A highly prevalent lupus risk haplotype increases IRF7-dependent induction of IFN-α, enhancing antiviral defense and exacerbating autoimmunity

Abstract

Genome-wide association studies have identified genetic polymorphisms at 11p15 associated with Systemic Lupus Erythematosus (lupus). Statistical fine mapping prioritizes a highly prevalent coding haplotype within IRF7. Analysis of ancient DNA confirms that this haplotype has persisted at high frequencies in the global population for millennia. The IRF7 risk haplotype is sufficient to increase nuclear localization of IRF7 and transcriptional activity downstream of pattern recognition receptor pathways. This risk haplotype increases IRF7 DNA binding strength and alters IRF7 DNA sequence specificity, resulting in genotype-dependent increases in IFN-α production in numerous biological systems, including monocytes and airway epithelial cells. CRISPR engineering of a homologous risk variant in mouse Irf7 results in both enhanced innate control of virus infection and increased autoantibody titers in a model of autoimmunity. Altogether, we establish a persistent and prominent IRF7 haplotype that amplifies IRF7 activity in a manner that has immunological risks and benefits.

Graphical Abstract

Virolainen et al. combine human genetics, functional genomics, and mouse models to show that a common lupus-risk IRF7 haplotype enhances IRF7 nuclear localization, DNA binding, and type I interferon production. The haplotype improves antiviral defense but increases autoreactivity, revealing an evolutionary trade-off between host protection and autoimmunity.

INTRODUCTION

Systemic lupus erythematosus (SLE or lupus) is a debilitating autoimmune disease driven by widespread inflammation, immune complex formation, and autoantibody production^1,2. The heritability of lupus is between 43 and 66% across ancestral groups, and over 200 genetic risk loci have been associated with lupus through genome-wide association studies (GWAS)^3–10. Genetic risk loci for lupus are highly enriched for type I interferon (IFN-I) signaling pathway components^11,12. IFN-I production is initiated upon sensing of viruses by pattern recognition receptors (PRRs), leading to an inflammatory signaling cascade and a gene expression program that restricts viral growth and influences innate and adaptive immune responses¹³. Thus, potent IFN-I responses are critical to clear viral infections, but dysregulation of this pathway contributes to autoimmunity and lupus¹⁴.

Interferon regulatory factor 7 (IRF7) is sometimes referred to as the master regulator of the IFN-I response^13,15. IRF7 is activated downstream of pattern recognition receptors (PRRs)^15–19 that include Toll-like receptors (TLRs), RNA sensor RIG-I (encoded by RIGI), and cyclic GMP-AMP synthase-STING pathway (encoded by CGAS and STING1). IRF7 activation leads to its nuclear localization and binding to DNA, with subsequent induction of interferon genes and IFN-I secretion^{16,18,20–22}. Secreted IFN-I binds the IFN-I receptor (IFNAR) to initiate autocrine and paracrine signaling cascades that restrict virus replication, induce interferon stimulated genes (ISGs), propagate inflammatory responses, and contribute to adaptive immune responses^13,23. Robust IRF7 driven IFN-I responses are crucial to human health, as IRF7 deficiency and autoantibodies that neutralize IFN-I each increase the risk of hospitalization and death from viruses such as influenza virus A (IAV) and SARS-CoV-2^24–27.

IRF7-driven IFN-I responses also contribute to pathogenesis of autoimmune diseases, including lupus^28,29. More than 50% of patients with lupus have elevated IFN-I in peripheral blood, a phenomenon known as the “IFN-I Signature” of lupus³⁰. In addition, autoantibody production in mouse models of lupus is Irf7 dependent³¹. Variants at the chromosome 11p15 locus encoding IRF7 are associated with differential risk for lupus development^32–35. Specifically, a human variant (NM_001572.5:c.1556A>G; p.Gln412Arg; rs1131665) results in a coding amino acid change from Arg to Gln at position 412 of IRF7, which is the most robustly associated variant at the IRF7 locus across numerous studies^32,33,36. Further, this variant is highly prevalent in the global population, with major allele frequencies approaching nearly 98% in certain populations^32,33. However, much remains to be explored about the biological consequences of variation at the IRF7 locus and its impact on lupus risk. As IRF7 is a critical regulator of both antiviral and autoimmune-related IFN-I responses, and IRF7 genetic risk alleles are highly prevalent in the human population, an improved understanding of the causal variants and their mechanisms of action is necessary to inform the development of new therapies and response to treatment in these important health contexts.

In this study, we explore the evolutionary persistence of this highly prevalent lupus-associated haplotype in IRF7 (Gln412Arg: Lys179Glu) and its relationship with the magnitude of IFN-I responses. Ancient DNA datasets are combined with viral challenge experiments to test the hypothesis that coding variants in IRF7 persisted across human evolution based on their ability to enhance IFN-mediated antiviral immunity, with increased risk of lupus being an evolutionary consequence of this benefit to host defense. Mechanistically, we explore the effect of IRF7 coding risk variants on IRF7 nuclear translocation, interactions with DNA, transcriptional regulatory activity, and IFN-α production downstream of innate pattern recognition receptors. Altogether, we establish a persistent and prominent IRF7 haplotype that amplifies IRF7 activity in a manner that has immunological risks and benefits.

METHODS

Genotyping of Variants

In this study, we genotyped 120 SNPs covering the IRF7 region, spanning GRCh37 chr11: 550,028–697,661 as part of a larger collaborative study, the Large Lupus Association Study 2 (LLAS2). The samples were collected from individuals in the United States, Asia, Europe, and Latin America. They were genotyped using the Illumina iSelect platform located at the Lupus Genetics Studies Unit at the Oklahoma Medical Research Foundation. The subjects were grouped into four ancestral groups. All lupus patients met the American College of Rheumatology criteria for the classification of lupus³⁷ and were enrolled in this study through an informed consent process approved through the local Institutional Review Boards. LLAS2 included genotyping of other lupus risk loci, and the analyses of those loci from this same collection, with and without lupus, have been published separately. Variants were filtered to include only those with a call rate >0.95 and a Hardy-Weinberg P-value of >0.0001 in controls. Following variant filtration, samples that did not meet the >0.95 threshold were removed. Variants are described according to HGVS nomenclature using the IRF7 RefSeq transcript NM_001572.5.

Ascertainment of Population Stratification

The ancestries of the subjects in this study were self-identified. Genetic outliers from each ancestry group were removed from further analysis as determined by principal component (PC) analysis and admixture estimates, as described previously^38–40. A PC analysis of the remaining samples (after outlier removal) confirmed that no sample had a PC1–3 more than 2 standard deviations outside of the mean. We used 347 ancestral informative markers (AIMs) from the same custom genotyping study that passed quality control in both EIGENSTRAT⁴⁰ and ADMIXMAP^41,42 to distinguish the four continental ancestral populations: African, European, Admixed American, and East Asian, allowing identification of the substructure within the sample set^43,44. We utilized PCs from EIGENSTRAT outputs to identify outliers of each of the first three PCs for the individual population clusters through visual inspection [see Fig. 1 of reference³⁸]. Three PCs were used because they accounted for 95% of the eigenvalues. Because the four admixture estimate proportions sum to 1, any three of the four provide the full set of information, so only three proportions were necessary to include as covariates; the East Asian proportion was omitted from this analysis, as it corresponded to the largest ancestry segment of the combined population. Admixture estimates were included as covariates in the stepwise logistic regression analysis.

Genotype Imputation

To detect associated variants that were not directly genotyped, we imputed the IRF7 region with Beagle 4.1⁴⁵ using a composite imputation reference panel based on the 1,000 Genomes Project Phase 3⁴⁶ sequence data. Imputed genotypes were included in the analysis if they met or exceeded a probability threshold of 0.5 and the same quality-control criteria thresholds described for the genotyped markers. The most likely genotype was used for variants passing quality controls in all analyses.

Genetic Association Analysis

To identify groups with sufficient statistical power to identify true genetic associations, we performed a Power Calculation (α=0.05) for the Gln412Arg haplotype. We did not assess genetic association at the IRF7 locus in the East Asian and Admixed American ancestry groups due to insufficient power, as presented in Table S1.

We tested each variant for its association with lupus using logistic regression that included three admixture proportion estimates as covariates, as implemented in Golden Helix SVS (8.9.1). The additive genetic model was assessed as the initially tested model of inheritance. Using Golden Helix SVS, stepwise logistic regression was performed to identify variants independently associated with the development of lupus. For these analyses, admixture estimates were also used as covariates.

A trans-ancestral meta-analysis was performed using Golden Helix SVS (8.9.1). The P-values and odds ratios for each variant were included, and the analysis was weighted by the number of individuals with data for each variant in an individual cohort as part of an inverse variance model. For the trans-ancestral meta-analysis, only the European and African ancestral cohorts were used. Plots for genetic data were generated using locuszoomr (version 0.2.0)⁴⁷ and annotated using ensembldb (version 2.99.0)⁴⁸.

Ancient DNA Analysis

A dataset containing ancient and present-day DNA was downloaded from the Ancient Genome Diversity Project⁴⁹ for haplotype analysis (16,389 unique individuals: 9,990 ancient, 6,399 present-day). This dataset contains ~1.23 million positions in the genome (in hg19 coordinates) and is in a modified Eigensoft format. To extract haplotypes, the dataset was converted with the convertf program (EIGENSOFT 8.0.0 package;^40,50) into a ped file and then converted with PLINK⁵¹ into a BED file. All samples younger than 1,950 CE (age “0” years before present, reflecting radiocarbon date correction) were removed to only represent “ancient” samples. All ages were adjusted to “calendar year” for figures.

As this is not a whole genome dataset, but instead a sampling of SNPs across the genome, we searched for any lupus SNPs in perfect LD with the lead SNPs contained within this dataset. Of the SNPs within the lupus IRF7 risk haplotype block, only the rs11246213 locus was present. This locus was extracted from the larger dataset, and then all samples with missing data at this locus were removed (ancient genomes have high proportions of missing data due to degradation compared to modern genomes). The dataset was also filtered to remove duplicates when a genome was sequenced more than once (common for ancient datasets as sequencing technologies improve). Almost all duplicates had the same rs11246213 haplotype across replicates, but a small subset had different alleles inferred (N=7). For these seven samples, we assessed data quality to choose the most probable haplotype and deleted the duplicate(s) (e.g., highest coverage from sequencing, or removing any replicates that indicated contamination).

The final dataset contained 4,503 individuals across Africa, East Asia, West Eurasia (European), Oceania, Central Asia/Siberia, and across North, Central, and South America. Most of the samples (particularly the very old samples) are from West Eurasia (Europe) and Central Asia/Siberia, because those are the places with permafrost and cold/deep caves to preserve very old skeletal remains including Neanderthal bones. Most samples are younger than 10,000 BCE, but the oldest are over 40,000 BCE in our final dataset. The haplotype for each individual was converted to an allele frequency percentage (0 for non-risk G, 0.5 for heterozygous, 1 for homozygous risk allele A). This dataset was then used to assess population allele frequencies from each ancestry in 500-year time bins. This bin size is appropriate based on the uncertainty inherent in radiocarbon dating due to the need to calibrate for background C-14 levels⁵². Weighted loss curves were fit for each ancestry background from 10,000 BCE onward.

Human groups were assigned to ancestries based on definitions from the 1000 Genomes Project. The only populations that required additional attention were in the transition zone across Eurasia to divide populations into the historically relevant groups of (1) West Eurasia, (2) Central Asia and Siberia, (3) East Asia, and (4) Oceania. Recent ancient and modern DNA studies were used to assign populations across this region^53–59.

Cell Lines and Cell Culture

THP-1 IRF7-deficient (knockout) cells (InvivoGen, Cat. #thpd-koirf7), HEK-Blue hTLR7 cells (InvivoGen, Cat. #hkb-htlr7), and A549 IRF7-deficient (knockout) cells (Abcam, Cat. #ab267210) were used in this study. THP-1 cells were cultured in RPMI 1640 medium (Cytiva, Cat. #SH30255.01) supplemented with 10% fetal bovine serum (FBS; Gibco, Cat. #26140–079), 10,000 U/mL penicillin, 10,000 μg/mL streptomycin, and Normocin (InvivoGen, Cat. #ant-nr-2). Following transduction, THP-1 cell lines stably expressing IRF7 variants were maintained in medium containing puromycin (InvivoGen, Cat. #ant-pr-1) at 1 μg/mL. HEK-Blue hTLR7 cells were cultured in DMEM (Gibco, Cat. #11965–092) supplemented with 10% FBS, 10,000 U/mL penicillin, and 10,000 μg/mL streptomycin. A549 cells were cultured in Ham’s F12-K (Kaighn’s) medium (Thermo Fisher Scientific, Cat. #21127022) supplemented with 10% FBS, 10,000 U/mL penicillin, and 10,000 μg/mL streptomycin. All cell lines were maintained at 37°C in a humidified atmosphere containing 5% CO₂ and were confirmed to be free of mycoplasma contamination.

Expression Plasmid Generation

For generation of stable IRF7-expressing cell lines, a pUNO1-hIRF7a-HA3x expression vector was purchased from InvivoGen and the region containing IRF7-HAx3 was cloned into a lentiviral transfer plasmid (pLentiPuro). The GeneArt site-directed mutagenesis system (Thermo Fisher Scientific, Cat. #A13282) was used to generate the IRF7 lupus protective haplotype according to the manufacturer’s instructions. For transient transfections, an IRF7 open reading frame (wild type IRF7) was cloned into an overexpression vector from the Wolfgang Hammerschmidt lab. The GeneArt site directed mutagenesis system (Thermo Fisher Scientific) was used to generate the various IRF7 (Q412R and K179E) mutants on the overexpression vector as well as the pLentiPuro (described above) backbone for this study according to the manufacturer’s instructions using the following primer sets: (forward) SDM 179K->E [GCCCCAGCTGGTGACGAGGGGGACCTCCTG], (reverse) SDM 179K->E [GCAGGAGGTCCCCCTCGTCACCAGCTGGGGC], (forward) SDM 412Q->R [CTTCAGAGTCTTCTTCCGAGAGCTGGTGGAATTCC], (reverse) SDM 412Q->R [GGAATTCCACCAGCTCTCGGAAGAAGACTCTGAAG]. A construct on this same backbone containing IRF7 attached to an enhanced green fluorescent protein (eGFP) tag was used as a control to show IRF7 activity on transfection. All plasmids used in this study were verified using Sanger sequencing.

Cell Stimulations and Treatments

Cells were stimulated with resiquimod (R848; InvivoGen, Cat. #tlrl-r848–1) at 1 μg/mL, poly(dA:dT)-Lyovec (InvivoGen, Cat. #tlrl-patc) at 2.5 μg/mL, or low-molecular-weight poly(I:C) (InvivoGen, Cat. #tlrl-picw) at 2.5 μg/mL. Poly(I:C) was transfected using Lipofectamine 2000 (Invitrogen, Cat. #11668027) at 1 μL/mL. Type I interferon stimulation was performed using Human Universal Type I Interferon (PBL Assay Science, Cat. #11200) at 1000 U/mL.

Isolation of Nuclear Lysates for Western Blot

To obtain nuclear lysates for Western blots, respective cell lines were resuspended in 1 mL of cold PBS (Corning, Cat. #21–040-CV) per 10 million cells. Cells were then centrifuged at 4°C, 300×g for 5 min and the PBS was aspirated off. Cell pellets were resuspended in 400μl of CE buffer (10 mM HEPES (Thermo Fisher Scientific, Cat. #15630106), pH 8; 10 mM KCl (Thermo Fisher Scientific, Cat. #AM9640G); 0.1 mM EDTA (Thermo Fisher Scientific, Cat. #AM9260G); 1 mM DTT (Thermo Fisher Scientific, Cat. #OH2369); 1×Halt protease and phosphatase inhibitor (Thermo Fisher Scientific, Cat. #PI87786) and incubated for 15 min on ice. Afterwards, 25μl of 10% Nonidet P-40 (Abcam, Cat. #ab142227) were mixed into the solution and the samples were centrifuged at max speed (17,300×g) at 4°C for 3 min. After discarding the supernatant, the cell pellet was resuspended in 30μl of NE Buffer (20 mM HEPES, pH 8; 0.4 M NaCl; 1 mM EDTA; 1 mM DTT; 1×Halt phosphatase inhibitor). Cells were sonicated using the Q125 sonicator (Qsonica) at 20% power, 5 s pulses for a total of 15 s. The supernatant was stored at −70°C. Protein concentration was measured using the bicinchoninic acid assay (BCA) (Thermo Fisher Scientific, Cat. #23227) using a 5.5μl aliquot.

Quantitative Western Blot Using Nuclear Lysates and Total Protein Stain

Nuclear lysates of samples were mixed with loading buffer and DTT and heated at 95°C for 2 min. After cooling, the samples were loaded into a 4–12% Nu page Bis–Tris Gel (Invitrogen) and ran for 90 min at 130 V in MOPS buffer, or 65 min at 130v in MES buffer. The gel was then transferred to a nitrocellulose membrane using the iBlot Machine (ThermoFisher Scientific). For Western blot normalization, membranes were stained using the Revert^™ 700 Total Protein Stain (LI-COR Biosciences, Cat. #926–11021). The membrane was then blocked in Intercept Blocking Buffer (LI-COR Biosciences, Cat. #927–70001) for 1 h at room temperature. The primary antibody (IRF7 G-8; Santa Cruz Biotechnology, Cat. #sc-74472) was diluted in Intercept Blocking Buffer (with Tween 20 diluted at 1:1000) and the membranes incubated with rocking in primary antibody overnight at 4°C. Membranes were washed in a PBS/Tween 20 solution twice and were incubated in fluorescently labeled secondary antibody (RDye 800CW Goat Anti-Mouse IgG secondary antibody: LI-COR Biosciences, Cat. #926–32210) in Intercept Blocking Buffer (with Tween 20 (Sigma-Aldrich, Cat. #P9416) and SDS) for 60 min at room temperature with rocking. Membranes were washed again in a PBS/Tween 20 solution 2 times before being imaged using the Odyssey DLx Imaging System (Licor Bioscience). Protein expression levels were determined using Empiria Studio v2.2 (Licor Bioscience).

Quantification of ISRE (Interferon-stimulated response element) Transcriptional Activity Using Renilla Luciferase

IRF7 open reading frames (ORFs) with the various amino acid changes were integrated as described above into an overexpression vector that was a gift from the Wolfgang Hammerschmidt lab. These plasmids expressing IRF7 were co-transfected (50 ng per well) into HEK Blue hTLR7 cells seeded in 96-well plates (30,000 cells per well) using TransIT-2020 Transfection Reagent (Mirus Bio, Cat. #MIR5404) with ISRE reporter (BPS Bioscience, Cat. #60613, 50ng per well) or a negative control plasmid (50 ng per well) without an ISRE reporter from BPS Bioscience according to the manufacturer’s protocol. Post-transfection, cells were stimulated for 24-hours with R848 and the two-step luciferase (Firefly & Renilla, BPS Bioscience, Cat. #60683) was used to quantify firefly and renilla luciferase signals. The firefly luciferase signal was normalized to the renilla signal for each replicate and then further normalized to the ISRE negative control signal to calculate normalized ISRE luciferase activity for each replicate.

Quantification of Secreted IFNα

Quantification of secreted human IFNα from cell supernatants was measured by Human IFN-Alpha All Subtype ELISA (Enzyme-linked immunosorbent assay) Kit, High Sensitivity Serum, Plasma, TCM, Cat. #41115 from PBL Assay Science.

Lentiviral Vector Generation

Lentiviral transfer plasmids (pLentiPuro) containing IRF7 ORFs were given to the Viral Vector Core at Cincinnati Children’s Hospital for generation of VSV-G pseudotyped lentiviral vector. Viral titer was determined by puromycin titration.

Transduction of IRF7 Into Monocyte and Airway Epithelial Cell Lines

THP-1 IRF7 deficient cells were transduced with the viral vectors described above. 100,000 cells were plated in 12-well plates treated with RetroNectin-fibronectin fragment (Takara Bio, Cat. #T100B) and treated with viral supernatant followed by spin-fection at room temperature at 1,300 × g for 45 minutes. Following two days of incubation at 37°C and 5% CO₂, 1 μg/mL of puromycin (InvivoGen, Cat. #ant-pr-1) was added to cell culture media to select for clones containing IRF7. Initial titration showed that a MOI of 2 was sufficient for stable expression of IRF7 and all cell lines generated for this study utilized an MOI of 2 for transduction.

A549 IRF7 deficient cells were transduced with the viral vectors generated above. 100,000 cells were plated in 12-well plates treated with RetroNectin-fibronectin fragment (Takara) and treated with viral supernatant followed by spin-fection at room temperature at 1,300 × g. Following two days of incubation at 37°C and 5% CO₂, cells were isolated and diluted for monoclonal isolation. Briefly, cells were diluted to a concentration of 0.3 cell/100 μl of media and plated in multiple 96-well plates and allowed to sit. Wells were then scanned 10–14 days later for colonies emerging from a single clone. Additional clones were validated using ELISA (PBL Assay Science, as described above) for IFN-α detection.

RT-qPCR

RNA was extracted using the RNeasy mini kit (Qiagen, Cat. #74104) and 1μg of RNA was used to make cDNA using SuperScript II ReverseTranscriptase (Thermo Fisher Scientific, Cat. #18064014) according to the manufacturer’s protocol. cDNA was then diluted at a 1:20 ratio and qPCR was performed on a QuantStudio with iTaq Universal SYBR Green Supermix (Bio-Rad, Cat. #1725121) with the following forward and reverse primers.

Transcript	Forward (5’->3’)	Reverse (5’->3’)
GAPDH	ctgcaccaccaactgcttag	gtcttctgggtggcagtgat
IFNα (IFNα2, IFNα25, IFNα21)	tgaaggacagacatgactttgg	tcctttgtcctgaagagattga
IRF7	gctggacgtgaccatcatgta	gggccgtataggaacgtgc
MX1	ctgggattttggggcttt	gggatgtggctggagatg
OAS1	tcagaaataccccagccaaa	gagccaccctttaccacctt

Chemiluminescent Western Blot For pIRF7 and Total IRF7

Nuclear lysates of samples were mixed with loading buffer and DTT and heated at 95°C for 2 min. After cooling, the samples were loaded into a 4–12% Nu page Bis–Tris Gel (Invitrogen) and run for 90 min at 130 V in MOPS buffer, or 65 min at 130v in MES buffer. The gel was then transferred to a nitrocellulose membrane using the iBlot Machine (ThermoFisher Scientific). Samples were blocked in 5% milk in TBST (50 ml TBST plus 2.5 g dehydrated milk powder) for at least 1 h. The primary antibody (IRF7 D71E1W; Cell Signaling Technology, Cat. #12390) was diluted in 5% milk plus TBST at a 1:1000 ratio and the membranes incubated in primary antibody overnight at 4 °C with rocking. Membranes were washed in a TBST solution 3 times for 5 min each and were incubated in HRP conjugated secondary antibody (Cytiva, Cat. #NA934V) in 5% milk in TBST for 1 h at room temperature with gentle shaking. Membranes were washed again in a TBST solution 3 times for 5 min each. Then, 2 ml of SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific, Cat. #34095) was added to each blot and incubated for 2 min. Substrate was removed before being imaged with the ChemiDoc Touch Imaging System (Bio-Rad).

RNA-sequencing

We performed RNA-seq from experiments in both HEK and THP-1 cells. The mirVana RNA (Thermo Fisher Scientific, Cat. #AM1560) isolation kit was used to isolate total RNA from 6-well plates of 700,000 cells per well of each cell line in triplicate (at greater or equal to 90% viability) following the manufacturer’s recommended procedures. THP-1 libraries were prepared from RNA samples using the TruSeq Stranded Total RNA with Ribo-Zero Globin (Illumina, Inc., San Diego, CA) and sequenced at 150 paired end bases (50 M reads per sample) at the CCHMC Genomics Sequencing Facility. HEK libraries were prepared from RNA samples and sequenced at 150 paired end bases (50 M reads per sample) at Novogene.

RNA-seq datasets were processed using the nf-core pipeline^60,61 (v. 3.11.1) (https://nf-co.re/rnaseq). In brief, the samples underwent QC using FastQC v. 0.11.9 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and adaptors were trimmed and filtered using Cutadapt (v. 3.4) (https://cutadapt.readthedocs.io/en/stable/)⁶². Other tools, such as RSeQC^63,64, Qualimap^65,66, dupRadar⁶⁷, and preseq (https://github.com/smithlabcode/preseq), were used to assess the quality of the datasets. Sequencing data were aligned to the hg19 genome using STAR (v. 2.7.9a). Ribosomal RNA was removed from the aligned datasets using the Silva Library (https://www.arb-silva.de/) and SortMeRNA v. 4.3.4 (https://bioinfo.lif.fr/RNA/sortmerna/) to eliminate potential contamination from non-target species. Transcript abundance estimates were calculated using GENCODE (hg19; v19)⁶⁸ annotations with Salmon⁶⁹ (v1.10.1). The quantifications were imported into R (v4.1.3) and summarized at the gene level using tximport (v1.22.0)⁷⁰. Differential analyses were performed using DESeq2⁷¹ (v1.34.0) and genes were considered to be differentially expressed if the adjusted p-value was less than 0.05 and the absolute value of the log₂ fold change was greater than 0.585 (i.e., a 1.5-fold linear change).

RNA-seq library quality was assessed using multiple complementary QC tools, including FastQC, RSeQC, Qualimap, dupRadar, and preseq. All datasets passed standard QC procedures, with strong agreement between biological replicates and biologically coherent expression profiles. Expression-dependent duplication profiles further supported the quality of the HEK Blue hTLR7 and THP-1 RNA-seq libraries (Figure S1). RNA-seq-based gene pathway enrichment analysis was performed with Enrichr⁷² and gene sets from Reactome 2022⁷³.

Transcription factor motif enrichment analysis of differentially expressed genes

Enrichment of IRF7 DNA binding motifs within the promoters of differentially expressed genes (DEGs) was assessed using HOMER v4.9 (findMotifs.pl;⁷⁴). Promoter regions were defined as −1,500 to +500 bp relative to the transcription start site. Analyses were performed for all DEGs and separately for genes upregulated or downregulated in cells expressing the lupus-risk haplotype relative to the protective haplotype. These promoter sets were compared with promoters of all human genes as background. Rather than the default HOMER motif library, IRF7 position weight matrices obtained from CIS-BP build 2.00⁷⁵ were supplied as known motifs using -mknown, with the detection threshold for each matrix set to 70% of its maximum attainable log likelihood score.

Nuclear Extraction for Protein Binding Microarrays

50 million THP-1 cells stably expressing the lupus and protective haplotypes with a HA epitope tag (generated as described above) were maintained in T175 flasks prior to transfection with Poly(I:C). Three hours following 2.5 ug/mL Poly(I:C) transfection (see above), cells were centrifuged and washed in 1X PBS+ 1X protease inhibitor and centrifuged to remove PBS. Cell pellets were resuspended in Buffer A (10mM HEPES pH 7.9, 1.5 mM MgCl_2, 10mM KCl, 0.1 mM protease inhibitor cocktail, and 0.5mM of DTT) and incubated on ice for 10 minutes. Following incubation, 10% lgepal detergent was added and cell pellets were vortexed to ensure the detergent was spread throughout the sample. Following cell lysis, samples were centrifuged to pellet nuclei and cytosolic fractions were collected, snap frozen in liquid nitrogen, and stored at −80 °C. The nuclear pellet was washed with additional Buffer A. Pelleted nuclei were then resuspended in 4 aliquots of 45 μl of Buffer C per tube (20mM HEPES pH 7.9, 25% glycerol, 1.5mM MgCl₂, 0.2 mM EDTA, 420 mM NaCl, 0.1mM protease inhibitor cocktail, and 0.5 mM DTT) and vortexed briefly followed by incubation with rotation and agitation at 4C for 1 hour. Following incubation, nuclei were pelleted and the nuclear extract was isolated, snap frozen with liquid nitrogen, and stored at −80 °C until application for the protein binding microarray (PBM).

Protein Binding Microarray Experiments

Microarray design:

The microarrays used for the PBM experiments were described previously in Andrilenas et al.⁷⁶. The 10,410 synthetic, ISRE-type sequences were based on 2- and 3-bp-spacer ISREs defined by 5’-GANA-3’ type core elements separated by 2 or 3 bases, respectively. The microarray contained 6,954 2-bp ISRE variants based on 76 seed ISRE sites and all single-nucleotide variants of those seeds, as well as 3,456 3-bp ISRE variants based on 36 seed sites and all single-nucleotide variants.

PBM experiments:

The microarray double-stranding procedure and nuclear extract PBM experiments were performed as described previously^77,78. Briefly, double-stranded microarrays were pre-wetted in HBS (20 mM HEPES, 150 mM NaCl) containing 0.01% Triton X-100 for 5 min and then de-wetted in an HBS bath. Next, the array was incubated with 180 μL of protein sample mixture: 45 μL of nuclear extract sample, 135 μL of binding reaction buffer [20 mM HEPES, pH 7.9, 1 mM DTT, 0.2 mg/ml bovine serum albumin, 0.02% Triton X-100, 0.4 mg/ml salmon testes DNA, Sigma-Aldrich, Cat. #D7656]. Note that the final salt concentration was ~105 mM NaCl due to the high-salt nuclear extract sample. The protein was left to bind the microarray for 1 hour in the dark. The array was then rinsed in an HBS bath containing 0.05% Tween 20 and subsequently de-wetted in an HBS bath. After the protein incubation, the array was incubated for 20 min in the dark with 10 μg/ml primary antibody (Abcam, Cat. #ab9110) in 2% milk in HBS. After the primary antibody incubation, the array was rinsed in an HBS bath containing 0.1% Tween 20 and de-wetted in an HBS bath. Microarrays were then incubated for 20 min with 20 μg/ml of Alexa 488-conjugated secondary antibody (Thermo Fisher Scientific, Cat. #A-11008) in 2% milk in HBS. The array was rinsed in an HBS bath containing 0.05% Tween 20 and then placed in a Coplin jar containing 0.05% Tween 20 in HBS. The array was agitated in solution in a Coplin jar at 125 rpm on an orbital shaker for 3 min and then placed in a new Coplin jar with 0.05% Tween 20 in HBS to repeat the washing step. It was then placed in a Coplin jar containing HBS and washed for 2 min as described above. After the washes, the array was de-wetted in an HBS bath. Microarrays were scanned with a GenePix 4400A scanner and fluorescence was quantified using GenePix Pro 7.2. Exported fluorescence data were normalized with MicroArray LINEar Regression⁷⁹.

DNA-binding motif determination:

Binding motifs were constructed using a seed and single-nucleotide variant approach described previously^76,80. The approach highlights the individual base preferences when the protein is binding to a specific seed element and allows direct comparison of the base preferences between distinct proteins at a common DNA element. Data for purified IRF7 homodimers were obtained from Andrilenas et al.⁷⁶.

Generation of Mutant Mice

The single-guide RNA (sgRNA; spacer sequence GAAAGTGCTGAAGTCGAAGA) used to generate mice carrying the Arg334Gln risk allele was selected on the basis of predicted on- and off-target scores using CRISPOR⁸¹. The sgRNA was transcribed in vitro using the MEGAshortscript T7 Kit (Thermo Fisher Scientific), purified using the MEGAclear Kit (Thermo Fisher Scientific), and stored at −80 °C. To prepare the injection mixture, sgRNA and recombinant Cas9 protein (IDT) were incubated at 37 °C for 10 min to form ribonucleoprotein complexes, after which a single-stranded oligodeoxynucleotide donor template (ssODN; IDT; 5′-CCCATCTTtGAtTTCAGtACTTTCTTCCaAG-3′) encoding the Arg334Gln substitution was added. Silent mutations shown in lower case were introduced to create the ScaI restriction enzyme site for genotyping and to prevent gRNA recutting. Final concentrations were 75 ng/μL sgRNA, 150 ng/μL Cas9 protein, and 75 ng/μL ssODN donor. The injection mixture was introduced into the cytoplasm of one-cell-stage C57BL/6 embryos using piezo-driven microinjection⁸². Injected embryos were immediately transferred to the oviductal ampullae of pseudopregnant CD-1 females. Live-born pups were genotyped by PCR using primers 5′-CAGAAGCAGCTGCACTACACAG-3′ and 5′-GCCTCCCTGCTGACAAGTC-3′. The knock-in allele was identified by ScaI restriction digestion and confirmed by Sanger sequencing. Animals were maintained under a 12-h light/dark cycle. All animal studies were approved by the Cincinnati Children’s Hospital Medical Center Institutional Animal Care and Use Committee.

C57BL/6 mice are homozygous for nucleotide sequences encoding an Arg at position 334 in Irf7 (the homologous position to human IRF7 412). CRISPR-Cas9 genome-editing with homologous recombination was used to produce heterozygous Arg/Gln 334 mice from the C57BL/6 background. Mice were genotyped using Sanger sequencing with DNA from tail clips of the mice. Arg-Arg (protective) and Gln-Gln (risk) mice were bred for the subsequent experiments. Mice were bred and maintained in the vivarium adhering to IACUC guidelines.

Epicutaneous Application of R848

Eight-week-old female Arg-Arg and Gln-Gln mice were sensitized by epicutaneous application of 100 micrograms of R848 three times per week for 8 weeks. Serum samples were collected before and after 8 weeks of the treatment and tested for autoantibodies against double-stranded (ds) DNA using ELISA.

Anti-double-stranded DNA (ds-DNA) ELISA

Plates were coated with 6 μg/mL ds-DNA in DNA Coating solution, washed with Tris-buffered saline (TBS) containing 0.05% Tween 20, and blocked with 1% BSA at 4ᵒC in 96-well plates. Following incubation, plates were washed and serum samples were incubated for an hour at room temperature on a plate shaker. Plates were washed and goat anti-mouse IgG HRP (horseradish peroxidase, SouthernBiotech, CAT#1030–05) and incubated for one hour. Plates were washed and TMB substrate was added, and absorbances for each well were read at 450 nm on a Biotek plate reader. Absorbance values were graphed using Graphpad Prism.

Vesicular Stomatitis Virus (VSV) Infection

Eight-week-old male mice were intranasally infected with 1×10⁶ plaque forming units per mouse of the New Jersey serotype of VSV. Viral stocks were generated from a progenitor provided by Dave Masopust via propagation in BHK cells. Viral titers were determined using Vero cells. Mice were euthanized after 24 hours. Lung homogenates were prepared using a gentle tissue homogenizer and were stored at −80°C. Homogenates were aliquoted in small tubes to avoid frequent freeze-thaw cycles.

Viral Plaque Assay

Vero cells were cultured and maintained in Dulbecco’s Modified Eagle Medium (DMEM). To titer virus, single cell suspensions of Vero cells were prepared using 0.25% trypsin and 0.2 million cells were seeded per well in 6-well plates. After overnight culture of cells, lung homogenates were thawed on ice and 10-fold serial dilutions were prepared. A range of sequential 10-fold dilutions were added to individual wells of Vero monolayers, followed by a 90-minute incubation and then overlay of cells with 0.5% SeaKem agarose in Eagle’s minimum essential medium (EMEM). Serially diluted aliquots of VSV stock were used as a positive control. To visualize the plaques, plates were stained with a second overlay of 0.5% agarose containing 1% neutral red on day 2. Plaques were counted within 24 hours of staining. Viral plaques were counted, whereafter wells with counts between 10 and 60 were used to quantity virus concentration in the original volume of lung homogenate using the following formula: (average number of plaques) × (overall dilution factor).

RESULTS

Highly prevalent coding variants account for lupus risk at the IRF7 locus

Genetic variants at the 11p15 locus are associated with lupus risk across ancestries^10,32,33. To identify the likely causal variant(s) underlying the association signal at this locus, we interrogated a subset of the Large Lupus Association Study 2 comprised of 150 genotyped variants at the 11p15 locus in a multi-ancestral cohort of 7,137 individuals with lupus and 6,329 controls⁸³. We first stratified the population into four groups based on ancestrally informative markers and applied imputation to incorporate variants that were not genotyped (see Methods). We then performed an association study in European and African ancestries that conferred 96% and 80% power, respectively, to identify the variants in the 11p15 locus underlying risk for lupus (see Table S1 and Methods). Analysis of other ancestral groups had insufficient power (<40%) due to the combined effects of lower minor allele prevalence and a small cohort size. We employed a logistic regression model with admixture estimates as covariates to identify highly associated variants (Figure 1). In the European ancestry population, a haplotype of 17 variants in strong linkage disequilibrium (LD) exceeded the statistical threshold of genome-wide significance (Figure 1A, Table 1). This same haplotype reached modest (but not genome-wide) significance in the population of individuals with African ancestry (Figure 1B, Table 1). This haplotype achieved a P-value of 3.58×10⁻¹⁰ in a meta-analysis of the European and African ancestry populations (Figure 1C, Table 1). Two coding variants, rs1131665 (Gln412Arg) and rs1061502 (NM_001572.5:c.856A>G (p.Lys179Glu), were found within this robustly associated haplotype (Figure 1D–E). The additive effect sizes of 1.15–1.20 (Table 1) for these two variants are consistent with prior studies^32,33. In contrast to most autoimmune-associated variants¹¹, the lupus risk haplotype for these variants is highly prevalent in multiple human populations, ranging from 70 to 98% across ancestries (Figure 1F).

Figure 1. A highly prevalent haplotype in IRF7 tagged by rs1131665 (Gln412Arg) and rs1061502 (Lys179Glu) shows genome-wide association with lupus.

(A-C) Genome-wide association signal of variants in the IRF7-PHRF1 locus. Each variant is represented as a data point in the context of its genomic location within the locus and is colored based on its P-value, with red dots representing genome-wide significant variants (P<5×10–8) and blue dots signifying variants that do not meet this threshold. rs1131665 and rs1061502, which are characterized in this study, are identified in A-C. The variants were assessed in a logistic regression model using admixture estimates as a covariate. Genomic position is provided using GRCh37 (hg19) coordinates. (A) European Ancestry, (B) African Ancestry, (C) Trans-ancestral Meta-Analysis. (D) Schematic of IRF7 (isoform A) protein domains: DBD: DNA Binding Domain; CAD: Constitutive Activation Domain; VAD: Virus-Activated Domain; ID: Inhibitory Domain; SRD: Signal Response Domain. (E) Heatmap showing variants in the IRF7 haplotype assessed in the trans-ancestral meta-analysis demonstrating perfect linkage disequilibrium of IRF7-PHRF1 variants; (F) Pie charts quantifying the frequency of the lupus risk (“T”) and protective (“C”) haplotypes globally and in specified ancestries (adapted from the Ensembl database website: https://www.ensembl.org). Ancestry abbreviations (defined by Ensembl Gene): ALL = Global Population; EUR = European Ancestry; AFR = African Ancestry; AMR = Admixed American Ancestry (native people of the Americas); EAS = East Asian Ancestry; SAS = South Asian Ancestry.

Table 1. Lupus association for the IRF7 haplotype block containing rs1131665 and rs1061502.

P-values are based on a logistic regression model performed in each ancestry with admixture estimates as covariates. Odds ratios are calculated in terms of the risk haplotype. The trans-ancestral meta-analysis was performed using an inverse-variance model using data from the European and African cohorts only. Identical results were obtained when each variant in the haplotype was assessed individually because of the perfect LD (r²=1) between the variants.

European Ancestry		African Ancestry		Trans-ancestral meta-analysis
P-Value	OR	P-Value	OR	P-Value	OR
4.84×10−8	1.20	0.0013	1.15	3.58×10−10	1.18

LD: Linkage Disequilibrium; OR: Odds Ratio.

To assess whether the haplotype containing Gln412Arg (rs1131665) and Lys179Glu (rs1061502) accounts for lupus risk at 11p15, we performed a stepwise logistic regression analysis starting with the genotype of rs1131665 as a covariate in the cohort of European ancestry. The genotype of rs1131665 was sufficient to account for the entire association with a significance level of P<0.01 at this locus in the European ancestry cohort (Figure S2A). We next performed a separate stepwise logistic regression including rs1061502 as a covariate. As expected, the genotype of rs1061502 was also sufficient to account for the lupus association (P<0.01) at this locus (Figure S2B). This association is consistent with the inheritance of these two coding variants in a haplotype given their complete linkage disequilibrium (LD) (r²=1.0, Figure 1E). This finding is especially important given that this risk haplotype is the most prevalent haplotype globally (Figure 1F). Collectively, our statistical analyses support a model in which lupus risk at the 11p15 (IRF7) locus is driven by the rs1131665+rs1061502 (IRF7 Gln412Arg + Lys179Glu) haplotype. In this study, we will refer to the Gln412+Lys179 haplotype as “lupus risk” and the Arg412+Glu179 haplotype as “lupus protective.”

A common lupus IRF7 risk haplotype is highly conserved in ancient DNA

The high prevalence of the lupus-associated haplotype in multiple ancestries is striking, particularly because the frequency of disease-associated variants is often kept low by natural selection^84–87. Ancient DNA studies assessing rapidly changing allele frequencies in populations before and after pandemics or plagues can be used to reveal selection of immune-related genes⁸⁸. To assess the conservation of this haplotype through time, we performed an analysis of publicly available datasets of ancient human DNA from the Ancient Genome Diversity Project - Allen Ancient DNA Resource (AADR) (see Methods)⁸⁹. We used rs11246213 as a tag for the rs1131665+rs1061502 haplotype in this analysis. This dataset has global sampling (Figure 2A), with genomes as old as ~100,000 years before present. As this dataset represents only ~1.23 million positions in the genome, our two lead SNPs were not represented. However, one of the 11p15 SNPs in LD with our two lead SNPs (perfect LD (R²=1) in EUR and AFR ancestry and nearly perfect (R²=0.95) in East Asian ancestry) was present in this dataset (rs11246213). Thus, this SNP was used to track population frequencies of this haplotype through time. Across the timeframe best captured in this dataset (~10,000 BCE to ~1,800 CE), the lupus risk haplotype was already present at high frequencies and remained largely consistent across ancestries. These findings indicate that the lupus-associated IRF7 haplotype has persisted for millennia in human populations. Given the uneven temporal and geographic sampling of ancient genomes, with sampling density within each bin shown in Table S2, we interpret these data as evidence of long-term persistence rather than as evidence for a specific selective event (Figure 2B).

Figure 2. The IRF7 lupus risk haplotype containing rs1131665 and rs1061502 have high population frequencies in the oldest human samples available, and these high frequencies are maintained over time across populations.

(A) Global map of ancient DNA sampling locations included in this analysis (compiled from the Ancient Genome Diversity Project). Colors indicate ancestries used in analyses (see Methods). Sampling is most comprehensive in West Eurasia (Red) and Central Asia and Siberia (Blue) due to long-term preservation of DNA in colder climates. World map is depicted using a Cylindrical Equal Area Projection. (B) Plot of the mean population frequencies of the SLE risk haplotype for each 500-year bin are shown (dots), and each population is fit with a weighted loess curve and 95% confidence intervals to show trends of the frequency of the SLE risk allele over time. Confidence intervals for Oceania and Africa were not calculated due to small sample sizes. Genomes older than 10,000 BCE were sparse (N=36 across all ancestries) and were not included in the analysis. Instead, mean frequencies of the haplotype in each ancestry between 43,000 BCE and 10,000 BCE are represented by boxes on the left-hand y-axis. Current allele frequencies are shown as boxes on the right (taken from Ensembl). Each black vertical line along the x-axis represents an ancient DNA sample that passed quality control measures as described in the Methods section. Ancestry colors are the same as in panel A. Exact sample counts for each 500-year bin and ancestry group are provided in Table S2.

The lupus-associated IRF7 haplotype increases IFN-α responses downstream of TLR7

To gain a biological understanding of the IRF7 lupus risk haplotype, we next examined the effects of the haplotype on IRF7 activity. IRF7 is activated downstream of numerous PRR pathways^15,16,18,90, resulting in nuclear localization, binding to DNA, and subsequent transcription of IFN-I (Figure 3A). We hypothesized that the lupus risk haplotype would increase interferon stimulated regulatory element (ISRE) activity and IFN-I production downstream of these innate immune pathways.

Figure 3. HEK Blue hTLR7 cells transfected with the IRF7 lupus risk haplotype exhibit higher IFN-α induction and secretion compared to other allelic combinations.

(A) Schematic of IRF7 activation, nuclear localization, and DNA binding downstream of multiple pattern recognition receptor pathways. (B) Overview of experimental approach employed in HEK Blue hTLR7 cells. C) Representative Western blot image of HEK Blue hTLR7 nuclear lysates following stimulation with R848 for 3-hours. β-actin and total H3 are included as loading controls. (D) Densitometry analysis of band intensities for risk and protective haplotype bands based on the quantitative Western blot presented in Figure S3. IRF7 bands were normalized to a total protein stain to calculate normalized signal as described in Methods. Welch’s t-test was used to determine significance. (E) Normalized ISRE luciferase activity (ratio of ISRE firefly to Renilla luciferase relative light units (RLU)) following 24-hours of R848 stimulation. Welch’s t-test was used to determine significance. (F) IFN-α measured in cell-free supernatants from cells stimulated with R848 for 3- and 24-hours. Bars represent mean, error bars represent ±SEM of triplicate samples (unless otherwise indicated) and are representative of at least three independent experiments. Significance was assessed with two-way ANOVA with a Holm-Šídák’s multiple comparisons correction. *P<0.05; **P< 0.01; ***P<0.005; ns indicates non-significant (P>0.05). ISRE: Interferon stimulated response element; EV: Empty Vector; R848: Resiquimod.

We first assessed IRF7 activity downstream of endosomal Toll-like Receptor 7 (TLR7), which is activated by host and viral nucleic acids^91,92. The role of TLR7 signaling in the etiology of lupus is well-established⁹³, and TLR7 gain-of-function and increased signaling alone is sufficient to cause a monogenic form of lupus⁹⁴. We employed a system using cells expressing one of the four possible combinations of IRF7 genotypes in HEK cells that stably express TLR7 (HEK Blue hTLR7) (Figure 3B). Since one of the amino acid changes within the risk haplotype (Gln412Arg) is located within the IRF7 inhibitory domain (Figure 1D), which controls IRF7 nuclear localization²¹, we hypothesized that the risk haplotype would increase nuclear localization of IRF7 downstream of TLR7. Indeed, when we stimulated HEK Blue hTLR7 cells expressing the IRF7 coding constructs with TLR7/8 agonist Resiquimod (R848), nuclear IRF7 was higher for the risk haplotype compared to the protective haplotype (Figure 3C–D, Figure S3). To determine if this difference in nuclear localization leads to genotype-dependent transcriptional activity, we next assessed R848-stimulated cells expressing the risk or protective haplotype of IRF7 in the context of a luciferase reporter containing three copies of the interferon stimulated response element (ISRE) regulatory sequence recognized by IRF7. We observed genotype-dependent ISRE regulatory activity, with the risk haplotype leading to increased ISRE transcriptional activity compared to the protective haplotype (Figure 3E). We next assessed whether this increased IRF7 activity leads to increased IFN-α protein secretion. Strikingly, we observed robust genotype-dependent differences in IFN-α production after stimulation with R848 (Figure 3F). Because total IRF7 protein levels were similar across genotypes in this HEK Blue hTLR7 transfection system (Figure 3C; Figure S3), the increased nuclear localization and activity observed for the lupus risk haplotype is unlikely to be explained by differences in overall IRF7 expression levels.

When we assessed combinations of the two mixed haplotypes (that rarely occur in humans), we found that neither combination fully replicated the increase in IRF7 nuclear localization observed for the risk-only compared to the protective-only haplotype, with the largest change coming from the 412 risk (Gln) variant (Figure 3C, Figure S3). We further found that the Gln412 risk allele alone increased IFN-α protein production (Figure S4A), ISRE reporter activity (Figure S4B), and IFN-I-responsive transcript abundance (Figure S4C–G).

Based on the additive biological impact of the two variants and the fact that the risk alleles of the haplotype are in perfect linkage disequilibrium and almost always co-occur in humans, we conclude that the risk haplotype contributes to heightened TLR7-dependent IFN-I production. Altogether, the results of these experiments establish that the lupus risk haplotype increases IRF7 nuclear localization, ISRE-based transcriptional regulation, and IFN-I secretion relative to the lupus protective haplotype.

THP-1 monocytes stably expressing physiological levels of IRF7 exhibit genotype-dependent interferon production

Given our results in a transient transfection system in HEK Blue hTLR7 cells, we next assessed the effects of the IRF7 lupus risk haplotype in monocytes, a cell type that is critical to the IFN-I signature in lupus^95,96. THP-1 monocytes were transduced at a multiplicity of infection (MOI) that results in IRF7 expression levels approximating physiological levels (Figure 4A, Figure S5). The IRF7 constructs used in this study employed an HA epitope tag and were tested for IFN-α induction downstream of TLR7 activation to ensure sustained IRF7 transcription factor activity (Figure S6).

Figure 4. THP-1 monocytes stably expressing the IRF7 lupus risk haplotype exhibit higher IFN-α induction at the transcript and protein levels.

(A) Schematic showing the generation of THP-1 monocytes stably expressing the lupus risk and protective haplotypes in IRF7. (B) Western blot assessment of IRF7 (total) and pIRF7 (nuclear) in KO, lupus risk, and lupus protective monocytes. (C) Quantitative PCR analysis of mRNA expression of IFN-α (subtypes IFNα2, IFNα25, and IFNα21) following transfection with Poly(I:C) at multiple time points in IRF7-deficient (KO; Knockout) cells engineered to express the risk or protective haplotypes of IRF7. (D) IFN-α measured in cell-free supernatants from cells transfected with Poly(I:C). (E) Differentially expressed genes in THP-1 monocytes expressing either the lupus risk or the lupus protective haplotypes 3-hours post-transfection of Poly(I:C). (F-I) Gene expression profiles [TPM values] of IFNA subset genes differentially expressed in IRF7 deficient (KO), lupus protective, and lupus risk monocytes 3-hours post transfection with Poly(I:C). (J) Pathway enrichment analysis of differentially expressed genes from E. (K) Motif enrichment for a representative IRF7 DNA binding motif for genes differentially expressed in the analysis presented in E. The motif image is taken from CIS-BP data build 2.00. Bars represent mean, error bars represent ±SEM of triplicate samples and are representative of at least three independent experiments, except the RNA-seq experiment in Figure 4E, which was performed once for all replicates. Statistical comparisons between risk and protective are based on a two-tailed unpaired Student’s t-test; *P<0.05; **P<0.01; ****P<0.0001; ns indicates non-significant (P>0.05).

Previous studies have shown that THP-1 monocytes produce IRF7-dependent IFN-α in response to a variety of agonists, including Poly(I:C)^97,98. We confirmed that Poly(I:C) transfection leads to IRF7 phosphorylation in our system and that risk and protective monocytes express equal amounts of total IRF7 (Figure 4B). THP-1 monocytes expressing the lupus risk haplotype produced >2-fold more IFN-α mRNA compared to the lupus protective haplotype at 3-hours (P<0.05) and 12-hours (P<0.01) post-transfection (Figure 4C). The genotype-dependence of increased IFN-α was attenuated and no longer statistically significant at 24 hours, likely due to post-transcriptional negative feedback of the signaling pathway (Figure 4C). The interferon-stimulated gene (ISG) response of THP-1 monocytes to exogenous universal IFN-I administration was not fully dependent on the presence of IRF7 (Figure S7), consistent with previous studies⁹⁹. This suggests that the IRF7 lupus risk haplotype amplifies the primary production of IFN-α, with indirect amplification of ISGs downstream of this IFN-I secretion. We also examined secreted IFN-α in our system. At 3 hours post transfection with Poly(I:C), no detectable IFN-α protein was observed (Figure 4D), consistent with the relatively low levels of mRNA induction at this time point (Figure 4C). Twenty-four hours after transfection, we observed a >2-fold (P<0.0001) increase of IFN-α protein production in cells expressing the risk haplotype compared to the protective haplotype (Figure 4D). This two-fold difference is consistent with our data obtained in the HEK Blue hTLR7 system as well as previous data³³. As IRF7 deficient cells failed to produce detectable IFN-α, these results support the role of IRF7 as the central regulator of IFN-α production downstream of PRR pathways in this cellular context.

To assess transcriptional differences driven by the lupus haplotype in an unbiased manner, we next performed RNA-seq on lupus risk, lupus protective, and IRF7-deficient monocytes 3-hours post-transfection with Poly(I:C) (Tables S3 and S4). As expected, Poly(I:C) stimulation is a bigger source of variation than genotype in these experiments (Figure S8). We observed 306 differentially expressed genes (DEGs, alpha threshold α=0.05, 1.5-fold or greater difference in expression levels), with 200 genes upregulated in the lupus risk haplotype and 106 genes upregulated in the protective haplotype (Figure 4E, Figure S9, Table S5). This genotype-dependence was not due to differences in IRF7 expression levels (Figure 4B) or differences in transcript levels of IRF7 or IRF3 (Figure S10). We observed significant upregulation of multiple IFNA transcripts in the lupus risk haplotype compared to the protective haplotype (Figure 4F–I), and IRF7 deficient cells did not produce IFNA transcripts, consistent with an IRF7 intrinsic effect. Pathway enrichment analysis of DEGs revealed significant enrichment for the “Interferon alpha/beta signaling” (P_adj= 3.9×10⁻¹³), “Interferon signaling” (P_adj= 8.7×10⁻⁹), and “IRF7 activation” (P_adj=3.2×10⁻⁴) pathways (Figure 4J, Table S6). When pathway analysis was performed separately on genes upregulated by the risk and protective haplotypes, we observed a similar pattern of enrichment in those genes upregulated by the risk haplotype, while genes upregulated by the protective haplotype were not enriched for these pathways (Figure S11). The promoters of DEGs upregulated by the risk haplotype were significantly enriched for predicted IRF7 binding sites, consistent with a mechanism involving stronger DNA binding of lupus risk IRF7 leading to stronger induction of gene expression (Figure 4K). Taken together, these data suggest that the IRF7 lupus risk haplotype drives a distinct transcriptional program in monocytes characterized by roughly two-fold higher IFN-α RNA and protein levels than the protective haplotype.

The IRF7 lupus risk haplotype increases DNA binding strength and alters the DNA sequence specificity of IRF7

Next, we sought to identify allele-dependent differences in IRF7 DNA binding. To this end, we used the PBM technology¹⁰⁰. This array-based method measures the DNA binding of a tagged transcription factor to thousands of double stranded DNA probes, providing a comprehensive survey of DNA-binding strength and specificity for a given protein. We used a previously published PBM array design that contains thousands of DNA elements recognized by IRF proteins (see Methods)⁷⁶. Because IRF7 requires post-translational modifications, and because we observed allele-dependent IRF7 nuclear localization (Figure 3C and D), we used the nextPBM technology, which profiles DNA binding using nuclear extracts¹⁰¹. We performed nextPBM experiments using extracts from untreated or Poly(I:C) transfected THP-1 monocytes expressing the IRF7 lupus risk haplotype or the protective haplotype attached to a HA-tag, as employed in the experiments in Figure 4.

The PBM probe intensities for the IRF7 lupus risk haplotype after Poly(I:C) transfection correlate strongly with those obtained in a previously published study for purified IRF7 homodimers (Pearson correlation coefficient (PCC) = 0.95)⁷⁶ (Figure 5A, Figure S12). The motif derived from the risk haplotype in the transfected condition matches the known motif for the IRF7 homodimer (Figure S13). DNA binding of the IRF7 lupus risk haplotype in the untreated condition was weaker than the Poly(I:C) condition, in agreement with lower levels of nuclear IRF7 in this condition (Figure S12), and was still highly concordant with IRF7 homodimer binding (PCC = 0.89). Strikingly, the binding strength of the IRF7 lupus risk and protective haplotypes were nearly identical in untreated conditions (Figure S12). In contrast, IRF7 binding strength for these haplotypes deviated dramatically with Poly(I:C) transfection, with a significant increase in binding of the risk haplotype to the IFN-type probes (2bp-type ISRE elements) but little change for the protective haplotype (Figure 5B, Figure S12). Distinct DNA binding preferences for the IRF7 lupus risk and protective haplotypes were observed for elements with 2- or 3-bp spacers between 5’-GAAA core elements (i.e., 5’-GAAAXXGAAA-3’ versus 5’-GAAANNNGAAA-3’) (Figure 5B). Further, motif analysis confirmed that IRF7 lupus risk haplotype binding agrees with that of the IRF7 homodimer, while the protective haplotype exhibits distinct base preferences (Figure S13). In particular, for the IRF7 lupus risk haplotype and IRF7 homodimer, we found stronger preference for a “C” nucleotide at positions 5 and 11, and a weaker preference for a “G” at position 16. Collectively, these results indicate that, in the Poly(I:C) transfected condition, the lupus risk-associated IRF7 binds more strongly to IFN-type sites in a manner consistent with an IRF7 homodimer. In contrast, while the lupus protective IRF7 also accumulates in the nucleus upon Poly(I:C) transfection, it exhibits weaker binding to IFN-type sites than the risk haplotype, along with altered DNA preferences at several base positions. While this could reflect direct changes in IRF7 interaction with DNA, the amino acid distinctions between risk and protective IRF7 variants may also alter interactions with other proteins in the nuclear extracts that shape DNA binding characteristics.

Figure 5. The IRF7 lupus risk haplotype has increased DNA binding strength and altered DNA binding specificity compared to the protective haplotype.

Comparison of DNA binding strength for the IRF7 lupus risk haplotype to purified IRF7 dimer (A) or the IRF7 protective haplotype (B). Binding for IRF7 lupus risk and protective haplotypes was determined via nextPBM assays using extracts from lupus risk and protective THP-1 monocytes following 3-hours of Poly(I:C) transfection. Data for the purified IRF7 dimer is from Andrilenas et al.⁷⁶, which used a mutant form of IRF7 that is constitutively dimeric. Binding is shown to 6,954 2-bp-spacer variant IRF7 binding sites (dark) and 3,456 3-bp-spacer variants (light) (see Methods) and is quantified by median PBM probe fluorescence values for each DNA sequence. (C) Increased binding of lupus risk haplotype IRF7 to the promoters of genes assessed on the PBM leads to increased gene expression (RNA-seq) of the 8 indicated genes. Log₂ fold change values for these genes are plotted along with the adjusted P-value from the RNA-seq analysis (Figure 4). *Padj<0.005, **Padj<0.0005, ***Padj<1.0×10–5.

We next compared allele-dependent IRF7 DNA binding events to allele-dependent gene expression levels. To this end, we used the subset of PBM array probes that contain DNA sequences representing the promoters of 26 previously identified interferon-responsive genes⁷⁶. Among these 26 genes, nine had significantly higher expression levels in the IRF7 lupus risk versus protective haplotype RNA-seq data, four were unchanged, none had lower expression, and 13 were not detected in THP-1 cells. These data are consistent with prior studies showing that THP-1 cells contain deletions in IFNA genomic regions, including IFNA1, IFNA2, IFNA5, IFNA6, IFNA8, and IFNA13¹⁰². Strikingly, eight of the nine “high in lupus risk” genes also have significantly stronger IRF7 risk haplotype binding to their promoters in Poly(I:C) transfected conditions (Figure 5C). Stronger binding of risk IRF7 to select interferon-responsive genes was also accompanied by higher transcript abundance of multiple IFNA genes (Figure 4F–I) and increased total secretion of IFN-α protein (Figure 4C). In contrast, only one of the four “unchanged” genes had higher IRF7 lupus risk haplotype binding (P=0.011, one-sided proportions test). These results suggest that the stronger overall binding of the IRF7 lupus risk haplotype directly impacts gene expression levels, leading to elevated expression of IFN-I genes including IFNA10 and IFNB1.

Human lung epithelial cells expressing the IRF7 lupus risk haplotype exhibit genotype-dependent IFN-α secretion

The IFN-I response is critical to mounting effective immune responses, especially in response to viruses¹³. We therefore hypothesized that the IRF7 lupus risk haplotype could also increase IFN-I responses in airway epithelial cells^25,103. We generated A549 cell lines stably expressing the risk and protective haplotypes using the same lentiviral vectors employed in the THP-1 monocyte system, starting with A549 cells deficient in IRF7 (Figure S14, see Methods). Because we could not use a chemical selection strategy, we performed single cell cloning to generate three independent clones of both the risk and protective haplotypes. Because A549 cells produce IRF7-dependent IFN-α in response to transfection of Poly(dA:dT)^90,104,105, we hypothesized that the risk haplotype would increase secreted IFN-α, as we observed with our monocyte and HEK Blue hTLR7 systems. Indeed, we again observed a 2-fold (P<0.005) difference in secreted IFN-α between the risk and protective haplotypes 24-hours after transfection with Poly(dA:dT) (Figure S14B), with no detectable IFN-α produced in the IRF7 deficient cells.

An engineered risk variant in mouse Irf7 enhances antiviral immunity and autoantibody production in a lupus model

Human IRF7 has 69% amino acid identity with murine Irf7¹⁰⁶. Importantly, the highest amino acid identity (>95%) is found in the DNA-binding and rs1131665-bearing inhibitory domains of IRF7. All inbred laboratory strains (e.g., C57BL/6J, BALB/cJ, 129S1/SvImJ, C3H/HeJ), as well as those genetically susceptible to autoimmunity (e.g., SJL/J and MRL/MpJ), are homozygous for the protective C/G allele at the mouse position orthologous to the human IRF7 variant rs1131665. Thus, both humans and mice with the protective variant are characterized by an arginine residue in the IRF7/Irf7 inhibitory domain at positions 412 and 334, respectively. Using CRISPR targeting and homologous recombination, we mutated C/G to T/A at position 141,263,661 on chromosome 7 (Genome Reference Consortium Mouse Build 38 or mm10), resulting in an Arg334Gln substitution in mouse Irf7 that is homologous to the rs1131665 SLE risk-variant Arg412Gln substitution in human IRF7.

To test the hypothesis that homologous lupus-risk variants in mouse Irf7 would exacerbate risk for autoimmune disease, we applied an epicutaneous resiquimod (R848) model that can trigger lupus-like autoimmunity in non-autoimmune prone mouse strains (e.g., C57BL/6)^107,108 (Figure 6A). After eight weeks of R848 treatment, mice with the lupus-risk Irf7 homologue (Gln-Gln) developed significantly more anti-double stranded DNA autoantibodies compared to mice with the lupus-protective Irf7 homologue, Arg-Arg (Figure 6B).

Figure 6. Mice with the Irf7 lupus risk variant develop autoantibodies in a model of lupus and have a lower viral titer with Vesicular stomatitis virus infection.

Using CRISPR targeting and homologous recombination, we mutated C/G to T/A at position 141,263,661 on chromosome 7 (Genome Reference Consortium Mouse Build 38), resulting in an Arg334Gln substitution in mouse Irf7 that is analogous to the rs1131665 SLE risk-variant Arg412Gln substitution in human IRF7. (A) 100 micrograms of resiquimod (R848) was epicutaneously applied three times per week to the ears of genome edited C57BL/6 mice with the lupus risk or non-risk IRF7 genotype for a total of eight weeks. (B) Sera was collected and used to quantify anti-double-stranded DNA (ds-DNA) antibodies by ELISA. Each point represents one mouse; mean ± SD is shown. Statistical significance was assessed using a two-tailed Mann–Whitney test (P = 0.0043). (C) Eight-week-old male mice with the lupus risk or non-risk Irf7 genotype were intranasally infected with 1 million plaque-forming units of VSV-NJ. (D) After 24 hours, virus was quantified in lung homogenates using viral plaque assays. Each point represents one mouse; mean ± SD is shown. Statistical significance was assessed using a two-tailed Welch’s t-test (P = 0.016). (E) Model for the lupus risk genotype in IRF7 enhancing both viral responses (with the lupus risk genotype enhancing viral clearance) and autoimmunity (with the lupus risk genotype enhancing the production of autoantibodies). *P<0.05; **P<0.01.

We next tested the hypothesis that the Irf7 lupus-risk homologue could enhance innate immunity against virus infection (Figure 6C). After pulmonary infections with the interferon-sensitive vesicular stomatitis virus (VSV), IFN-sufficient mice (e.g., C57BL/6) rapidly eliminate replicating virus from the lungs in an IFN-dependent manner^15,109. We found that mice with the lupus-risk Irf7 homologue (Gln-Gln) exhibited significantly lower titers of VSV in lungs one day after infection compared to non-risk Arg-Arg control mice (Figure 6D).

Taken together, these results suggest that the lupus risk haplotype increases IRF7 activity in a manner that can enhance antiviral defense but also increase risk for autoimmunity (Figure 6E). These results offer a plausible biological explanation for the strong conservation of this haplotype in ancient DNA (Figure 2).

DISCUSSION

Lupus arises from a convergence of genetic and environmental factors. The mechanisms by which specific risk alleles perturb immune regulation remain incompletely defined^110–112. Here, we focused on the highly prevalent IRF7 coding haplotype associated with lupus risk in modern populations and conserved in ancient genomes, delineating its impact on IRF7-dependent IFN-I responses. Across multiple cellular contexts, the lupus risk haplotype increases IRF7 gene regulatory activity, culminating in augmented IFN-α transcription and secretion.

There is substantial evidence that IFN-I drives disease in lupus and plays a central role in the control of many viral infections^95,96,113. Consistent with this dual role, mice harboring the Irf7 lupus risk variant that enhances IFN production exhibited both elevated autoantibody titers in response to autoimmune triggers and improved control of pulmonary virus infections. The frequency of the lupus-associated IRF7 haplotype differs across present-day ancestral groups, with lower frequency in African ancestry populations than in some non-African populations. Our data do not identify the cause of this difference. Possible contributors include demographic history, differences in haplotype structure and linkage disequilibrium, and broader ancestral genetic diversity. In addition, because the available ancient DNA data capture only a limited time window and are unevenly distributed across populations, they do not resolve when or why these frequency differences arose. Accordingly, we interpret the ancestry-specific frequency patterns cautiously. The persistence of this lupus-associated IRF7 haplotype in ancient and modern populations is notable because many disease-associated alleles are uncommon. Our ancient DNA analysis indicates that this haplotype was already common during the period captured by available ancient genomes and remained so across time. These data do not identify the selective pressure responsible for its rise in frequency, and the uneven geographic and temporal distribution of ancient samples warrants caution in interpretation. However, when considered together with our functional data showing enhanced IFN-I induction and improved antiviral control, these findings are consistent with a model in which this common IRF7 haplotype confers an immunologic benefit that may help explain its long-term persistence despite increasing risk for lupus.

IRF7 itself is an interferon-responsive gene whose expression is upregulated downstream of IFN-I signaling. In this study, we introduced variants of IRF7 whose expression was not altered by IFN-I so that we could decouple the immunologic consequences of increased IRF7 transcription factor activity from those of increased IFN-I production and signaling. Using this system, we observed consistent ~2-fold or greater amplification of IFN-I production from cells expressing the lupus risk haplotype compared to those with the protective haplotype. In physiological systems, where increased IFN-I would be expected to further induce IRF7 expression, these primary differences are likely amplified through positive feedback loops that increase IRF7 levels and enhance IFN responses. Although we did not use pDCs directly in the present study, these cells are among the most relevant contexts in which to consider the impact of this haplotype because they depend strongly on IRF7 for IFN-α production⁹⁸. We therefore expect that the gain-of-function effects of the lupus-associated IRF7 haplotype identified here will be relevant to pDC biology as well.

The genotype-dependent increase in IFN-α observed in our cell-based systems was most apparent at early stimulation time points. This does not argue against relevance to the chronic interferon signature of lupus. Rather, our findings support a model in which the IRF7 lupus risk haplotype increases the magnitude of IFN-I production each time nucleic-acid sensing pathways are engaged. In the setting of lupus, where endogenous nucleic acids, immune complexes, and inflammatory feed-forward circuits can repeatedly activate these pathways, recurrent higher-amplitude IFN-I responses could cumulatively sustain chronic interferon-associated immune activation.

In models of autoimmunity and in patients with lupus, IRF7 is involved in the enhanced production of IFN-I, contributing to autoantibody production, tissue damage, and inflammation^31,33,34,114. Our findings extend these observations by demonstrating that a common IRF7 coding haplotype quantitatively tunes IFN-I output in cell types central to lupus pathogenesis, including monocytes and airway epithelial cells, and that the corresponding murine homologue augments autoantibody production in a TLR7–driven lupus model. These results fit within a broader framework in which endosomal TLR7 gain-of-function is sufficient to cause lupus in humans⁹⁴, and in which monocytes and plasmacytoid dendritic cells act as major sources of IRF7-dependent IFN-I in both infection and autoimmunity^95,96,113. Amplified IFN-I can arise downstream of both pathogens and self-nucleic acids, each contributing to chronic pathway activation in lupus^115–117. IRF7-mediated type I interferon signaling promotes lupus across compartments, driving autoreactive germinal center and plasma cell responses in B cells¹¹⁸. In parallel, IRF7 and STAT1 are required for control of infection with IFN-sensitive viruses, including VSV and influenza A virus⁹⁸, consistent with our observation that the lupus risk Irf7 allele increases innate control of pulmonary VSV infection.

IRF7 thus emerges as a key nodal point integrating genetic risk for lupus with host defense against viral pathogens. Loss-of-function mutations in IRF7 affect susceptibility to viral infections such as SARS-CoV-2, influenza A virus, and hepatitis C by attenuating transcriptional amplification of type I IFN genes^24,26,27. At the opposite end of this functional spectrum, our data indicate that a gain-of-function IRF7 haplotype enhances IFN-I induction and accelerates viral clearance, but at the cost of increased autoantibody production. We initially hypothesized that a selective pressure may have led to the increased IRF7 lupus risk frequency in the East Asian ancestry dataset (98%) compared to African ancestry (42%). However, the risk allele frequency was consistently high throughout the sampling period starting from the Antonine Plague of CE 165 to 180 in the Roman Empire and including the cycles of Bubonic Plague in the 14^th-18^th centuries. Thus, the change in frequency from a presumably lower ancestral population must have occurred before the time frame captured in ancient DNA genomes. The protective role of the lupus risk haplotype of IRF7 that we observe in IFN-I production is consistent with both the effect size of increased lupus risk and the persistence of this haplotype in human populations over time.

DATA AND CODE AVAILABILITY

Cell lines and constructs generated in this study are available from the corresponding authors upon request. Sperm from the Irf7 Arg334Gln knock-in mouse line generated in this study is available from the corresponding authors upon reasonable request. RNA-seq data are available in the NCBI Gene Expression Omnibus under accession GSE312190, and a UCSC Genome Browser session is available at https://genome.ucsc.edu/s/Ledsall/IRF7. All other data are included in the article and supplemental information. No custom code was generated for this study.