IRF5 Risk Polymorphisms Contribute to Inter-Individual Variance in Pattern-Recognition Receptor-Mediated Cytokine Secretion in Human Monocyte-Derived Cells

Abstract

Monocyte-derived cells display highly variable cytokine secretion upon pattern-recognition receptor (PRR) stimulation across individuals; such variability likely affects inter-individual inflammatory/autoimmune disease susceptibility. To define mechanisms for this heterogeneity, we examined pattern recognition receptor (PRR)-induced monocyte-derived-cell cytokine secretion from a large cohort healthy individuals. Although cytokine secretion ranged widely among individuals, the magnitude of cytokine induction after individual Nod2 and TLR2 stimulation (a cohort of 86 individuals) or stimulation of multiple TLRs (a cohort of 77 individuals), either alone or in combination with Nod2, was consistent intra-individually across these stimuli. Nod2 and TLRs signal through interferon-regulatory-factor-5 (IRF5) and common IRF5 polymorphisms confer risk for autoimmunity. We find that cells from rs2004640 IRF5 risk-associated allele carriers secrete increased cytokines upon individual or synergistic PRR stimulation in a gene dose- and ligand dose-dependent manner in both monocyte-derived dendritic cells and macrophages. IRF5 expression knockdown in IRF5-risk-allele carrier cells significantly decreases PRR-induced cytokines. Moreover, we find that IRF5 knockdown profoundly decreases Nod2-mediated MAPK and NF-κB pathway activation, whereas the PI3K and mTOR pathways are not impaired. Finally, the IRF5 rs2004640 polymorphism is a major determinant of the variance (r²=0.53) in Nod2-induced cytokine secretion by monocyte-derived cells from different individuals. We therefore show a profound contribution of a single gene to the variance in inter-individual PRR-induced cytokines. The hyper-responsiveness of IRF5 disease-associated polymorphisms to a wide spectrum of microbial triggers has broad implications on global immunological responses, host defenses against pathogens and inflammatory/autoimmune disease susceptibility.

Introduction

The interplay between microbial and genetic susceptibility factors is particularly important for development of autoimmune/inflammatory diseases. Inter-individual differences in host responses to microbes confer some of the largest risk effects in autoimmunity; pattern recognition receptor (PRR) pathway associations are one such example. For instance, the sensor of bacterial peptidoglycan Nod2 is associated to Crohn’s disease(1), a subtype of inflammatory bowel disease (IBD). Although IBD has traditionally been associated with dysregulation of responses to bacteria(2), recent studies show that viruses also play a role in modulating intestinal inflammation(3); therefore, polymorphisms in PRR pathways that modulate both bacterial and viral responses may have a particularly significant role in IBD risk. Additional evidence for microbial responses contributing to autoimmune/inflammatory diseases include identification of microbial signature pathways (e.g. Type I IFN) (4), and genetic associations in microbial response pathways including IFIH1 in type I diabetes mellitus (T1DM) (5) and IRF5 in multiple autoimmune/inflammatory diseases (including systemic lupus erythematosus (SLE), IBD, rheumatoid arthritis, Sjogren's syndrome, primary biliary cirrhosis, systemic sclerosis and multiple sclerosis (MS)) (6–11). An important consequence of PRR activation by bacterial and viral products is induction of cytokine secretion, and autoimmune/inflammatory diseases are characterized by dysregulated cytokines(1,12,13).

In this study we sought to define the nature and etiology of differences in PRR-mediated cytokine induction across a large cohort of healthy individuals. We performed our study in healthy individuals given the increased or dysregulated cytokine production in autoimmune/inflammatory diseases such as IBD and SLE (1,12,13), and that immune-modulating medications administered for these diseases can alter cytokine regulation. We and others have found that PRR stimulation of primary human monocyte-derived cells shows large inter-individual variability in the magnitude of cytokine induction(14–16). We now establish that this variability is consistent within a given individual over time, in monocyte-derived dendritic cells (MDDC) and macrophages (MDM), over a wide range of Nod2 and TLR2 ligand concentrations, and between distinct cytokines. Importantly, we establish that the rs2004640 disease-associated risk allele in the IRF5 gene region leads to significantly increased individual and synergistic Nod2- and TLR-induced cytokine secretion and is a major contributor to the variance in PRR-induced cytokine secretion across individuals. Significantly, IRF5 knockdown in human MDDC results in decreased PRR-mediated MAPK and NF-κB pathway activation and cytokine secretion. That select IRF5 polymorphisms dramatically modulate Nod2- and TLR-induced cytokine protein secretion and are a uniquely major determinant of the variance in Nod2-induced cytokine secretion between individuals, highlights one mechanism whereby these polymorphisms may contribute to autoimmune/inflammatory diseases.

Materials and Methods

Patient recruitment and genotyping

Informed consent was obtained per protocol approved by the institutional review board at Yale University. We recruited participants with no personal or family history of autoimmune/inflammatory disease including psoriasis, SLE, rheumatoid arthritis, MS, T1DM, Crohn’s disease, and ulcerative colitis, or history of HIV. Additional self-reported information collected included gender, race and age. Our cohort included individuals of self-declared European (88%), African (7%), Asian (2%) or other ancestries (3%). Two separate cohorts of 86 and 77 individuals were recruited for Nod2/TLR2 dose-response studies in MDDC and MDM, and Nod2/TLR synergy studies in MDDC, respectively. We selected 35 individuals to overlap between the two cohorts so as to examine a subset of the responses at two distinct times. Based on this treatment design, MDDC from 128 non-overlapping individuals were treated with 1 µg/ml MDP. Twenty-five Crohn’s disease-risk polymorphisms (10,17–19) were initially selected for analysis based on their location in regions containing genes with known or potential roles in mediating PRR responses. Genotyping was performed by TaqMan SNP genotyping (Applied Biosystems, Foster City, CA) or Sequenom platform (Sequenom Inc., San Diego, CA); individuals were sequenced for the IRF5 CGGGG-indel polymorphism at the Yale University Keck facility.

Primary MDM and MDDC cell culture

Monocytes were purified from human peripheral blood mononuclear cells by positive CD14 selection (Miltenyi Biotec, Auburn, CA), tested for purity and cultured for 7 days with IL-4 (40ng/ml) and GM-CSF (40 ng/ml) (R&D Systems Inc. Minneapolis, MI, USA) (for MDDC differentiation) or with 10 ng/ml M-CSF (for MDM differentiation).

MDM stimulation

Cultured MDM or MDDC from a cohort of 86 individuals were treated with increasing doses of MDP (Bachem, King of Prussia, PA) or Pam₃Cys (Calbiochem, La Jolla, CA) for 24 h. MDDC from a cohort of 77 individuals were treated with a single dose of MDP, Pam₃Cys, lipid A (Peptides International, Louisville, KY), flagellin, CL097, CpG or poly I:C (Invivogen, San Diego, CA) for 24 h alone or in combination. Following the treatment of the cells, supernatants were frozen and processed in large batches. Supernatants were assayed for TNF-α, IL-8 (BD Biosciences), IL-12p40 and IL-10 (R&D Systems) or IL-1β (Pierce Biotechnology, Rockford, IL) by ELISA.

Constructions and transfection of small interfering RNAs (siRNAs)

300 nM scrambled siRNA or siRNA against IRF5 (Dharmacon, Lafayette, CO) were transfected into MDDC using Amaxa nucleofector technology (Amaxa, San Diego, CA). Cells were cultured for an additional 48 h and then treated as indicated.

Phosphoprotein activation

Phosphorylation of MAPKs, Akt, p70-S6K and IκBα was determined by flow cytometry using Alexa Fluor 647-labeled phospho-ERK, phospho-p38, phospho-JNK, phospho-Akt, phospho-p70-S6K and phospho-IκBα (Cell Signaling, Danvers, MA) along with isotype controls.

mRNA expression analysis

Following stimulation, total RNA was isolated, reverse transcribed and quantitative PCR performed as in (20) on the ABI Prism 7000 (Applied Biosystems). Each sample was run in duplicate and normalized to GAPDH. Primers sequences are available upon request.

Statistical analysis

Correlation was calculated using the Spearman rank correlation coefficient. Significance for differences in cytokine protein, cytokine mRNA and phosphoprotein induction was assessed using t-test. p < 0.05 was considered significant. The effects of polymorphisms on PRR-induced cytokine secretion were analyzed by ordinal logistic regression and a forward stepwise strategy was used to select the polymorphisms remaining in the final model. As the pseudo r² obtained from ordinal logistic regression cannot be directly interpreted as a measure of percent of variance, we performed linear regression applying an additive model between the three genotypes and utilized r² to determine inter-individual variance in cytokine secretion. Haplotype phasing was performed utilizing PLINK package version 1.07 (21).

Results

Nod2 stimulation results in highly variable cytokine expression across individuals

To define the spectrum of inter-individual differences in PRR responses in primary monocyte-derived cells we utilized well-powered cohorts of healthy individuals. Given the prominent role of nucleotide oligomerization domain 2 (Nod2) in host responses to microbes and Crohn’s disease susceptibility(1), we examined the range and magnitude of cytokine induction to Nod2 stimulation by its ligand muramyl dipeptide (MDP)(22,23) in MDDC. MDDC phenotype was verified by absence of B cell and T cell markers (Supplementary Fig. 1A), upregulation of the surface markers HLA-DR, CD1c, CD1d, CD11b, CD11c, CD33, CD40, CD45RO, CD54, CD80, CD83 and CD86, and downregulation of CD14 relative to monocytes (Supplementary Fig. 1B&C). We stimulated cells from 128 healthy controls with 1 µg/ml of the muramyl dipeptide for 24 h and measured TNF-α secretion. Unstimulated cells show very low baseline cytokine secretion, and MDP stimulation induced a wide range of TNF-α secretion between individuals, ranging from no increase to 1440-fold increase over no treatment (Fig 1A). To better assess the response distribution, we reanalyzed the values using log2 transformation (Fig. 1B), and we use log2 transformed data in our subsequent figures. Interestingly, cytokine secretion levels cluster into two groups: a group with very low TNF-α secretion to MDP stimulation and a group that shows a normalized distribution over moderate to high TNF-α secretion levels (Fig. 1B). Similar to our results, prior studies using monocytes identify low and high responders to Toll-like receptor (TLR) stimuli(14–16). We found poor correlation between the magnitude of TNF-α induction to age (data not shown). Whole blood studies observed decreased cytokines upon TLR stimulation when specifically comparing separate cohorts of older (>65 years) to younger (21–32 years) individuals(15). The lack of age to cytokine level correlation in our study is likely due to the low number of ‘older’ individuals.

Figure 1. Nod2 stimulation of primary human MDDC and MDM results in inter-individual heterogeneity of cytokine secretion, and MDDC from the same individual show similar cytokine induction at separate time points.

Human MDDC from 128 healthy individuals were stimulated with 1 µg/ml MDP for 24 h. Supernatants were assayed for TNF-α, IL-12p40, IL-8, IL-1β and IL-10. The data are represented as the (A) scatter plot of fold TNF-α induction upon Nod2 stimulation and (B) histogram of fold TNF-α induction (log2 transformed) upon Nod2 stimulation compared to the cytokine levels of untreated cells. (C) Correlation plot of TNF-α and IL-12p40 fold-secretion (log₂ transformed) across different individuals with Spearman correlation coefficient (r) indicated. (D) Table summarizing the correlation between fold cytokine induction (log₂ transformed) for various cytokines as calculated by Spearman correlation coefficient. (E) Correlation of TNF-α secretion (log₂ transformed) between MDM and MDDC generated from a subset of 86 individuals with Spearman correlation coefficient indicated. (F, G) Human MDDC (n=35) from the same individuals drawn at two different time points within a six month period were stimulated with 1 µg/ml MDP for 24 h. Supernatants were assayed for TNF-α, IL-12p40, IL-8, IL-1β and IL-10. (F) Shown is a correlation plot of IL-12p40 fold-induction (log₂ transformed) relative to untreated cells at two different times with Spearman correlation coefficient indicated. (G) Table summarizing the correlation between cytokine induction at two different times as calculated by Spearman correlation coefficient values. ***; p<0.001; †, p<1×10⁻⁴; ††, p<1×10⁻⁵. Gray dot (indicated by arrow) represents (C) 46, (E) 31 and (F) 14 clustered individuals who showed minimal cytokine response following Nod2 stimulation. r, Spearman rho coefficient.

Magnitude of cytokine secretion shows intra-individual correlation across multiple cytokines and different cell types

Importantly, upon Nod2 stimulation the magnitude of induction for TNF-α and other pro-inflammatory cytokines, including IL-12p40 (Fig. 1C&D), IL-8 and IL-1β (Fig. 1D), significantly correlates within each individual. These cytokines were chosen as they are upregulated in a number of autoimmune/inflammatory diseases, including Crohn’s disease (2, 24), and they are secreted upon MDP stimulation of monocyte-derived cells (14,25–27). As anti-inflammatory cytokines suppress pro-inflammatory cytokines, we questioned if these two cytokine types inversely correlate upon Nod2 stimulation. Interestingly, we find that IL-10 induction levels are similar to those of pro-inflammatory cytokines in each individual (Fig. 1D). To assess if results were similar in other cell types, we examined monocyte-derived macrophages from a subset of 86 individuals and find excellent correlation in the Nod2-induced TNF-α levels secreted by MDM and MDDC from the same individual (Fig. 1E). Importantly, we found significant intra-individual correlation in MDP-induced cytokine secretion in MDDCs isolated upon repeated blood draws (Fig. 1F&G). Taken together, Nod2 stimulation induces differential inter-individual cytokine secretion, and the magnitude of induction of different cytokines, in different monocyte-derived cells and over time is highly correlated within each individual.

The magnitude of cytokine secretion shows correlation across Nod2 and TLR2 ligand doses

We next questioned if higher MDP doses abolish the inter-individual differences in secreted cytokine levels upon Nod2 stimulation. We therefore stimulated MDDC from a cohort of 86 individuals with 1, 10 and 100 µg/ml MDP. While there is an MDP dose-dependent induction of TNF-α (Fig. 2A), we observe excellent correlation in the magnitude of cytokine secretion between the MDP doses within each individual (Fig. 2B). Therefore, individuals that cluster in the lower range of cytokine secretion with low MDP dose stimulation, also cluster in the lower range of cytokine secretion upon higher MDP dose stimulations.

Figure 2. The magnitude of TNF-α induction by MDDC shows correlation across Nod2 and TLR2 ligand doses.

Human MDDC (n=86) were stimulated for 24 h with 1, 10 or 100 µg/ml of MDP (A, B) or Pam₃Cys (C, D). Supernatants were assayed for TNF-α. (A, C) Graphs show fold TNF-α induction (log₂-transformed) upon Nod2 or TLR2 stimulation compared with the cytokine levels of untreated cells + SEM. (B, D) Spearman correlation coefficients of induction of TNF-α (log₂-transformed) across indicated MDP and Pam₃Cys doses. ††, p<1×10⁻⁵. (E) Correlation of the fold TNF-α induction (log₂ transformed) comparing 1 µg/ml MDP and 1 µg/ml Pam₃Cys stimulation with Spearman correlation coefficient indicated. Gray dot (indicated by arrow) represents 27 clustered individuals who showed minimal cytokine response following stimulation. r, Spearman rho coefficient.

The Gram-positive bacterial cell wall contains both the Nod2 ligand peptidoglycan and the TLR2 ligand lipotechoic acid such that Gram-positive infection activates both receptors. Therefore, we investigated Pam₃Cys-mediated TLR2 stimulation of MDDC. Similar to Nod2 stimulation, we observe inter-individual variability upon TLR2 stimulation of MDDC with a Pam₃Cys dose-dependent induction of TNF-α (Fig. 2C). Moreover, there was excellent correlation in the degree of TNF-α secretion between the Pam₃Cys doses within each individual (Fig. 2D). Importantly, we observe significant correlation in TNF-α induction levels between Nod2 and TLR2 stimulation in each individual (r=0.86) (Fig. 2E). Given this striking correlation, we considered a genetic contribution to the inter-individual variability in cytokine secretion in pathways downstream PRRs.

A polymorphism in IRF5 gene shows strong association with inter-individual variation in TNF-α production

We assessed a subset of Crohn’s disease-associated polymorphisms in regions that included genes likely to regulate PRR signaling with respect to cytokine induction. As allele frequencies vary between ancestry groups, we focused this analysis on MDDC from our most represented ancestry within the overall cohort, 111 European-ancestry individuals. We used regression to examine TNF-α secretion by MDDC upon stimulation with 1 µg/ml MDP. Of the 25 polymorphisms examined, we observed highly significant evidence for the effects of only the association examined within the IRF5 region, rs2004640 (uncorrected p-value = 3.7×10⁻¹³; corrected p-value = 9.2×10⁻¹¹) (Table I). Importantly, the r² or variance explained by the IRF5 rs2004640 genotype on TNF-α secretion is 0.53. This indicates that the IRF5 rs2004640 genotype is a major genetic factor modulating the inter-individual variance in host cytokine induction upon Nod2 stimulation. We therefore decided to more closely examine the relationship of the IRF5 disease-associated polymorphisms with PRR-induced cytokine secretion.

Table I.

Contribution of Crohn's disease-associated polymorphisms to Nod2-induced cytokine secretion in MDDC

dbSNP ID	Chr.	Risk alleleA	Expected allele frequency in European controlsB	Observed allele frequency	Major allele homoz (n)	Het (n)	Minor allele homoz (n)	Corrected p value	r2	Positional candidate gene of interest
rs2476601	1p13	G	0.833	0.923	96	13	2	>0.05	4.00×10−3	PTPN22
rs4656940	1q23	A	0.814–0.833	0.860	83	25	3	>0.05	6.04×10−5	ITLN1,CD244
rs10181042	2p16	T	0.412–0.432	0.437	33	59	19	>0.05	1.63×10−2	C2orf74,REL
rs6738825	2q33	A	0.456–0.521	0.500	31	49	31	>0.05	3.41×10−3	PLCL1
rs3792109	2q37	A	0.517–0.604	0.545	42	37	32	>0.05	6.20×10−4	ATG16L1
rs11167764	5q31	C	0.783–0.825	0.766	64	42	5	>0.05	1.43×10−2	NDFIP1
rs7714584	5q33	T	0.042–0.044	0.081	93	18	0	>0.05	1.96×10−4	IRGM
rs359457	5q35	T	0.592–0.617	0.536	30	59	22	>0.05	3.06×10−4	CPEB4
rs1847472	6q15	G	0.620–0.690	0.649	47	50	14	>0.05	3.04×10−4	BACH2
rs1456896	7p12	T	0.608–0.658	0.662	51	45	15	>0.05	4.43×10−3	IKZF1,ZPBP,FIGNL1
rs2004640	7q32	T	0.440–0.560	0.468	36	46	29	9.2×10−11	0.53	IRF5
rs4077515	9q34	T	0.492–0.518	0.550	38	46	27	>0.05	2.41×10−4	CARD9,SNAPC4
rs1250550	10q22	G	0.633–0.700	0.662	49	49	13	>0.05	3.40×10−4	ZMIZ1
rs4409764	10q24	T	0.455–0.483	0.486	25	64	22	>0.05	1.01×10−2	NKX-2
rs102275	11q12	C	0.230–0.392	0.306	52	50	9	>0.05	9.73×10−4	FADS1
rs694739	11q13	A	0.608–0.649	0.671	51	47	13	>0.05	1.28×10−2	PRDX5,ESRRA
rs11564258	12q12	A	0.017–0.027	0.058	98	13	0	>0.05	5.80×10−3	MUC19,LRRK2
rs4902642	14q24	G	0.567–0.603	0.572	36	55	20	>0.05	3.01×10−3	ZFP36L1
rs2066844	16q12	T	0.021–0.083	0.045	101	10	0	>0.05	6.00×10−3	NOD2
rs2066845	16q12	C	0.043–0.150	0.049	100	11	0	>0.05	9.10×10−3	NOD2
rs5743293	16q12	insC	0.025	0.040	102	9	0	>0.05	1.59×10−2	NOD2
rs2872507	17q21	A	0.458–0.475	0.455	30	61	20	>0.05	3.16×10−2	GSMDL,ZPBP2,PRMDL3,IKZF3
rs1893217	18p11	G	0.115–0.158	0.131	83	27	1	>0.05	5.53×10−3	PTPN2
rs740495	19p13	G	0.233–0.261	0.243	68	32	11	>0.05	5.67×10−4	GPX4,SBNO2
rs2413583	22q13	C	0.842–0.854	0.865	82	28	1	>0.05	3.28×10−4	MAP3K71P1

^AThe risk allele is for independent Crohn's disease-associated polymorphisms (or SNPs in LD) (References 6,17)

^BAs reported in dbSNP version 135 (Oct 12, 2011) and References 8–10

Regression analysis was conducted to examine potential PRR-response-regulating genes associated to Crohn's disease for the effects on TNF-α secretion upon 1 µg/ml MDP stimulation of MDDC derived from 111 European ancestry healthy controls. p values were adjusted by Bonferroni correction for multiple comparisons.

Our cohort represents random healthy controls. Therefore, some disease-associated or less frequent polymorphisms may not be well represented. For this reason, polymorphisms in genes such as Nod2, known to regulate cytokines upon MDP stimulation References 22, 23), may not be adequately powered in this cohort to detect differences, in particular if the effects are seen mainly in minor allele homozygote carriers. Chr., chromosome; homoz., homozygotes; het., heterozygotes.

The disease-associated rs2004640 and CGGGG-indel IRF5 risk alleles show significantly increased cytokine secretion following Nod2 or TLR2 stimulation

Previous studies found that the commonly distributed rs2004640 T allele is associated with multiple autoimmune diseases including SLE (odds ratio 1.31–1.69)(6–11). The rs2004640 T allele, located at the intron-exon border of exon 1B in the IRF5 gene, is proposed to create a splice site resulting in expression of the variant exon 1B IRF5 mRNA transcript and increased IRF5 mRNA expression (8,9). IRF5 disease-associations have been shown in multiple ancestries, including African American, Chinese Han, Japanese and European ancestries (7,9,28,29), such that we initially included individuals of all ancestries collected in our study to examine the functional immunological outcomes of this polymorphism. We find that MDDC (Fig. 3A) and MDM (Fig. 4) from rs2004640 T risk allele carriers significantly increase TNF-α protein induction upon MDP and Pam₃Cys stimulation at 1 µg/ml. We observe a dramatic gene dose effect, with cells from GT heterozygotes demonstrating cytokine secretion intermediate between TT and GG homozygotes (Fig. 3A & Fig. 4). We next asked if increasing the doses of MDP and Pam₃Cys could overcome the wide differential in TNF-α secretion between different genotypes. While the greatest differential between the rs2004640 alleles was observed at low MDP and Pam₃Cys doses, a significant gene dose effect remained even at the highest PRR ligand concentrations examined (Fig. 3A & Fig. 4). Importantly, at low ligand doses, cells from individuals homozygous for the protective GG allele secreted negligible cytokine amounts in MDDC (Fig. 3A) and MDM (Fig. 4). Of note is that untreated cells secrete negligible levels of cytokines, and the effects of the rs2004640 genotype on cytokine secretion are observed specifically upon PRR stimulation of MDDC (Supplementary Fig. 2A). Re-analysis of the correlation plot shown in Fig. 1E further demonstrates that rs2004640 IRF5 GG carriers mostly corresponded to low cytokine producers, and rs2004640 IRF5 TT carriers to high cytokine producers (Supplementary Fig. 2B). To verify that the results are not affected by mixing individuals from various ancestral backgrounds, we analyzed our MDDC data from Fig. 3A for the subset of European ancestry-only individuals that represent the majority within our cohort (Supplementary Fig. 3A). Moreover, we observe similar trends of IRF5 rs2004640 genotypes on cytokine production in the limited numbers of the non-European ancestry cohort (data not shown).

Figure 3. The disease-associated rs2004640 and CGGGG-indel IRF5 risk alleles result in increased TNF-α secretion upon Nod2 or TLR2 stimulation of primary human MDDC.

The cytokine responses from Fig 2 were stratified based on the genotype of the individuals for rs2004640 (A) or CGGGG-indel (B) in IRF5. The data are represented as fold TNF-α induction (log₂ transformed) upon dose-dependent Nod2 (left) or TLR2 (right) stimulation stratified on the IRF5 genotype + SEM. *, p<0.05; **, p<0.01; ***; p<0.001; †, p<1×10⁻⁴; ††, p<1×10⁻⁵.

Figure 4. The rs2004640 IRF5 T risk allele is associated with increased TNF-α induction upon Nod2 or TLR2 stimulation of primary human MDM.

The cytokine responses from MDM derived from the same individuals as in Figure 2 were stratified on rs2004640 IRF5 genotype. The data are represented as fold TNF-α induction (log₂ transformed) upon dose-dependent Nod2 (left) or TLR2 (right) stimulation stratified on rs2004640 IRF5 genotype + SEM. **, p<0.01; ***; p<0.001; †, p<1×10⁻⁴; ††, p<1×10⁻⁵.

Given this important functional outcome of rs2004640, we examined additional IRF5 disease-associated polymorphisms that affect IRF5 through different putative mechanisms. The CGGGG-indel IRF5 variant is associated with various autoimmune/inflammatory diseases, including SLE (odds ratio 1.42–2.02)(10), and Crohn’s disease(6). In contrast to rs2004640, CGGGG-indel IRF5 provides an additional SP1 transcriptional factor binding site(6). Immune responses from cells carrying this allele have not been examined. We find that MDDC from CGGGG-insertion carriers show dramatically increased expression of TNF-α (Fig. 3B), IL-12p40, IL-8, IL-1β and IL-10 following stimulation of Nod2 or TLR2 with gene-dose and PRR ligand-dose effects. Taken together, we demonstrate that specific disease-associated IRF5 polymorphisms increase cytokine induction upon PRR stimulation in a ligand-dose and gene-dose dependent manner and contribute to the heterogeneity in PRR-mediated cytokine induction across healthy individuals.

IRF5 knockdown significantly reduces PRR-mediated cytokine induction from carriers of high risk IRF5 variants

IRF5 deficiency results in decreased cytokines upon stimulation with N-glycolyl-MDP(30) and TLR ligands(31,32). To examine if the production of cytokines in human MDDC is directly regulated by IRF5, we silenced IRF5 (Fig. 5A) in individuals that were homozygous for the rs2004640 T allele and CGGGG-insertion allele (high cytokine producers) using siRNA. We observe that decreasing IRF5 expression in cells from IRF5 risk allele carriers results in significantly attenuated RNA (Fig. 5B) and protein (Fig. 5C) expression of TNF-α, IL-12p40, IL-8, IL-1β and IL-10 upon Nod2 or TLR2 stimulation, indicating that IRF5 is crucial for modulating cytokine responses through these receptors.

Figure 5. IRF5 knockdown reduces cytokine mRNA expression and protein secretion following Nod2 or TLR2 stimulation.

MDDC from individuals homozygous for both IRF5 rs2004640 T allele and CGGGG-insertion allele (n=8) (high cytokine-producer cohort) were transfected with scrambled or IRF5 siRNA. (A) IRF5 mRNA expression was assessed 48 h after transfection to ensure efficacy of knockdown. The data are represented as the fold mRNA expression compared to cells transfected with scrambled siRNA (represented by the dotted line at 1) + SEM. 48 h after transfection cells were treated with 1 µg/ml MDP or 1 µg/ml Pam₃Cys for (B) 4 h or (C) 24h. The data are represented as the fold (B) mRNA expression or (C) protein expression (log₂ transformed) of TNF-α, IL-12p40, IL-8, IL-1β and IL-10 compared to untreated cells transfected with scrambled siRNA + SEM. Δ, change; M, MDP; P, Pam₃Cys; scr, scrambled siRNA **, p<0.01; ***, p<0.001, †, p<1×10⁻⁴.

IRF5 knockdown significantly reduces MDP-mediated MAPK and NF-κB activation in MDDC

We next questioned how IRF5 regulates pathways leading to cytokine induction. MAPK and NF-κB participate in cytokine induction downstream of PRR(33), but interestingly, the activation of these pathways was not affected in B cells from IRF5-deficient mice upon PRR stimulation(31). However, to our knowledge, how IRF5 regulates these and other pathways in dendritic cells, including primary human MDDC, has not been studied. We therefore utilized siRNA to IRF5 to investigate the consequences of decreased IRF5 expression on Nod2-initiated signaling pathways in primary human MDDC. We find that in contrast to the mouse B cell studies, IRF5 silencing upon Nod2 stimulation of MDDC resulted in significant attenuation of ERK, p38 and JNK activation (Fig. 6A). IRF5 knockdown cells also showed significantly reduced NF-κB pathway activation upon Nod2 stimulation as evidenced by decreased IκBα phosphorylation (Fig. 6B). We and others have found that PI3K and one of its downstream targets, mTOR, also regulate Nod2-mediated cytokine induction(25,34), such that we examined the effect of IRF5 silencing on these pathways. We find that neither the activation of the PI3K substrate Akt, nor of the mTOR substrate p70-S6K, was significantly attenuated upon IRF5 knockdown in MDDC (Fig. 6C). Taken together, we find that IRF5 is necessary for optimal Nod2-mediated activation of both the MAPK and NF-κB pathways and that IRF5 knockdown in high risk IRF5 variant-carrier cells decreases the elevated PRR-mediated cytokine induction observed in these cells.

Figure 6. IRF5 knockdown reduces MAPK and NF-κB pathway activation following Nod2 stimulation.

(A–C) MDDC were transfected with scrambled or IRF5 siRNA and stimulated with 100 µg/ml MDP for 10 min. Left: Shown are representative flow cytometry plots with MFI values as indicated for (A) phospho-MAPK (n=6–8), (B) phospho-IκBα (n=6) or (C) phospho-Akt and phospho-p70S6K (n=11–14). Shaded region indicates untreated cells, solid dark line indicates scrambled siRNA-transfected cells treated with MDP and dotted line indicates IRF5 siRNA-transfected cells treated with MDP. Right: Summarized data are represented as the fold phospho-protein induction normalized to untreated cells (represented by the dotted line at 1) + SEM. scr, scrambled siRNA. *, p<0.05; **, p<0.01; †, p<1×10⁻⁴.

Disease-risk IRF5 polymorphisms show decreased cytokine production upon individual or synergistic stimulation of a broad range of PRR

To determine whether disease-associated IRF5 polymorphisms affect cytokine secretion by a broad range of PRRs, we isolated MDDCs from a cohort of 77 individuals and examined cytokine secretion after stimulation with various PRR ligands. We selected PRR ligand doses that are commonly used and that result in a similar range of cytokine induction. We observed a high degree of intra-individual correlation across Nod2, TLR2, TLR3, TLR4, TLR5, TLR7 and TLR9 ligand stimulation (Fig. 7A). Therefore, if MDDC from a given individual produce high levels of cytokines in response to stimulation of one PRR, they similarly do so in response to stimulation of other PRRs. As low doses of Nod2 and TLR ligands synergize to induce more robust cytokine induction(20,27), the synergy of Nod2 with each of the above PRRs was examined. We observed high intra-individual correlation in levels of cytokine secretion upon stimulation of MDDC with MDP combined with each of the individual TLR ligands examined (Fig. 7B). Finally, stratification on rs2004640 (Fig. 7C) and CGGGG indel (Fig. 7D) IRF5 polymorphisms revealed that the risk allele carriers secrete significantly higher levels of TNF-α upon TLR2, TLR4, TLR5, TLR7 and TLR9 alone (Fig. 7C&D, left) or in combination (Fig. 7C&D, right) with Nod2, indicating that these IRF5 polymorphisms affect TNF-α secretion by a broad range of PRRs. Similar results were observed for IL-12p40, IL-8, IL-1β and IL-10. Moreover, similar results were observed when analyzing on European ancestry only individuals (Supplementary Fig. 3B). To corroborate our results, we stratified our data on two additional IRF5 disease-associated polymorphisms, rs7808907 and rs4728142, and find similar results with the disease-risk alleles for these IRF5 polymorphisms (Supplementary Fig. 4A–D).

Figure 7. Cytokine secretion upon PRR stimulation and Nod2/TLR synergistic stimulation of primary human MDDC shows excellent inter-individual correlation and the disease-associated rs2004640 and CGGGG-indel IRF5 risk alleles result in increased TNF-α secretion upon stimulation of multiple PRR.

Human MDDC (n=77) were stimulated with 1 µg/ml MDP 1 µg/ml Pam₃Cys, 0.1 µg/ml polyI:C, 0.01 µg/ml lipid A, 0.5 ng/ml flagellin, 0.1 µg/ml CL097 or 0.1 µg/ml CpG DNA for 24 h alone (A) or in combination (B). Supernatants were assayed for TNF-α. (A, B) Correlation in fold TNF-α induction (log₂ transformed) between PRR stimulation alone or in combination as calculated by Spearman correlation coefficients. (C, D) The cytokine responses were stratified based on the genotype of the individuals for rs2004640 (C) or CGGGG-indel (D) in IRF5. The data are represented as fold TNF-α induction (log₂ transformed) upon individual (left) and combined (right) Nod2 and TLR stimulation stratified on the IRF5 genotype + SEM. *, p<0.05; **, p<0.01; ***; p<0.001; †, p<1×10⁻⁴; ††, p<1×10⁻⁵. r, Spearman rho coefficient.

The IRF5 rs2004640 polymorphism is a major contributor to inter-individual variance of cytokine expression

The four IRF5 markers examined in our cohort are contained in the same haplotype block, and are in strong linkage disequilibrium(Supplementary Fig. 4E), similar to observations in other studies(9,10) and the International HapMAP CEU data. To more thoroughly analyze the effects of IRF5 polymorphisms, we increased the density of mapping in the IRF5 region to include a total of twenty markers that have been assessed in prior studies (6–11,17,35,36)(Supplementary Fig. 4E). To assess the relative importance of the twenty IRF5 variants in the observed inter-individual cytokine variation, we analyzed each of the variants individually by regression with respect to their effect on TNF-α secretion upon 1 µg/ml MDP stimulation of MDDC from European ancestry individuals. We found that the rs2004640 SNP had the most significant effect on the magnitude and inter-individual variance of TNF-α secretion (Table II). We next questioned whether the twenty IRF5 polymorphisms, or a subset of these polymorphisms, in combination lead to an increased contribution to cytokine variance relative to the contribution of the single polymorphisms. Using a forward stepwise regression model on all twenty polymorphisms, we found that rs2004640 fully accounts for the IRF5 contributions to TNF-α secretion upon 1 µg/ml MDP stimulation of MDDC. Taken together, IRF5 polymorphisms are a major determinant of the inter-individual variance in cytokine expression.

Table 2.

Analysis of effects of multiple polymorphisms in the IRF5 region on MDP-induced TNF-α secretion

IRF5 snp	Expected minor allele frequencyA	Observed minor allele frequency	Major allele homoz. (n)	Het. (n)	Minor allele homoz. (n)	p value	r2
rs960633	0.292–0.327	0.265	60	43	8	2.10×10−3	0.08
rs729302	0.310–0.325	0.351	49	46	16	1.81×10−6	0.23
rs11768806	0.183	0.171	79	26	6	4.94×10−5	0.15
rs4728142	0.398–0.417	0.428	42	43	26	8.33×10−10	0.36
CGGGG indel	0.430–0.470	0.432	41	44	26	3.98×10−11	0.42
rs2004640	0.440–0.560	0.468	36	46	29	3.71×10−13	0.53
rs3807307	0.475	0.446	39	45	27	4.55×10−12	0.42
rs3807306	0.492–0.496	0.482	35	45	31	2.11×10−11	0.46
rs7808907	0.458–0.467	0.473	36	45	30	3.80×10−10	0.35
rs1874328	0.300–0.341	0.360	48	46	17	2.30×10−5	0.13
rs2070197	0.150	0.135	85	22	4	1.95×10−4	0.15
rs10954213	0.467	0.401	46	41	24	2.77×10−8	0.27
rs11770589	0.375	0.473	36	45	30	4.50×10−3	0.05
rs10954214	0.400	0.360	51	40	20	6.55×10−10	0.35
rs7800687	0.400	0.360	51	40	20	6.55×10−10	0.35
rs10488630	0.283–0.314	0.342	50	46	15	1.23×10−5	0.14
rs10488631	0.124–0.158	0.126	86	22	3	2.92×10−5	0.20
rs2280714	0.385–0.400	0.360	51	40	20	6.55×10−10	0.35
rs12539741	0.150	0.126	86	22	3	2.92×10−5	0.20
rs17166351	0.383	0.473	36	45	30	4.50×10−3	0.05

^AAs reported in dbSNP version 135 (Oct 12, 2011) and References 8–10

Regression analysis was conducted on twenty markers in the IRF5 region individually in European ancestry individuals (n=111) to examine effects on TNF-α secretion upon 1 µg/ml MDP stimulation of MDDC. chr., chromosome; homoz., homozygotes; het., heterozygotes.

Discussion

The dramatic effects of IRF5 genotypes on PRR responses observed here reflect several unique features of IRF5. First, although numerous genetic loci have demonstrated associations across multiple autoimmune disorders, IRF5, together with the PTPN22 and IL23R polymorphisms, confer uniquely high disease odds ratios(6–11). Secondly, IRF5 polymorphisms are unique in their effects on gene expression; in a genome-wide expression quantitative trait loci screen, cis-acting IRF5 polymorphisms demonstrated one of the most significant correlations to mRNA expression in the entire transcriptome(37). Finally, the frequencies of both the risk T and the non-risk G allele of IRF5 rs2004640 range from 0.44 – 0.56(8–10); therefore, the distinct TT, TG and GG rs2004640 genotypes are commonly distributed throughout population. This common distribution, combined with the dramatic effect of rs2004640 on PRR-induced cytokines in MDDC in a gene-dose dependent manner, accounts for the marked contribution of IRF5 rs2004640 to inter-individual variance in cytokine secretion. Taken together, utilizing monocyte-derived cells from a large cohort of individuals, we demonstrate that rs2004640, a common disease-associated polymorphism in IRF5, confers increased responsiveness to a broad array of PRRs and is a major genetic contributor to the variation in PRR-induced cytokine secretion observed between individuals.

Although the IRFs have generally been associated with viral responses and secretion of Type I IFN, IRF5 contributes to induction of additional cytokines by other microbial ligands(31,32), implicating IRF5 in responses to multiple classes of pathogens in human and mouse cells. PRR stimulation results in IRF5 phosphorylation and translocation into the nucleus(38). In the nucleus, IRF5 binds to promoters of pro-inflammatory cytokine genes, thereby inducing their expression(31,32,38). Consistent with its contributions to cytokine secretion from innate cells, IRF5 influences subsequent Th1 and Th17 responses(32) and assists in responses to and/or clearance of various pathogens(30,39,40). At the same time, IRF5 deficiency leads to decreased cytokines and amelioration of disease progression in a mouse model of SLE(41).

A recent study identifies an IRF5 haplotype in SLE patients that is associated with higher anti-dsDNA and anti-Ro antibodies and increased serum IFN-α(35). This haplotype, consisting of rs2004640, rs3807306, rs10488631 and rs2280714, accounts for four distinct mechanisms of regulating IRF5(9,35). While stratifying our results on this haplotype showed a highly significant contribution of the previously implicated(35) TATA and TACA (rs2004640, rs3807306, rs10488631 and rs2280714) haplotypes (p-value=5.64×10⁻¹²) to MDP-induced TNF-α secretion, these results did not improve on the rs2004640 genotype alone (Table II).

In our study we also show that lowering IRF5 expression in human MDDCs from IRF5-risk allele carriers (high cytokine producers) dramatically decreases secretion of multiple cytokines following Nod2 and TLR2 stimulation (Fig. 5C). Moreover, we determine that decreased IRF5 expression in MDDC results in decreased Nod2-mediated MAPK and NF-κB pathway activation, whereas PI3K and mTOR pathway activation was not significantly impaired (Fig. 6). In contrast to our findings, a previous study in B cells from IRF5-deficient mice examining MAPK activation did not demonstrate a defect in this pathway upon PRR stimulation relative to WT B cells(31). This discrepancy highlights fundamental functional differences between mouse B cells and our studies in primary human MDDC, as well as consequences to new pathways examined.

Polymorphisms in other genes have been associated with modifying cytokine secretion upon stimulation of select PRR(1,42). However, we find that the commonly distributed rs2004640 IRF5 risk-allele variant has the most profound influence on the variance in inter-individual PRR-induced cytokine secretion among not only the polymorphisms we examined (Table I), but, to our knowledge, among the other common disease risk polymorphisms reported to date. The fine-tuning of cytokine secretion following infection likely significantly influences susceptibility to both infection and autoimmune/inflammatory disease. The highly correlated hyper-responsiveness of PRR-induced cytokines and IRF5 disease-associated polymorphisms implicates a wide spectrum of microbial triggers in autoimmune disease pathogenesis and has broad implications in global immunological responses, host defenses against pathogens and disease susceptibility.

Abbreviations

Nod

nucleotide-binding oligomerization domain

PRR

pattern-recognition receptor

IBD

inflammatory bowel disease

T1DM

type I diabetes mellitus

SLE

systemic lupus erythematosus

MS

multiple sclerosis

MDP

muramyl dipeptide

MDDC

monocyte-derived dendritic cells

MDM

monocyte-derived macrophages