TNIP1, encodes the ABIN1 protein. ABIN1 functions in the immune system by restricting NF-κB signaling. The present study was undertaken to investigate the genetic factors that influence association with SLE in genes that regulate the NF-κB pathway.


We analyzed a dense set of genetic markers spanning TNIP1 and TAX1BP1, as well as the TNIP1 homolog TNIP2, in case–control populations of diverse ethnic origins. TNIP1, TNIP2, and TAX1BP1 were fine-mapped in a total of 8,372 SLE cases and 7,492 healthy controls from European-ancestry, African American, Hispanic, East Asian, and African American Gullah populations. Levels of TNIP1 messenger RNA (mRNA) and ABIN1 protein in Epstein-Barr virus–transformed human B cell lines were analyzed by quantitative reverse transcription–polymerase chain reaction and Western blotting, respectively.


We found significant associations between SLE and genetic variants within TNIP1, but not in TNIP2 or TAX1BP1. After resequencing and imputation, we identified 2 independent risk haplotypes within TNIP1 in individuals of European ancestry that were also present in African American and Hispanic populations. Levels of TNIP1 mRNA and ABIN1 protein were reduced among subjects with these haplotypes, suggesting that they harbor hypomorphic functional variants that influence susceptibility to SLE by restricting ABIN1 expression.


Our results confirm the association signals between SLE and TNIP1 variants in multiple populations and provide new insight into the mechanism by which TNIP1 variants may contribute to SLE pathogenesis.

Members of the NF-κB family of transcription factors are key mediators of innate and adaptive immune responses. A diverse array of surface receptors including tumor necrosis factor α (TNFα) and Toll-like receptors (TLRs) converge on NF-κB (1, 2), and precise control of NF-κB is therefore needed to effectively interpret and transmit these signals in order to produce an effective defense against invading pathogens and viruses. The ubiquitin editing enzyme A20, encoded by TNFα-induced protein 3 (gene TNFAIP3; OMIM ID 191163), is a critical negative regulator of NF-κB (3, 4). Termination of NF-κB signaling by A20 leverages adapter proteins such as Tax1 (human T cell leukemia virus type I) binding protein 1 (gene TAX1BP1; OMIM ID 605326), and the A20 binding inhibitor of NF-κB1 (ABIN1), which facilitate the interaction of A20 with target molecules (5, 6). Breakdown of this system leads to unrestrained NF-κB transactivation that can result in autoimmune disease, sepsis, and/or malignancy (7–10).

Systemic lupus erythematosus (SLE; OMIM ID 152700) is an autoimmune disease that exhibits a robust but complex genetic architecture. Candidate gene and genome-wide association studies have identified >30 genetic loci that are convincingly associated with SLE (11–13). Among these are TNFAIP3 and TNFAIP3 interacting protein 1 (TNIP1; OMIM ID 607714), which encodes ABIN1, emphasizing the importance of these genes in restricting autoimmunity. Variants in the vicinity of TNFAIP3 are associated with multiple autoimmune diseases in several different ethnic populations (14–16). Moreover, multiple independent genetic effects associated with SLE appear to be operating in the region (17–19). Our group recently identified a functional polymorphism in a regulatory element ∼25 kb telomeric of the TNFAIP3 coding region that can explain the association signal of one of these independent effects (20).

Variants in the region of TNIP1 are also associated with multiple autoimmune diseases, including psoriasis (OMIM ID 177900) (21), psoriatic arthritis (OMIM ID 607507) (22), systemic sclerosis (OMIM ID 181750) (23), and SLE (24, 25). However, in contrast to TNFAIP3, less is known about the genetic architecture of TNIP1. For instance, association with TNIP1 has only been evaluated in SLE cases of European and Asian ancestry (24, 25), and thus it is not known if TNIP1 is a risk locus in African American and Hispanic populations. Furthermore, TNIP1 has not been thoroughly fine-mapped to determine the number of risk effects present in the region, nor have any functional mechanisms been attributed to TNIP1-associated risk haplotypes.

In order to gain a more comprehensive understanding of the TNIP1 locus in SLE, we performed a genetic fine-mapping study in 5 ethnically diverse SLE case–control populations. We also included single-nucleotide polymorphisms (SNPs) in TAX1BP1, an adapter molecule for A20 (5) located adjacent to the previously described SLE susceptibility gene JAZF zinc finger 1 (gene JAZF1; OMIM ID 606246) and in the TNIP1 homolog, TNFAIP3-interacting protein 2 (gene TNIP2; OMIM ID 610669). Our results demonstrate a complex genetic architecture within the TNIP1 locus that is shared, in part, across multiple ethnic populations. We identified 2 independent functional risk haplotypes that result in decreased expression of TNIP1 messenger RNA (mRNA) and ABIN1 protein, providing insight into the mechanism by which variants in TNIP1 may contribute to SLE pathogenesis.



The following independent groups of cases and controls were included in the study: European ancestry (4,248 cases and 3,818 controls), African American (1,569 cases and 1,893 controls), Hispanic enriched for American Indian/European admixture (1,622 cases and 887 controls), East Asian (1,328 cases and 1,348 controls), and African American Gullah (155 cases and 131 controls). The majority of the subjects in the East Asian population were from Korea (906 cases and 1,012 controls) but this cohort also included Chinese, Japanese, Taiwanese, and Singaporean individuals. The African American Gullah population is a group of African Americans with low genetic admixture who live in the Sea Islands of the Carolinas, whose ancestry originates from Sierra Leone. All cases met the 1997 American College of Rheumatology revised criteria for the classification of SLE (26). Samples were supplied from multiple institutions with approval from their respective institutional review boards (IRBs); consent forms were obtained at each location under IRB guidelines. Samples were then assembled at the Oklahoma Medical Research Foundation (OMRF) under the approval of the OMRF IRB. Only individuals for whom there were signed informed consent forms were included in the study.

Genotyping and quality control.

Genotyping for 88 SNPs within and flanking TNIP1 on chromosome 5q33, 22 SNPs within and flanking TAX1BP1 on chromosome 7p15, and 52 SNPs within and flanking TNIP2 on chromosome 4p16, as well as for 347 genome-wide ancestry-informative markers (AIMs) (27, 28) was performed using the Illumina iSelect platform at OMRF. Details on the 347 AIMs are available at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf.

For a SNP to be included, we required that it have well-defined cluster scatter plots, a call rate of >90%, a minor allele frequency of >0.001, and a Hardy-Weinberg proportion test P value in controls of >0.001. We excluded AIMs that had low call rates (<90%) or low minor allele frequencies (<0.001) or were in linkage disequilibrium (LD) (r2 > 0.2) with each other. The Hardy-Weinberg proportion test was not performed, to avoid removing AIMs due to monomorphic states in one of the populations. A total of 1,135 samples were excluded because they were sample heterozygosity outliers (>5 standard deviations from the mean), were extreme population outliers (based on global ancestry estimation and principal components analysis), were sample duplicates (proportion of alleles with shared identity by descent >0.4), had a low call rate (<90%), or had discrepancies between reported sex and genetic data. The numbers of excluded samples by population group and reason for exclusion are shown at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf. Using 262 AIMs, principal components analysis (29) (calculated using the R program) and global ancestry (estimated using AdmixMap [30,31]) were used to identify population outliers (>4 standard deviations from the mean of principal components 1 and 2 with ancestral allele frequencies from African, European, American Indian, and East Asian populations) and estimate percent ancestry for inclusion in association analysis as an adjustment for population substructure. After applying sample and SNP quality control measures, the final data set comprised the samples and SNPs shown in Table 1 and at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf.

Table 1. Summary of the population sample following quality control adjustments
PopulationNo. of samplesCasesControlsMaleFemale
  • *

    Enriched for American Indian/European admixture.

European ancestry7,4273,9363,4911,4955,932
African American3,3381,5271,8116952,643
East Asian2,5251,2651,2602532,272
African American Gullah27515212333242

Association analyses.

Single-marker associations were assessed using the logistic regression function in Plink, version 1.07 (32) and R, version 2.12.0, assuming an additive model and adjusting for sex and either global ancestry (African, European, and East Asian) or the first 3 principal components, with no observable difference. We set a stringent Bonferroni corrected P value threshold of <3.21 × 10−4 based on multiple tests of 156 genotyped SNPs (0.05/156). The association results were plotted using LocusZoom (33).

Resequencing, variant detection, and quality control.

Genomic DNA (3–5 μg) from 296 European, 41 African American, and 40 Hispanic individuals enriched for known SLE risk haplotypes was sheared and prepared for sequencing using an Illumina Paired-End Genomic DNA Sample Prep Kit. Region-of-interest enrichment was performed using a SureSelect Target Enrichment System (Agilent) with a custom-designed bait pool. Resequencing and generation of fastq sequencing reads were performed on the Illumina GAIIx platform with Illumina Pipeline software, version 1.7, using standard procedures.

We removed duplicate reads using a custom script, followed by alignment to the Human Reference Genome build hg19 using BWA alignment software, version 0.5.9 (34). Realignment of reads around insertion/deletion sites and problematic areas, base quality score recalibration, and variation detection were performed using the GATK software suite, version 1.0 (35, 36). We excluded variants with a quality score of <30, a quality by depth score of <5, inclusion within a homopolymer run of 5 or more bases, or a strand bias score of >−0.1, as well as variants clustered within 10 basepairs. All samples were sequenced to minimum average fold coverage of 25×. We compared sequence-based variant calls with SNPs previously genotyped on the Illumina iSelect platform and found >99% concordance between platforms. The program Beagle, version 3.3 (37) was used to determine variant phase. Plink, version 1.07 and Impute2 (38) format files were generated using the vcftools software suite, version 0.1.3 (39).

Imputation and haplotype analyses.

Imputation was performed over a 500-kb interval flanking the TNIP1 region for each population, using the Impute2 program with data from iSelect genotyping as the source of observed genotypes, and using the haplotypes from the 1000 Genomes Project Phase I interim release (June 2011) for 1,094 individuals from Africa, Asia, Europe, and the Americas (40) (detailed information on the composition of this panel available at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf) and our in-house sequencing as reference. Impute2 estimates posterior probabilities for the 3 possible genotypes of imputed SNPs (i.e., AA, AB, and BB). Association analyses of imputed SNPs were performed using a missing data likelihood score test implemented in SNPTest, version 2.3.0 by taking into account the genotype uncertainty of imputed SNPs and adjusting for sex and global ancestry estimates (41). Conditional association analyses were also performed in SNPTest, adjusting for sex, global ancestry estimates, and SNP(s) of interest within the risk haplotypes. For haplotype analyses, the posterior probabilities were converted to the most likely genotypes with a threshold of 0.8. Imputed SNPs with an information measure of <0.5, an average maximum posterior genotype call probability of <0.9, or failed quality control measures were removed. LD between variants (confirmed by r2 values) and haplotypes was estimated, followed by haplotypic association using Haploview, version 4.2 (42). The number of variants imputed from the 1000 Genomes Project and our sequencing data that passed quality control measures are shown at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf.

Cell culture.

Epstein-Barr virus (EBV)–transformed B cell lines from European individuals were obtained from the Lupus Family Registry and Repository (43) at OMRF with IRB approval and were selected based on the genotypes of rs7719549 (a proxy of the H1 haplotype) and rs33934794 (a proxy of the H2 haplotype). Cell lines are either homozygous (carry 2 copies) for the nonrisk haplotype or homozygous (carry 2 copies) for each risk haplotype. Cell lines were cultured in RPMI 1640 supplemented with 10% fetal bovine serum, penicillin, streptomycin, L-glutamine, and 55 μM β-mercaptoethanol. We harvested equal numbers of cells under basal conditions in log-phase growth.

RNA isolation and quantitative reverse transcription–polymerase chain reaction (RT-PCR).

Total RNA was isolated using TRIzol total RNA isolation reagent (Invitrogen). Total RNA was measured using a NanoDrop spectrophotometer and was diluted with 20 ng/μl of MS2-RNA (Hoffmann-La Roche) to a final concentration of 0.5 μg/μl. Total RNA was treated with DNase, and complementary DNA was synthesized using an iScript cDNA Synthesis Kit (Bio-Rad). Quantitative PCR was performed using the SYBR Green method to determine the expression of TNIP1 messenger RNA (mRNA). Primer pairs were designed and synthesized (sense 5′-AAATCCAAATCAGAGCTCCCAA-3′, antisense 5′-CAAATGACACAATCTGGTCTCACT-3′). The PCR product corresponds to 2407–2519 bp of TNIP1 mRNA. HMBS (the human hydroxymethylbilane synthase gene) was used in quantitative RT-PCR as a reference (RT2 qPCR Primer Assay-SYBR Green Human HMBS Kit; SABiosciences). TNIP1 mRNA expression was normalized to HMBS.

Analysis of protein expression.

We harvested and lysed EBV-transformed B cells in Whole Cell Extraction Buffer (25 mM Tris, 1% Triton X-100, 150 mM NaCl, 1 mM EDTA, and protease inhibitors). Protein concentrations in each cell line were measured using Quick Start Bradford Protein Assay Kits and were adjusted to a final concentration of 2 mg/ml. Anti-ABIN1 antibodies (kindly provided by Drs. Sambit Nanda and Philip Cohen, University of Dundee, Dundee, Scotland, UK) were used to detect ABIN1 protein expression in EBV cell lines. The generation and characterization of this antibody have been described previously (44). Anti-GAPDH antibody (Cell Signaling Technology) was used to detect GAPDH protein expression, and expression of ABIN1 protein was normalized to GAPDH. An ECL Plus Western blotting detection kit (GE Healthcare) was used to visualize horseradish-peroxidase–conjugated antibodies. Band intensities were analyzed using ImageJ software (National Institutes of Health).

Statistical analysis.

Odds ratios (ORs) and 95% confidence intervals (95% CIs) for expression differences were calculated. Statistical significance was assessed by unpaired t-test, performed using Prism 5.0.


To test for genetic association, we performed single-marker logistic regression analysis with adjustment for sex and global ancestry estimated using AIMs. We found no convincing association in TAX1BP1 or TNIP2 that exceeded the Bonferroni corrected threshold of P < 3.21 × 10−4 for any of the populations. However, rs232722, located downstream of TNIP2, exhibited suggestive association in the East Asian cohort (OR 0.83 [95% CI 0.74–0.92], P = 8.91 × 10−4). These results are graphically displayed at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf. In contrast, significant association was observed for SNPs in TNIP1 in the European-ancestry population (rs6889239; P = 2.24 × 10−11), the African American population (rs13168551; P = 5.86 × 10−5), and the Hispanic population (rs7708392; P = 2.02 × 10−6) (Figures 1A–C). In the East Asian population we observed association slightly below the Bonferroni-corrected threshold (rs4958435; P = 5.49 × 10−4) (Figure 1D), while no significant association was observed in the African American Gullah population. Seventy-six percent of the East Asian study subjects were Korean. When the Korean members of this study population were analyzed independently, there were no marked differences in associations from those observed in the overall East Asian data set; hence, subsequent analyses were performed on the full East Asian data set. The presence of multiple associated SNPs with variable pairwise LD in the European, African American, and Hispanic populations (Figure 1) suggested that multiple independent SLE-associated haplotypes are present in the region.

Figure 1.

Single-nucleotide polymorphisms (SNPs) in and around the TNIP1 region associated with systemic lupus erythematosus in A, European-ancestry, B, African American, C, Hispanic, and D, East Asian populations. Genotyped SNPs are represented by triangles and imputed SNPs by circles. Horizontal red lines show the Bonferroni threshold of significance. The color of each triangle or circle represents the level of linkage disequilibrium between the SNP and the most significant SNP.

Before proceeding to evaluate TNIP1 using haplotype and conditional analyses, we sought to enrich our genotype data set by imputing untyped variants that were in LD with SLE-associated SNPs. To do this, we imputed variants from the 1000 Genomes Project using reference panels of individuals of European, African, Asian, and American Indian ancestry. This procedure increased the number of variants in all populations and added to the data set 19–30 additional variants in LD with directly genotyped SLE-associated SNPs. As a further enrichment step for causal variants we imputed variants from an independent resequencing study of European SLE cases (n = 159) and controls (n = 137), African American cases (n = 21) and controls (n = 20), and Hispanic cases (n = 38), performed at OMRF. This procedure added 13–30 novel variants not present in the 1000 Genomes Project reference panels, 5 of which were in moderate-to-high LD (2 variants with an r2 of 0.71 and 3 variants with an r2 of 0.99) with SLE-associated genotyped SNPs. In total, imputation of variants from the 1000 Genomes Project and our own resequencing introduced 24–35 (depending on the population) SLE-associated variants into consideration. Details are available at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf.

Association analyses using these enriched data sets demonstrated enhanced granularity of the association signals, although the most significant signals in the European, African American, and Hispanic populations remained constant (Figure 1). In the case of the East Asian population, however, the peak association signals shifted ∼6 kb centromeric from rs4958435 to rs2112635 (OR 0.74 [95% CI 0.63–0.88], P = 2.00 × 10−4) (Figure 1D).

To investigate the presence of multiple SLE-associated haplotypes in TNIP1, we constructed haplotypes using SNPs that surpassed the Bonferroni-corrected threshold of P < 3.21 × 10−4 (Table 2). In the European population we observed 2 risk haplotypes, H1 and H2, spanning 29 kb of the TNIP1 region. Of the 42 SNPs carried on H1 and the 19 SNPs carried on H2, 11 were found on both risk haplotypes and produced the most significant association signals (Figure 2). To determine if these haplotypes defined independent association signals, we performed conditional analyses. Conditioning on variants unique to H1 did not significantly change the magnitude of association for SNPs unique to H2, but did reduce the magnitude of the association for the SNPs shared by both risk haplotypes, by approximately one-half. Likewise, adjusting for the H2 haplotype failed to reduce the magnitude of association at variants unique to H1, while again reducing the magnitude of association for the shared SNPs. Adjusting for both H1 and H2 effectively eliminated the association signals in the region, thus confirming the presence of 2 independent risk haplotypes. Results of these analyses are depicted at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf.

Table 2. Association evidence for variants within TNIP1 meeting the Bonferroni threshold of P < 3.21 × 10−4 in the European-ancestry study population
Single-nucleotide polymorphismbp*Variant statusImputation information scoreAlleles, major/minorMinor allele frequencyOR (95% CI)P§
  • *

    From Human Reference Genome build hg19.

  • Genotyped (g), imputed from the 1000 Genomes Project data (i-1TGP), or imputed from sequencing data (i-seq).

  • The odds ratio (OR) was calculated with respect to the minor allele. 95% CI = 95% confidence interval.

  • §

    Adjusted for sex and global ancestry estimates.

rs2001542150432859i-TGP0.880C/T0.1111.28 (1.16–1.42)1.57 × 10−5
rs4958879150434422i-1TGP0.936A/G0.4821.19 (1.11–1.26)2.64 × 10−5
rs3792794150434722i-1TGP0.961C/T0.0841.32 (1.17–1.49)1.66 × 10−4
rs6579837150434894i-1TGP0.929G/T0.1161.28 (1.16–1.42)2.45 × 10−5
rs2233290150436503i-1TGP0.984G/C0.1101.31 (1.18–1.45)1.63 × 10−5
rs4958435150438284g1.000G/T0.4871.16 (1.08–1.24)2.60 × 10−5
rs4958880150438477g1.000C/A0.2251.27 (1.17–1.38)1.15 × 10−8
rs1422673150438988i-1TGP0.991C/T0.2251.32 (1.22–1.42)9.35 × 10−9
rs2042234150439131i-1TGP0.990A/G0.1101.31 (1.18–1.45)1.43 × 10−5
rs3805431150439539i-1TGP0.973G/A0.0841.32 (1.17–1.48)1.82 × 10−4
rs2233287150440097g1.000G/A0.1101.28 (1.14–1.42)1.15 × 10−5
rs3792790150442171i-1TGP0.970C/A0.4861.18 (1.11–1.26)2.06 × 10−5
rs4958436150442829i-1TGP0.976T/C0.2371.32 (1.22–1.42)4.81 × 10−9
rs17111708150443507i-1TGP0.982G/A0.1111.32 (1.19–1.47)5.65 × 10−6
rs7732451150444212i-1TGP0.982A/G0.1411.26 (1.15–1.38)2.57 × 10−5
rs73272818150444843i-1TGP0.986T/C0.1101.33 (1.2–1.48)3.40 × 10−6
rs10057690150445215i-1TGP0.984T/C0.1121.34 (1.21–1.49)1.91 × 10−6
rs1422674150445609i-1TGP0.960T/G0.1011.3 (1.17–1.45)1.98 × 10−4
rs1559127150446753g1.000T/C0.1311.28 (1.16–1.42)2.35 × 10−6
rs6880110150447090g1.000A/G0.1691.25 (1.14–1.37)1.47 × 10−6
rs6861227150447128g1.000T/G0.1681.24 (1.13–1.36)4.88 × 10−6
rs59926079150447743i-1TGP0.982A/G0.1121.34 (1.21–1.49)1.32 × 10−6
rs58474444150447880i-1TGP0.988T/G0.1101.34 (1.2–1.48)1.70 × 10−6
rs1862364150448376i-1TGP0.987A/G0.1701.27 (1.16–1.38)1.33 × 10−6
rs73272828150449220i-1TGP0.985C/T0.1121.35 (1.21–1.49)1.17 × 10−6
rs4958881150450236g1.000T/C0.1401.31 (1.19–1.45)9.34 × 10−8
rs10700649150451347i-seq0.984C/CCT0.1521.32 (1.2–1.45)3.90 × 10−8
rs3792785150451650i-1TGP0.983T/C0.1231.34 (1.21–1.48)3.37 × 10−7
rs13160369150452196i-1TGP0.985C/G0.1801.27 (1.16–1.38)5.16 × 10−7
rs13180950150452553i-1TGP0.986T/C0.1821.27 (1.16–1.38)4.39 × 10−7
rs6869605150452866g1.000A/C0.1471.29 (1.17–1.43)2.56 × 10−7
rs73272841150453384i-1TGP0.978C/T0.1231.33 (1.2–1.47)5.42 × 10−7
rs73272842150453888i-1TGP0.983G/A0.1491.31 (1.2–1.44)6.66 × 10−8
rs1107239150454606i-1TGP0.985T/A0.1231.33 (1.21–1.47)3.35 × 10−7
rs4958882150454787g1.000C/G0.1491.31 (1.19–1.44)5.64 × 10−8
rs3792784150455672g1.000A/G0.1191.29 (1.16–1.43)3.06 × 10−6
rs3792783150455732i-1TGP0.988A/G0.1781.26 (1.16–1.37)7.94 × 10−7
rs5872188150456053i-seq0.988CAG/C0.1501.32 (1.2–1.43)5.51 × 10−8
rs7731150150456392i-1TGP0.986G/A0.1501.31 (1.2–1.44)6.53 × 10−8
rs7708392150457485g0.989G/C0.2841.29 (1.2–1.39)5.51 × 10−11
rs6889239150457771g1.000T/C0.2851.3 (1.2–1.4)2.24 × 10−11
rs10036748150458146i-1TGP0.988C/T0.2831.33 (1.24–1.43)2.50 × 10−11
rs918498150459788g1.000C/T0.1291.23 (1.11–1.37)7.29 × 10−5
rs7719549150460047i-1TGP0.965C/T0.1261.32 (1.2–1.46)5.81 × 10−7
rs33934794150460088i-seq0.984GA/G0.1291.27 (1.15–1.39)1.14 × 10−4
rs960709150461049g1.000A/G0.2821.28 (1.19–1.39)2.25 × 10−10
rs78717966150462573i-seq0.932GGC/G0.0941.33 (1.19–1.49)7.67 × 10−5
rs62383766150462576i-seq0.913C/A0.0921.32 (1.18–1.48)9.75 × 10−5
rs13168551150462638g1.000T/C0.4411.2 (1.12–1.29)1.66 × 10−7
rs62383767150462705i-1TGP0.924G/A0.0951.32 (1.18–1.48)7.10 × 10−5
Figure 2.

TNIP1 haplotypes and linkage disequilibrium (LD) for systemic lupus erythematosus–associated variants that surpassed the Bonferroni correction threshold of P < 3.21 × 10−4 in the European-ancestry study population. Haplotypes observed at a frequency of >3% are shown. In the haplotype block, white boxes represent the major alleles and colored boxes represent the minor alleles. The groups of variants unique to H1, unique to H2, or shared by the H1 and H2 haplotypes are shown in blue, red, and green, respectively. Pairwise LD relationships (r2) are shown below the haplotype block, with different shading intensity according to degree of correlation between 2 variants. OR = odds ratio.

We then investigated for these haplotypes in the other study populations. H1 was present in the African American and Hispanic populations, but was not observed at a haplotype frequency of >3% in the East Asian population. The frequency of H1 differed across populations and was most prevalent in the African American population. H2 was observed in the African American, Hispanic, and East Asian populations. The East Asian population had the highest prevalence of the H2 haplotype, but it was evenly distributed between cases and controls and not associated with SLE. The ORs for both haplotypes in the African American and Hispanic populations ranged from 1.31 to 1.45, indicating that they conferred risk for SLE. Conditional analyses in these populations confirmed that they were independent. Detailed results are shown in Table 3 and at∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf.

Table 3. Summary of haplotype analysis results in the European-ancestry, African American, Hispanic, and East Asian study populations*
PopulationH1 risk haplotypeH2 risk haplotype
FrequencyFrequency in casesFrequency in controlsPORFrequencyFrequency in casesFrequency in controlsPOR
  • *

    Two independent risk haplotypes were observed in the European-ancestry, African American, and Hispanic populations, whereas only the H2 risk haplotype was found in the East Asian population. OR = odds ratio; NA = not applicable.

European ancestry0.1020.1150.0909.32 × 10−71.310.0790.0910.0685.33 × 10−71.37
African American0.1330.1530.1213.00 × 10−41.310.0200.0240.0170.0471.42
Hispanic0.0900.1010.0720.0021.450.2850.3140.2451.66 × 10−61.41
East AsianNANANANANA0.5310.5420.5270.2951.06

To evaluate the effects of the TNIP1 risk haplotypes on mRNA expression, quantitative RT-PCR was performed in an independent set of EBV-transformed B cell lines derived from individuals of European ancestry who were homozygous for H1 (n = 7), H2 (n = 7), or the nonrisk haplotype (n = 8). Compared with cells from subjects with the nonrisk haplotype, cells from subjects who were homozygous for H1 or H2 exhibited decreased expression of TNIP1 mRNA (P = 0.044 and P = 0.0035, respectively) (Figure 3A). Western blotting was also performed in the same independent set of EBV-transformed B cells, to evaluate the effects of the risk haplotypes on ABIN1 protein expression. In accordance with the mRNA measurements, reduced ABIN1 expression was observed in cells from subjects who were homozygous for H1 (P = 0.0463) or H2 (P = 0.0002) compared with those from subjects with the nonrisk haplotype, (Figure 3B and∼gaffneyp/ar-12-0551_TNIP1_Supporting_Information_PMG.pdf.). Taken together, these findings suggest that both haplotypes harbor hypomorphic risk variants that influence susceptibility to autoimmunity by reducing expression of ABIN1.

Figure 3.

Effect of the risk haplotypes H1 and H2 on TNIP1 mRNA expression (A) and ABIN1 protein expression (B). Each data point represents an individual subject; bars show the mean ± SEM. P values were determined by unpaired t-test.


In this study, we genetically dissected the TNIP1 locus in a multiethnic SLE case–control sample collection. Using haplotype and conditional analyses and a comprehensive variant data set derived from direct genotyping and imputation of variants from the public domain and targeted deep resequencing, we identified 2 independent risk haplotypes (H1 and H2) in the European-ancestry, African American, and Hispanic populations. These risk haplotypes likely carry hypomorphic functional alleles since cell lines derived from European-American individuals with these haplotypes exhibited reduced expression of TNIP1 mRNA and ABIN1 protein.

In a study of patients with systemic sclerosis, reduced expression of TNIP1 mRNA and ABIN1 protein was also demonstrated, in both lesional skin tissue and cultured dermal fibroblasts (23). The investigators identified rs2233287, rs4958881, and rs3792783 (all within the H1 haplotype) as being associated with disease. While the samples were not stratified by genotype, decreased expression in diseased tissue and association with what appears to be one of the risk haplotypes we have identified in SLE suggest the possibility of a shared mechanism in the etiologies of systemic sclerosis and SLE.

TNIP1 encodes the adapter protein ABIN1, which recruits A20 to polyubiquitinated NF-κB essential modulator (NEMO) (IKKγ) and subsequently facilitates NEMO deubiquitination and restriction of NF-κB signaling (45), making it a compelling candidate for SLE susceptibility. The importance of ABIN1 in restricting NF-κB signaling has been demonstrated in studies of mice deficient in ABIN1, which succumb to hepatocyte apoptosis, anemia, and bone marrow hypoplasia in utero (46). Moreover, mice expressing mutated ABIN1 that is defective in polyubiquitin binding develop lupus-like autoimmunity (44). Embryonic fibroblasts from ABIN1-deficient mice or mice expressing polyubiquitin binding–defective ABIN1 also exhibited hypersensitivity to TNF-induced apoptosis (44, 46). The ABIN1 interaction with polyubiquitin chains, including linear and K63 polyubiquitin chains, suppressed the activation of TLR–myeloid differentiation factor 88 signaling that is important in the prevention of autoimmune disease (44). The loss of ABIN1 has also been shown to increase the expression of TLR-induced CCAAT/enhancer binding protein β, resulting in development of a lupus-like inflammatory disease (47).

The H1 haplotype also carries the minor allele of a coding missense variant, rs2233290, which results in a proline-to-alanine substitution at position 151 in ABIN1. The P151A polymorphism is predicted to be damaging according to PolyPhen-2 version 2.1.0 (48), and represents a putative causal variant for future functional studies.

The H1 and H2 haplotypes include SNPs near the TNIP1 promoter, and it is therefore likely that causal variants located within regulatory elements affect TNIP1 gene expression. A recent study has functionally validated 5 NF-κB binding sites in the TNIP1 promoter that potentially influence TNIP1 expression (49). However, none of these NF-κB binding sites are modified by SNPs carried on the SLE-associated haplotypes.

We did not identify any significant association between SLE and variants in the other genes investigated (TNIP2 and TAX1BP1). We did, however, observe modest association in the TNIP2 region for the East Asian population, and this warrants further study in larger East Asian SLE sample collections. It should be noted that our TAX1BP1 fine-mapping SNP panel was near, but did not overlap sufficiently with, the JAZF1 locus, known to be associated with SLE in individuals of European ancestry (24). Thus, we cannot provide independent replication of the JAZF1 association. However, our results are likely sufficient to rule out an effect of TAX1BP1 being responsible for, or contributing to, the association in JAZF1.

In summary, the present results clarify the association signals in the TNIP1 locus in human SLE across multiple ethnic populations and suggest that reduced expression of ABIN1 contributes to SLE pathogenesis. Our data also inform ongoing efforts to identify TNIP1 causal variants in other autoimmune diseases, by clarifying the genetic architecture of the locus across multiple world populations. This transethnic mapping study design has also been successful in narrowing other SLE-associated regions, including the ITGAM and TNFAIP3 loci (20, 50). Given the importance of ABIN1 in restricting NF-κB signaling, a mechanistic understanding of how deleterious genetic variation in the TNIP1 locus influences susceptibility to autoimmune disease is needed. These results will serve to focus functional hypotheses that can be tested in the laboratory.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Gaffney had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Adrianto, Wang, Wiley, Lessard, Williams, Marion, Liang, Kaufman, Guthridge, Alarcón-Riquelme, Ramsey-Goldman, Edberg, Freedman, Jacob, Martín, Langefeld, Harley, Moser, Montgomery, Gaffney.

Acquisition of data. Adrianto, Wang, Wiley, Lessard, Kelly, Adler, Glenn, B. E. Wakeland, Liang, Kaufman, Guthridge, Alarcón-Riquelme, Alarcón, Anaya, Bae, Kim, Joo, Boackle, Brown, Petri, Ramsey-Goldman, Reveille, Vilá, Criswell, Edberg, Freedman, Gilkeson, Jacob, James, Kamen, Kimberly, Martín, Merrill, Niewold, Pons-Estel, Scofield, Stevens, Tsao, Vyse, Langefeld, Harley, E. K. Wakeland, Moser, Montgomery, Gaffney.

Analysis and interpretation of data. Adrianto, Wang, Wiley, Lessard, Williams, Ziegler, Comeau, Marion, B. E. Wakeland, Kaufman, Brown, Martín, Langefeld, E. K. Wakeland, Moser, Montgomery, Gaffney.


We would like to thank Drs. Sambit Nanda and Philip Cohen (MRC Protein Phosphorylation Unit, University of Dundee, Dundee, Scotland, UK) for providing the ABIN1 antibody, as well as Dr. David W. Powell (University of Louisville School of Medicine, Louisville, KY) for suggestions regarding the ABIN1 functional studies. We express our gratitude to the SLE patients and control subjects who participated in the study. We are thankful to the research assistants, coordinators, and physicians who helped in the recruitment of participants. We would like to thank the following groups/individuals for contributing samples genotyped in this study: BIOLUPUS Network, GENLES Network, Drs. Peter K. Gregersen (US), S. D'Alfonso and R. Scorza (Italy), P. Junker and H. Laustrup (Denmark), M. Bijl (The Netherlands), E. Endreffy (Hungary), C. Vasconcelos and B. M. da Silva (Portugal), A. Suarez, C. Gutierrez, and I. Rúa-Figueroa (Spain), and C. Garcilazo (Argentina). We are grateful to Drs. N. Ortego-Centeno, J. Jimenez-Alonso, E. de Ramon, and J. Sanchez-Roman from the Asociación Andaluza de Enfermedades Autoimmunes (Spain) for their collaboration, and to Drs. M. Cardiel, I. G. de la Torre, M. Maradiaga, and J. F. Moctezuma (Mexico), E. Acevedo (Peru), and C. Castel, M. Busajm, and J. Musuruana (Argentina) for collaboration on Hispanic populations enriched for American Indian/European admixture. We also thank the following other participants from the Argentine Collaborative Group: Drs. H. R. Scherbarth, P. C. Marino, E. L. Motta, S. Gamron, C. Drenkard, E. Menso, A. Allievi, G. A. Tate, J. L. Presas, S. A. Palatnik, M. Abdala, M. Bearzotti, A. Alvarellos, F. Caeiro, A. Bertoli, S. Paira, S. Roverano, C. E. Graf, E. Bertero, C. Guillerón, S. Grimaudo, J. Manni, L. J. Catoggio, E. R. Soriano, C. D. Santos, C. Prigione, F. A. Ramos, S. M. Navarro, G. A. Berbotto, M. Jorfen, E. J. Romero, M. A. Garcia, J. C. Marcos, A. I. Marcos, C. E. Perandones, A. Eimon and C. G. Battagliotti. We are grateful to Dr. P. S. Ramos and S. Frank for their assistance in genotyping, quality control analyses, and clinical data management, and to the staff of the Lupus Family Registry and Repository for collecting and maintaining SLE samples.