Structural insertiondeletion variation in IRF5 is associated with a risk haplotype and defines the precise IRF5 isoforms expressed in systemic lupus erythematosus.код для вставкиСкачать
ARTHRITIS & RHEUMATISM Vol. 56, No. 4, April 2007, pp 1234–1241 DOI 10.1002/art.22497 © 2007, American College of Rheumatology Structural Insertion/Deletion Variation in IRF5 Is Associated With a Risk Haplotype and Defines the Precise IRF5 Isoforms Expressed in Systemic Lupus Erythematosus Sergey V. Kozyrev,1 Susanna Lewén,1 Prasad M. V. Linga Reddy,1 Bernardo Pons-Estel2 and the Argentine Collaborative Group, Torsten Witte3 and the German Collaborative Group, Peter Junker,4 Helle Laustrup,4 Carmen Gutiérrez,5 Ana Suárez,5 Maria Francisca González-Escribano,6 Javier Mart´ın7 and the Spanish Collaborative Group, and Marta E. Alarcón-Riquelme1 Objective. To determine whether specific isoforms of IRF5 are transcribed in patients with systemic lupus erythematosus (SLE) who have risk genotypes in the exon 1B donor splice site at single-nucleotide polymorphism (SNP) no. rs2004640. Methods. Peripheral blood mononuclear cells were obtained from SLE patients and healthy controls from Argentina, Spain, and Germany and from trio families from Spain and Denmark. A reporter assay was used to investigate the role of SNP no. rs2004640. IRF5 expression in relation to the genotypes of functional SNPs was analyzed using quantitative polymerase chain reaction. Sequencing and genotyping of the IRF5 gene was performed. Results. Sequencing of complementary DNA from individuals with different genotypes showed 4 basic isoforms transcribed from all 5ⴕ-untranslated regions (5ⴕ-UTRs), suggesting no preferential isoform transcription based on rs2004640 genotypes. Analysis of translation efficiency showed that exon 1A was the most efficient in initiating protein synthesis. We identified a novel polymorphic insertion/deletion that defines the pattern of expression of isoforms of IRF5. The insertion consists of 4 repeats in exon 6 affecting the protein interaction domain. The insertion segregates in the risk haplotype with the high expression allele of a poly(A) site SNP no. rs10954213 and the exon 1B donor splice allele of the 5ⴕ-UTR SNP no. rs2004640. The poly(A) polymorphism correlated with levels of IRF5 in cells stimulated with interferon-␣. The SNP most strongly associated with SLE was SNP no. rs2070197 (P ⴝ 5.2 ⴛ 10ⴚ11), which is a proxy of the risk haplotype, but does not appear to be functional. Conclusion. None of the functional variants investigated in this study is strongly associated with SLE, with the exception of the exon 1B donor splice site, and its functional importance appears to be small. Our results suggest that there may be other functional polymorphisms, yet to be identified, in IRF5. We did not observe evidence of epistatic interaction between the functional SNPs. Supported by the Swedish Research Council, the Swedish Association Against Rheumatism, the 80-Year Foundation of King Gustaf V, the Marcus Borgströms Foundation, the Magnus Bergvalls Foundation, the Torsten and Ragnar Söderbergs Foundation, and the Danish Rheumatism Association. Dr. Kozyrev’s work was supported by the Gurli and Edward Brunnbergs Foundation for Rheumatic Research. The German Collaborative Group’s work was supported by a grant from BMBF Kompetenznetz Rheuma C2.12. Dr. AlarcónRiquelme is supported by the Knut and Alice Wallenberg Foundation through a Royal Swedish Academy of Sciences award. 1 Sergey V. Kozyrev, PhD, Susanna Lewén, MSc, Prasad M. V. Linga Reddy, MSc, Marta E. Alarcón-Riquelme, MD, PhD: Rudbeck Laboratory, Uppsala University, Uppsala, Sweden; 2Bernardo PonsEstel, MD: Sanatorio Parque, Rosario, Argentina; 3Torsten Witte, MD, PhD: Medical School Hannover, Hannover, Germany; 4Peter Junker, MD, Helle Laustrup, MD: Odense University Hospital, Odense, Denmark; 5Carmen Gutiérrez, MD, PhD, Ana Suárez, PhD: Hospital Universitario Central de Asturias, Universidad de Oviedo, Oviedo, Spain; 6Maria Francisca González-Escribano, PhD: Hospital Virgen del Rocı́o, Sevilla, Spain; 7Javier Martı́n, MD, PhD: Instituto de Biomedicina López Neyra, CSIC, Granada, Spain. Address correspondence and reprint requests to Marta E. Alarcón-Riquelme, MD, PhD, Department of Genetics and Pathology, Rudbeck Laboratory, Uppsala University, Dag Hammarskjolds vag 20, 751 85 Uppsala, Sweden. E-mail: firstname.lastname@example.org. Submitted for publication December 11, 2006; accepted in revised form December 21, 2006. 1234 SLE AND STRUCTURAL INSERTION/DELETION VARIATION IN IRF5 IRF5, a key component of the type I interferon pathway, has a strong genetic association with systemic lupus erythematosus (SLE). IRF5 is a transcription factor involved in the transcriptional activation of various proinflammatory cytokines and interferon-␣ (IFN␣) (1) and is primarily involved in host defense against viruses. The various isoforms of IRF5 have different effects on the interferon system (2). The genetic association of IRF5 with SLE was recently identified (3), and we have previously described a mechanism through which IRF5 contributes to genetic susceptibility to SLE (4). The IRF5 gene has 4 alternative exons in the 5⬘-untranslated region (5⬘-UTR), named exon 1A, exon 1B, exon 1C, and exon 1D. The T allele of the single-nucleotide polymorphism (SNP) no. rs2004640 introduces a donor splice site leading to the expression of exon 1B transcripts and the reduction of exon 1C–derived transcripts (4). A second SNP located 5 kb downstream of IRF5, rs2280714, has a T allele that is strongly associated with high levels of IRF5 (4). Alleles at both SNP positions comprise a haplotype associated with SLE (4). Our primary hypothesis was that the 5⬘-UTRs affected which of the alternative isoforms would be expressed, and that the splice donor mutation represented by SNP no. rs2004640 changed their pattern of expression. Identification of the major isoforms of IRF5 in patients with SLE is of major importance in understanding the mechanisms through which IRF5 leads to the development of SLE. In the present study, we found that a previously unknown structural insertion/deletion in exon 6 of the IRF5 gene leads to the precise expression of specific isoforms in the risk haplotype associated with SLE. PATIENTS AND METHODS Patients and controls. Peripheral blood mononuclear cells (PBMCs) were obtained from SLE patients and healthy controls from Argentina, Spain, and Germany. The groups of subjects from Argentina and Spain have been previously described (4). The group of subjects from Germany consisted of 288 SLE patients and 245 healthy controls from various centers in northern Germany. Samples from the German subjects were collected at the University of Hannover. The study was conducted with the participation of the members of the Argentine, German, and Spanish Collaborative Groups (Appendix A). Trio families were from 2 separate populations. Complete trios consisted of both parents and an affected child, and incomplete trios consisted of one parent and an affected child. One set of 89 families was from Asturias in northern Spain and the Madrid region (49 complete trios), and the second group of 1235 79 families was from Denmark (22 complete trios). All individuals fulfilled the American College of Rheumatology 1982 revised criteria for SLE (5). Control samples for expression analysis were obtained from healthy donors at the Uppsala Academic Hospital Blood Bank, as previously described (4). Stimulation of PBMCs. Freshly isolated PBMCs were stimulated with 1,000 units of IFN␣ (RayBiotech, Norcross, GA) for 6 hours in RPMI 1640 medium supplemented with penicillin/streptomycin and 10% fetal calf serum. Complementary DNA (cDNA) synthesis and quantitative polymerase chain reaction (PCR) of IRF5 UTR–specific transcripts. Total RNA from individuals with SLE carrying the various genotypes was purified from PBMCs with TRIzol (Invitrogen, San Diego, CA). We reverse-transcribed 2 g of total RNA with 2 units of MultiScribe reverse transcriptase in PCR buffer II containing 5 mM MgCl2, 1 mM dNTPs, 0.4 units of RNase inhibitor, and 5 M oligo(dT). All reagents were from Applied Biosystems (Foster City, CA). Synthesis was performed at 42°C for 45 minutes, followed by 95°C for 5 minutes. IRF5 isoforms with distinct 5⬘-UTRs were quantified using TaqMan real-time PCR on an ABI Prism 7700 Sequence Detector (Applied Biosystems) and SDS version 1.9.1 software. Primers used to distinguish PCR products with different UTRs have been published previously (4). The exon 1D– specific primer was 5⬘-GCTCAGCCCGGATCTGCAGTTGCCAG-3⬘. We used a common reverse primer lying in exon 3 and a common TaqMan probe labeled with FAM and TAMRA. We performed 45 cycles of 2-step PCR (95°C for 15 seconds and 63°C for 1 minute) in buffer containing 1.5 mM MgCl2, 200 M each dNTP, 0.5 units of Platinum Taq polymerase (Invitrogen), primer-probe mixture, and cDNA. Expression levels were normalized to levels of the human TBP gene (Applied Biosystems). Cloning and sequence analysis of IRF5 isoforms. PCR amplification of diverse isoforms of IRF5 was performed with the same forward primers used for the TaqMan assay. The common reverse primer 5⬘-GCAGCCTTGTTATTGCATGCCAGCTG-3⬘ was designed to allow amplification of the full-length isoforms. Cycle conditions were 95°C for 3 minutes, followed by 40 cycles of 95°C for 15 seconds, 60°C for 15 seconds, and 72°C for 1.5 minutes. PCRs were performed in a 25-l reaction volume with 0.5 units of High Fidelity Platinum Taq polymerase in the buffer supplied by the manufacturer (Invitrogen). Electrophoresis was performed on PCR products on a 1% agarose gel. PCR products were subcloned in pCR4-TOPO vector (Invitrogen), and 50 random positive clones were analyzed by sequencing. PCR was also used for direct analysis of insertion/deletion patterns in cDNA prepared from PBMCs. The following primers were used: forward 5⬘-TGCAGGAGAGGAGGAGGAAGAAGAG-3⬘, reverse 5⬘-AACTTGATCTCCAGGTCGGTCA-3⬘. In vitro translation efficiency. To determine the effect of alternative 5⬘-UTRs on protein synthesis, we amplified full-length UTRs using a common reverse primer, as for the quantitative PCR, and primers Ap1 and Ap2 and human spleen BD Marathon-Ready cDNA (Clontech, Palo Alto, CA). PCR products were gel purified with a kit from Qiagen (Chatsworth, CA) and subcloned in pCR4-TOPO vector for sequence analysis. The longest fragments were subcloned in pGL3 promoter vector. Plasmid DNA was purified with the 1236 KOZYREV ET AL EndoFree plasmid Maxi kit (Qiagen) and transfected with Lipofectamine 2000 into HEK 293 cells. To allow transfection normalization, pRL-TK vector (Promega, Madison, WI) was cotransfected with the reporter constructs. After 48 hours of incubation, cells were harvested and lysed in a passive lysis buffer, according to the recommendations of the manufacturer (Promega). Firefly and Renilla luciferase activity levels were measured in duplicate with the Dual Luciferase Reporter Assay (Promega). Experiments were repeated 4 times with 2 independent DNA preparations. Transterm (http://uther.otago.ac.nz/Transterm.html) was used to analyze the 3⬘-UTR of IRF5 (6). Genotyping. For genotyping, we used the TaqMan SNP Genotyping Assay (Applied Biosystems), and detection was performed using an ABI 9700 real-time PCR system. The primers and probes were designed by the Applied Biosystems Assays-on-Demand service for allele discrimination with the 5⬘-nuclease assay and fluorogenic probes. The average genotype completeness for SNP no. rs10954213 and no. rs2070197 was 98.8% for the samples from Argentine subjects, 99.6% for the samples from German subjects, and 86% for the samples from Spanish subjects. Genotyping for the insertion/deletion was performed by PCR analysis under the conditions specified above, with the following primers: forward 5⬘GCCGTCCACACGCACTCTCTGTAG-3⬘, reverse 5⬘-CTGAAGCCAGCAGGGTTGCCAG-3⬘. PCR products were resolved on 2.5% agarose gels. Accession numbers. Sequences for the full-length 5⬘UTRs of IRF5 have been deposited in GenBank under accession numbers DQ995491, DQ995492, DQ995493, and DQ995494. Accession numbers for the sequences of IRF5 with insertion and deletion are DQ995495 and DQ995496, respectively. Statistical analysis. Analyses of the genetic associations of the SNPs and haplotypes in patients and controls and the family-based analysis were performed using Haploview, version 3.2, which incorporates only complete trios. We used Family Based Association Testing software (7), which estimates genotypes for the nongenotyped parents so that information from the incomplete trios can be used. The software was used to analyze all Spanish and Danish trios jointly (n ⫽ 168) and corroborate the association results, taking into consideration all genetic information. Results were comparable with those obtained using Haploview. Five trios showed Mendelian inconsistencies and were excluded from the analysis. Odds ratios were calculated as previously described (4). Correlations of IRF5 messenger RNA (mRNA) levels were analyzed using the t-test included in GraphPad software (GraphPad Software, San Diego, CA), as previously described (4). R language was used for logistic regression analysis using the case control material. RESULTS Low contribution of exon 1B transcripts to the overall levels of IRF5. We performed quantitative analysis of IRF5 gene expression in human PBMCs upon stimulation with IFN␣. These studies clearly showed that exon 1A transcripts were expressed at higher levels than other exon 1 transcripts in both stimulated and unstimulated cells, while transcripts derived from exon 1B or exon 1C were produced at levels 100 times lower (Figure 1a). Exon 1D transcripts initially cloned from human spleen cDNA were also found at very low levels. We reasoned that despite the low levels of mRNA for some isoforms, the corresponding protein levels might be different and dependent on 5⬘-UTR– mediated translation efficiency. Therefore, we cloned cDNA for the 4 full-length 5⬘-UTRs in the luciferase reporter vector and analyzed their effect on in vitro protein expression. The most potent was the 5⬘-UTR encoded by exon 1A. Exon 1B and exon 1D did not show any effect compared with the control (pGL3 promoter vector), while the presence of exon 1C significantly inhibited the efficiency of protein translation (Figure 1b). Thus, mRNA for exons 1B, 1C, and 1D all have poor translation efficiency as compared with that for exon 1A. We then selected 18 individuals carrying the various genotypes of rs2004640. Using primers specific for each 5⬘-UTR, we performed PCR amplification and cloned cDNA from those individuals and sequenced 50 clones for each. We observed that all 4 promoters and 5⬘-UTRs led to expression of the same set of isoforms (V1, V4, V5, and V6 transcripts; results not shown). Hence, the effect of exons 1B, 1C, and 1D on the production of IRF5 protein isoforms, as compared with exon 1A, is negligible, since they code for the same isoform set as does exon 1A, which is expressed at far higher levels. Determination of isoform pattern by structural insertion/deletion and alternative 3ⴕ acceptor splice sites. Interestingly and unexpectedly, based on the sequences found in different samples, we could classify the individuals into 3 separate groups: those who had V1 and V4 isoforms, those who had V5 and V6 isoforms, and those who expressed all 4 isoforms (Figure 1c), suggesting Mendelian segregation and the effect of new structural genetic variation. We sequenced the complete IRF5 gene in 6 individuals from each of the 3 groups. We identified a novel insertion/deletion, which lacked 2 of 4 tandem repeats in exon 6 encoding for a proline-rich region within the putative PEST and interaction domains. The presence or absence of the repeats determined the isoforms to be expressed, such that individuals with 4 repeats (insertion) expressed isoforms V5 and V6, while individuals with 2 repeats (deletion) expressed isoforms V1 and V4 (Figure 1c). An alternative 3⬘ acceptor splice site in exon 6 defined the expression of V1 or V4 and V5 or V6 isoforms, respectively SLE AND STRUCTURAL INSERTION/DELETION VARIATION IN IRF5 1237 Figure 1. a, Expression levels of exon 1A, 1B, and 1C transcripts in unstimulated human peripheral blood mononuclear cells (PBMCs) and PBMCs stimulated with interferon-␣ (IFN␣). The data are typical of findings in individuals who have the T allele of rs2004640 and the A allele of rs10954213. Expression levels were normalized to levels of TBP. The expression level of the exon 1D transcript was detected with nested polymerase chain reaction (PCR) and is not shown on the plot. Values are the mean relative x-fold expression. b, Relative translation efficiency of the 4 alternative 5⬘-untranslated regions (5⬘-UTRs). HEK 293 cells were transfected with either pGL3 promoter vector (pGL3-prom) or constructs with alternative 5⬘-UTRs preceding luciferase gene. Firefly luciferase activity was measured with the Dual Luciferase Reporter Assay, and values were normalized to Renilla luciferase activity. Values are the mean ⫾ SD of the relative luciferase activity in 4 experiments. c, IRF5 gene structure. Solid boxes are protein coding exons. Open boxes are noncoding exons. Top, Location of the single-nucleotide polymorphisms and insertion/deletion (in-del) variation. Bottom, Two alternative 3⬘ acceptor splice sites in exon 6, giving rise either to long or short isoforms with 2 repeats, V1 or V4, or to long or short isoforms with 4 repeats, V5 or V6. The PCR gel shows the cDNA expression pattern obtained with PCR amplification of the segment from exon 4 to exon 7, in 6 individuals with insertion/deletion genotypes having 2 repeats (2R), 4 repeats (4R), or both. Each lane represents 1 individual. (Figure 1c). Further, all isoforms within each group (i.e., V1 and V4 in 1 group, V5 and V6 in 1 group, and V1, V4, V5, and V6 in the heterozygous group) were equally expressed, according to our comprehensive sequencing analysis of transcripts in various individuals (results not shown). Role of the length of the 3ⴕ-UTR and SNP no. rs10954213 in determining the level of IRF5. Through bioinformatic analysis of the IRF5 gene, we found that SNP no. rs10954213 is located in the 3⬘-UTR polyadenylation site AAT(A/G)AA. Thus, SNP no. rs10954213 potentially determines IRF5 expression levels, with the A allele predicting mRNA with a short 3⬘-UTR, and the G allele disrupting the poly(A) site and thus causing transcription to continue. Computational prediction with Transterm software revealed 2 strong AU-rich elements (AREs) located within the long 3⬘UTR (6). We tested whether any of the SNPs of IRF5 correlated with levels of IRF5 mRNA, and observed that, as expected, rs10954213, the 3⬘-UTR SNP, correlated with levels of IRF5 mRNA, in particular when cells were stimulated with IFN␣ (Figure 2). The highest correlations were observed with the A allele, but these did not reach statistical significance due to the number of samples used. No correlation was observed for rs2004640, insertion/deletion, or the risk haplotype (data not shown). 1238 KOZYREV ET AL Figure 2. Effect of rs10954213 on expression of the IRF5 gene in unstimulated PBMCs (non-stim) and PBMCs stimulated with 1,000 units IFN␣. Diamonds represent individual samples, and bars represent the mean level of expression. The y-axis shows the expression levels normalized to TBP. See Figure 1 for other definitions. Insertion in the risk haplotype. The insertion was found in the risk haplotype defined by the nonfunctional SNP no. rs2070197, which is the SNP in IRF5 with the highest level of association with SLE. SNP no. rs2070197 Table 1. divided the risk haplotype identified previously (4) into smaller haplotypes, with one (TCA) clearly segregating in patients and one (GTG) clearly protective (Table 1). The 2 novel SNPs, SNP no. rs10954213 and no. Genetic association of the risk haplotype of IRF5* Haplotype Spanish samples TTA GTG† GTA TCA‡ Argentine samples TTA GTG† GTA TCA‡ German samples TTA GTG† GTA TCA‡ Combined samples TTA GTG† GTA TCA‡ TICA‡ Case Frequency Control Frequency 2 P OR (95% CI) 0.432 0.217 0.142 0.146 0.402 0.288 0.152 0.106 1.55 16.344 0.522 9.163 0.28 5.2 ⫻ 10⫺5 0.4698 0.002 – 0.68 (0.57–0.82) – 1.44 (1.13–1.84) 0.309 0.363 0.094 0.178 0.303 0.431 0.129 0.091 0.042 5.616 3.636 19.026 0.836 0.0178 0.0565 1.28 ⫻ 10⫺5 – 0.75 (0.59–0.95) – 2.18 (1.53–3.12) 0.374 0.274 0.135 0.165 0.343 0.331 0.168 0.091 1.073 3.979 2.244 12.466 0.300 0.046 0.134 0.0004 – 0.76 (0.58–0.99) – 1.99 (1.35–2.93) 0.381 0.271 0.125 0.159 0.151 0.363 0.334 0.147 0.102 0.099 1.599 22.279 5.058 34.189 24.456 0.206 2.35 ⫻ 10⫺6 0.02 5.0 ⫻ 10⫺9 5.72 ⫻ 10⫺8 – 0.73 (0.65–0.83) – 1.67 (1.40–1.99) – * The order of the single-nucleotide polymorphisms is rs2004640, rs2070197, and rs10954213. Odds ratios (ORs) and 95% confidence intervals (95% CIs) are shown for the protective and risk haplotypes only. † Protective haplotype. ‡ Risk haplotype. I ⫽ insertion. SLE AND STRUCTURAL INSERTION/DELETION VARIATION IN IRF5 Table 2. 1239 Genetic association of the individual SNP risk alleles of IRF5 in combined samples* SNP Associated allele Case frequency Control frequency 2 P rs2004640 rs2070197 Insertion/deletion rs10954213 T C I A 0.602 0.166 0.507 0.660 0.516 0.096 0.515 0.606 34.857 43.083 0.23 11.263 3.5 ⫻ 10⫺9 5.2 ⫻ 10⫺11 0.626 0.0008 * SNP ⫽ single-nucleotide polymorphism. rs2070197, and the insertion/deletion were genotyped in trios, patients, and controls. The risk haplotype (TCA) was formed by the T allele of rs2004640, the C allele of rs2070197 (T⬎C), the insertion, and the A allele of rs10954213 (G⬎A) (i.e., TICA). As shown in Table 1, the same risk haplotype was found in 3 separate sets of patients and controls, from Spain, Germany, and Argentina, with a combined P value of 5 ⫻ 10⫺9. SNP no. rs2070197 was by itself strongly associated with SLE (P ⫽ 5.2 ⫻ 10⫺11) (Table 2), but bioinformatic analysis of the sequence did not provide any potential evidence that this SNP was functional. SNP no. rs10954213 was weakly associated (P ⫽ 0.0008), while the insertion/deletion was not by itself associated with SLE (Table 2). Of the functional SNPs, rs2004640 was the most strongly associated with SLE (P ⫽ 3.5 ⫻ 10⫺9). Using a set of trio families, we confirmed the haplotype and the presence of the insertion in the risk haplotype (Table 3). Thus, individuals with the risk haplotype (TICA) expressed V5 and V6 isoforms. The insertion was also present in a protective haplotype (GITG). Using multivariate logistic regression, we tested whether the 3 functional polymorphisms increased the risk of developing SLE. We did not find evidence of this. The 3 SNPs together did not improve the association found for rs2004640 alone (data not shown). Most Table 3. Genetic association of the risk haplotype of IRF5 in trio families* Haplotype† TDTA GITG GDTA TICA‡ Transmitted/ untransmitted Frequency 2 P 21.6/29.1 17.8/29.9 9.1/13.6 30.0/8.0 36.1 26.8 13.5 16.4 1.101 3.118 0.887 12.737 0.5491 0.0189 0.4028 0.000137 * The order of the single-nucleotide polymorphisms is rs2004640, rs2070197, and rs10954213. The trios from Spain and Denmark were from a total of 168 families with 71 complete trios. Transmitted alleles are expressed as absolute numbers. Haplotype frequencies and P values were obtained using Family Based Association Testing software. † D ⫽ deletion; I ⫽ insertion. ‡ Risk haplotype. individuals who were homozygous for rs2004640 and rs10954213 were heterozygous for the insertion (data not shown). DISCUSSION We have identified a novel structural variation determining the expression of the IRF5 isoforms. It was previously thought that isoforms in IRF5 resulted from the 5⬘-UTR and splicing at exon 6 (2). However, in the present study we found that a polymorphic insertion/ deletion in exon 6 and the alternative 3⬘ acceptor splice site in this exon determine the 2 major isoform groups in IRF5. Further, we showed that all 5⬘-UTRs of the IRF5 gene can lead to equal transcription of the 4 main isoform types, V1, V4, V5, and V6. Thus, the particular exon 1 used makes no difference in terms of the protein identity, and only exon 1A transcripts are present at high amounts in unstimulated cells and cells stimulated with IFN␣. It is well known that 5⬘-UTRs play an important role in the regulation of gene expression (8,9). Alternative UTRs with different indexes of translation efficiency can influence the amount of protein expressed. Some structural elements in the 5⬘-UTR determine the efficiency with which the protein synthesis will be initiated and elongated. These include the length and composition of the UTR, secondary structure, presence of upstream initiation codons, and the conformity of the sequence surrounding the initiation codon, with Kozak consensus defining the strength of the latter. In vitro analysis of the alternative 5⬘-UTRs of the IRF5 gene showed that the exon 1A–encoded UTR had a much higher translation potential than the other 3 UTRs, thus further supporting the hypothesis that the IRF5 protein isoform pool is made up of mainly exon 1A transcripts and very little or no exon 1B that is used when the associated T allele is present. Taken together, these data support the conclusion that SNP no. rs2004640 is not functionally important in determining the isoforms of IRF5. We also found, in the polyadenylation site of IRF5, an SNP (G⬎A) determining the length of the 1240 3⬘-UTR. The A allele leads to a short 3⬘-UTR, while the G allele encodes for a 1.5-kb long 3⬘-UTR and has 2 strong AREs known to be responsible for short half-life and rapid RNA degradation (10). Messenger RNA with a shorter 3⬘-UTR bearing no such signals should be more stable and present at high levels. Indeed, we found that the A allele of rs10954213 correlated with high levels of IRF5, particularly when cells were stimulated with IFN␣. AREs are abundant in cytokines and growth factor genes, for which rapid mRNA turnover is crucial for gene function (for review, see refs. 10 and 11). In mice, mutations disrupting AREs in the tumor necrosis factor ␣ (TNF␣) gene lead to elevated levels of circulating TNF␣ and hypersensitivity to stimulation with lipopolysaccharide (12,13). Thus, it is possible that SNP no. rs10954213 in the 3⬘-UTR poly(A) site of IRF5 could be the functional mutation responsible for the level of IRF5. The previously identified SNP no. rs2280714 is a perfect proxy for rs10954213, but is located 5 kb 3⬘ of IRF5 and therefore cannot clearly account for the high levels of IRF5. The isoforms of IRF5 do not differ in their DNA binding sites, but differ in the PEST and interaction domains (2). This may lead to the use of different coactivator or inhibitor proteins acting in specific cells or tissue, leading in turn to the promotion of a particular set of IRF5 targets. Exon 6 repeats forming the insertion/deletion polymorphism encode for the prolinerich region. Such structures, often present in transcription factors, may define the variety and affinity of other cofactors during transcription activity (14–16). Whether and how the insertion/deletion in the IRF5 gene influences gene function remain to be shown. IRF5 is clearly an SLE susceptibility gene, and in the present study we identified a risk haplotype, using samples from several populations. However, neither the insertion nor the highly expressed allele of rs10954213 were associated with SLE or appear to explain the genetic effect of IRF5 on susceptibility. Also, rs2004640, the functional polymorphism with the strongest association with SLE identified to date for IRF5, has a small functional impact. Therefore, it is highly conceivable that we have not identified all functional variation in IRF5 contributing to disease susceptibility and that one or more additional risk haplotypes in linkage disequilibrium with rs2004640 are yet to be found. Alternatively, despite its lower effect on genetic risk, rs10954213 may have a stronger functional impact on disease expression. This remains to be shown. Our results may explain why apparently functional polymorphisms may not be replicated in some KOZYREV ET AL studies, in which haplotypes have not been unambiguously defined. Defining the risk haplotypes in disease will be necessary for understanding of the functional genetics behind disease susceptibility. Addendum. Since the time this paper was submitted for publication, an article describing the polyadenylation site SNP of IRF5, no. rs10954213, has been published (Graham DS, Manku H, Wagner S, Reid J, Timms K, Gutin A, et al. Association of IRF5 in UK SLE families identifies a variant involved in polyadenylation. Hum Mol Genet 2006. E-pub ahead of print). ACKNOWLEDGMENTS We would like to thank Hong Yin for technical assistance with the samples, and the Uppsala Genome Center for sequencing. We acknowledge Rob Graham and David Altshuler for sharing information on the typing of SNP no. rs10954213 and SNP no. rs2070197 prior to publication. SNP no. rs10954213 was partly typed at the Broad Institute. We thank Adriana I. Scollo, Armando M. Perichon, and Mariano C. R. Tenaglia of CEDIM Diagnóstico Molecular y Forense SRL (Rosario, Argentina) for their help in DNA preparation of the Argentine samples, and Dr. Anne Voss for clinical help with the Danish samples. We would like to particularly thank the Lupus Patient Association of Asturias for help in the collection of family samples. AUTHOR CONTRIBUTIONS Dr. Alarcón-Riquelme had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study design. Kozyrev, Pons-Estel, Junker, Laustrup, AlarcónRiquelme. Acquisition of data. Kozyrev, Lewén, Reddy, Witte, Junker, Laustrup, Gutiérrez, Suárez, González-Escribano, Martı́n. Analysis and interpretation of data. Kozyrev, Lewén, Reddy, Junker, Laustrup, Martı́n, Alarcón-Riquelme. Manuscript preparation. Kozyrev, Pons-Estel, Witte, Junker, Laustrup, Alarcón-Riquelme. Statistical analysis. Kozyrev, Reddy, Alarcón-Riquelme. REFERENCES 1. Taniguchi T, Ogasawara K, Takaoka A, Tanaka N. IRF family of transcription factors as regulators of host defense. Annu Rev Immunol 2001;19:623–55. 2. Mancl ME, Hu G, Sangster-Guity N, Olshalsky SL, Hoops K, Fitzgerald-Bocarsly P, et al. Two discrete promoters regulate the alternatively spliced human interferon regulatory factor-5 isoforms: multiple isoforms with distinct cell type-specific expression, localization, regulation, and function. J Biol Chem 2005;280: 21078–90. 3. Sigurdsson S, Nordmark G, Goring HH, Lindroos K, Wiman AC, Sturfelt G, et al. Polymorphisms in the tyrosine kinase 2 and interferon regulatory factor 5 genes are associated with systemic lupus erythematosus. Am J Hum Genet 2005;76:528–37. 4. Graham RR, Kozyrev SV, Baechler EC, Reddy MV, Plenge RM, Bauer JW, et al. A common haplotype of interferon regulatory factor 5 (IRF5) regulates splicing and expression and is associated SLE AND STRUCTURAL INSERTION/DELETION VARIATION IN IRF5 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. with increased risk of systemic lupus erythematosus. Nat Genet 2006;38:550–5. Tan EM, Cohen AS, Fries JF, Masi AT, McShane DJ, Rothfield NF, et al. The 1982 revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum 1982;25:1271–7. Jacobs GH, Stockwell PA, Tate WP, Brown CM. Transterm— extended search facilities and improved integration with other databases. Nucleic Acids Res 2006;34:D37–40. Horvath S, Xu X, Laird NM. The family based association test method: strategies for studying general genotype—phenotype associations. Eur J Hum Genet 2001;9:301–6. Wang G, Guo X, Floros J. Differences in the translation efficiency and mRNA stability mediated by 5⬘-UTR splice variants of human SP-A1 and SP-A2 genes. Am J Physiol Lung Cell Mol Physiol 2005;289:L497–508. Gray NK, Wickens M. Control of translation initiation in animals. Annu Rev Cell Dev Biol 1998;14:399–458. Clark AR, Dean JL, Saklatvala J. Post-transcriptional regulation of gene expression by mitogen-activated protein kinase p38. FEBS Lett 2003;546:37–44. Khabar KS. The AU-rich transcriptome: more than interferons and cytokines, and its role in disease. J Interferon Cytokine Res 2005;25:1–10. Jacob CO, Lee SK, Strassmann G. Mutational analysis of TNF-␣ gene reveals a regulatory role for the 3⬘-untranslated region in the genetic predisposition to lupus-like autoimmune disease. J Immunol 1996;156:3043–50. Kontoyiannis D, Pasparakis M, Pizarro TT, Cominelli F, Kollias G. Impaired on/off regulation of TNF biosynthesis in mice lacking TNF AU-rich elements: implications for joint and gut-associated immunopathologies. Immunity 1999;10:387–98. Randall RA, Germain S, Inman GJ, Bates PA, Hill CS. Different Smad partners bind a common hydrophobic pocket in Smad2 via defined proline-rich motif. EMBO J 2002;21:145–56. Swingler TE, Bess KL, Yao J, Stifani S, Jayaraman PS. The proline-rich homeodomain protein recruits members of the Groucho/Transducin-like enhancer of split protein family to co-repress transcription in hematopoietic cells. J Biol Chem 2004;279: 34938–47. Prado F, Vincent G, Cardalda C, Beato M. Differential role of the proline-rich domain of nuclear factor 1-C splice variants in DNA binding and transactivation. J Biol Chem 2002;277:16383–90. APPENDIX A: THE ARGENTINE, GERMAN, AND SPANISH COLLABORATIVE GROUPS Members of the Argentine Collaborative Group are as follows: Hugo R. Scherbarth, MD, Pilar C. Marino, MD, Estela L. Motta, 1241 MD (Hospital Interzonal General de Agudos Dr. Oscar Alende, Mar del Plata), Susana Gamron, MD, Cristina Drenkard, MD, Emilia Menso, MD (UHMI 1, Hospital Nacional de Clı́nicas, Universidad Nacional de Córdoba, Córdoba), Alberto Allievi, MD, Guillermo A. Tate, MD (Organización Médica de Investigación, Buenos Aires), Jose L. Presas, MD (Hospital General de Agudos Dr. Juán A. Fernandez, Buenos Aires), Simon A. Palatnik, MD, Marcelo Abdala, MD, Mariela Bearzotti, PhD (Universidad Nacional de Rosario y Hospital Provincial del Centenario, Rosario), Alejandro Alvarellos, MD, Francisco Caeiro, MD, Ana Bertoli, MD (Hospital Privado, Centro Medico de Córdoba, Córdoba), Sergio Paira, MD, Susana Roverano, MD (Hospital José M. Cullen, Santa Fe), Cesar E. Graf, MD, Estela Bertero, PhD (Hospital San Martı́n, Paraná), Cesar Caprarulo, MD, Griselda Buchanan, PhD (Hospital Felipe Heras, Concordia, Entre Rı́os), Carolina Guillerón, MD, Sebastian Grimaudo, PhD, Jorge Manni, MD (Instituto de Investigaciones Médicas Alfredo Lanari, Buenos Aires), Luis J. Catoggio, MD, Enrique R. Soriano, MD, Carlos D. Santos, MD (Hospital Italiano de Buenos Aires y Fundación Dr. Pedro M. Catoggio para el Progreso de la Reumatologı́a, Buenos Aires), Cristina Prigione, MD, Fernando A. Ramos, MD, Sandra M. Navarro, MD (Hospital Provincial de Rosario, Rosario), Guillermo A. Berbotto, MD, Marisa Jorfen, MD, Elisa J. Romero, PhD (Hospital Escuela Eva Perón, Granadero Baigorria, Rosario), Mercedes A. Garcia, MD, Juan C. Marcos, MD, Ana I. Marcos, MD (Hospital Interzonal General de Agudos General San Martı́n, La Plata), Carlos E. Perandones, MD, Alicia Eimon, MD (Centro de Educación Médica e Investigaciones Clı́nicas, Buenos Aires), and Cristina G. Battagliotti, MD (Hospital de Niños Dr. Orlando Alassia, Santa Fe). Members of the German Collaborative Group are as follows: K. Armadi-Simab, MD, Wolfgang L. Gross, MD (University Hospital of Schleswig-Holstein, Campus Luebeck, Rheumaklinik Bad Bramstedt, Luebeck), Erika Gromnica-Ihle, MD (Rheumaklinik BerlinBuch, Berlin), Hans-Hartmut Peter, MD (Medizinische Universitaetsklinik, Abteilung Rheumatologie und Klinische Immunologie, Freiburg), Karin Manger, MD (Medizinische Klinik III derFAU Erlangen-Nuernberg, Erlangen), Sebastian Schnarr, MD, Henning Zeidler, MD (Abteilung Rheumatologie, Medizinische Hochschule Hannover, Hannover), and Reinhold E. Schmidt, MD (Abteilung Klinische Immunologie, Medizinische Hochschule Hannover, Hannover). Members of the Spanish Collaborative Group are as follows: Norberto Ortego, MD (Hospital Clı́nico San Cecilio, Granada), Enrique de Ramón, MD (Hospital Carlos Haya, Malaga), Juan JiménezAlonso, MD (Hospital Virgen de las Nieves, Granada), Julio SánchezRomán, MD (Hospital Virgen del Rocio, Sevilla), and Miguel Angel López-Nevot, MD (Hospital Virgen de las Nieves, Granada).