To perform a genetic association study using markers in the interleukin-1 (IL-1) gene cluster and the IL-4/IL-4 receptor system genes, seeking evidence for involvement in the onset or the erosive outcome of rheumatoid arthritis (RA).
To perform a genetic association study using markers in the interleukin-1 (IL-1) gene cluster and the IL-4/IL-4 receptor system genes, seeking evidence for involvement in the onset or the erosive outcome of rheumatoid arthritis (RA).
We tested the allelic distribution of IL-1A (+4845), IL-1B (−511), IL-1B (+3954), IL-1RN (+2018), IL-4 variable number of tandem repeat (VNTR), and IL-4R (+1902) in 233 patients with RA, 99 with polymyalgia rheumatica, and 148 ethnically matched controls. We analyzed the frequency of these gene variants in respect to presence of disease, but also to the degree of radiologic erosions (Larsen score) as a function of disease duration in 157 patients who had available radiographs of both hands.
None of the 6 genetic polymorphisms was significantly different in frequency between RA patients and healthy controls or patients with polymyalgia rheumatica. Among RA patients, the rarer (#2) alleles of IL-4 VNTR and IL-1B (−511) were both associated with a milder Larsen score progression: The slope of Larsen progression in the rare allele groups diverged significantly from those of the frequent allele groups after approximately 20 years of disease duration (P < 0.001).
None of the markers tested were shown to be associated with increased or decreased risk of RA. The rarer alleles of IL-4 VNTR and IL-1B (−511) appear to be associated with a less severe course in RA of long duration.
Rheumatoid arthritis (RA) is a multifactorial disease with a clear genetic background. Multiple genes are thought to be involved in disease susceptibility, whereas others could be more important as modulators of disease severity (1, 2). As such, they may account for the wide spectrum of clinical expression, ranging from mild, nondestructive forms to severe and rapidly debilitating disease.
HLA–DRB1 is associated with RA (3) and to some extent with disease severity (2, 4), but probably accounts for only about one-third of the genetic component of RA (5). Even when combined with other indicators of destruction, such as rheumatoid factor (RF), the specificity and sensitivity of HLA–DRB1 alleles as prognostic factors for destructive arthritis are frequently considered low and of limited value to clinicians (6). Other genetic markers of disease severity would therefore be welcome, especially in the light of the present tendency to treat patients with severe disease earlier and more aggressively.
Cytokines have been shown to play a key role in the pathogenesis of RA. Among them, interleukin-1 (IL-1) is thought to be a central mediator of joint destruction (7–11) and IL-1 receptor antagonist (IL-1Ra) proved protective in animal and human clinical studies (12, 13). IL-1 family factors are encoded by at least 3 genes clustered in human chromosome 2q12: IL-1A, IL-1B, and IL-1RN, respectively coding for IL-1α, IL-1β, and IL-1Ra. IL-4, whose gene is located in the long arm of chromosome 5, inhibits production of proinflammatory cytokines and proteases, and has shown antiinflammatory and protective properties (14, 15). IL-4 is also a key factor in the polarization of T helper cells toward Th2 differentiation, thought to be deficient in RA (16). IL-4 can bind at least 2 types of receptors (IL-4Rα-γc and IL-4Rα-IL-13Rα1), both comprising the common α chain.
Cytokine gene variants have been associated with disease severity in a number of chronic inflammatory diseases (17) including erosive RA (18). In this study, we investigated possible associations of joint destruction in RA with IL-1A (+4845), IL-1B (−511), IL-1B (+3954), IL-1RN (+2018), interleukin-4 variable number of tandem repeats (IL-4 VNTR), and interleukin-4 receptor (IL-4R) (+1902). IL-1A (+4845) is a single nucleotide polymorphism (snp) in exon V of IL-1A, which is in 100% linkage with IL-1 (−889), a promoter polymorphism associated with early pauciarticular juvenile rheumatoid arthritis and chronic iridocyclitis (19). IL-1B (−511) and IL-1B (+3954) have been associated with erosive RA (18). IL-1RN (+2018) is a single nucleotide polymorphism in 100% linkage disequilibrium with a VNTR polymorphism found to be associated with other autoimmune diseases, such as systemic lupus erythematosus (20) and ulcerative colitis (21). The IL-4 VNTR described in the third intron of IL-4 (22) has been associated with age at onset of multiple sclerosis (23). IL-4R (+1492) is an A to G transition shown to be associated with hyper IgE syndrome and atopy (24).
We first analyzed a possible correlation of these polymorphisms with the predisposition to RA. Comparisons were made with a control group and with a group of patients with polymyalgia rheumatica (PMR), a disease genetically (25) and clinically related to RA but in which joint erosion does not occur. We then investigated the association of these polymorphisms with joint damage progression in RA, using the Larsen score as a function of disease duration.
A total of 233 consecutive patients with RA and 148 healthy matched controls among volunteer bone marrow donors from the Swiss Registry were recruited in the 5 university hospitals of Switzerland (Basel, Bern, Geneva, Lausanne, and Zurich), as well as in 3 nonuniversity primary referral centers (Table 1). All RA patients fulfilled at least 4 of the 7 American College of Rheumatology 1987 criteria (26). A group of 99 PMR patients, previously described elsewhere (25), were also recruited in the same centers during the same period of time. For each center, the number of controls was proportional to the number of RA patients, to take into account possible local minor variations in allelic frequencies. The majority of patients and controls were of European descent (less than 2% were not).
|All patients (n = 233)||Patients with radiographs (n = 157)||P†|
|Age in 1996, mean (SD)||63.3 (13.7)||62.0 (13.5)||0.040|
|Women, n (%)||174 (75)||117 (75)||1.00|
|Age at disease onset, years (SD)||48.3 (15.8)||47.0 (15.9)||0.066|
|Disease duration, years (SD)||15.0 (12.3)||15.0 (11.9)||0.95|
|Rheumatoid factor + (%)||172 (74)||120 (76)||0.21|
|With nodules (%)||65 (28)||42 (27)||0.64|
|HLA–DRB1 shared epitope (%)|
|absent||53 (23)||29 (18)||overall: 0.009|
|one copy||126 (54)||96 (61)||trend test: 0.54|
|two copies||53 (23)||32 (20)|
We collected clinical and demographic data from RA patients, including date of the first symptoms, presence or absence of RF and nodules, self-report information on joint destruction, and enough blood to perform the genetic analysis. Because data on joint destruction and erosions proved difficult to interpret, we subsequently asked for recent hand radiographs (less than 2 years old) for all patients, and obtained them for 157 patients.
All radiographs were analyzed independently by 2 trained rheumatologists (SS and PAG) according to a standardized Larsen method (27) using standard reference films. The 2 rheumatologists were blinded to the genetic background of the patients. In case of divergence in their scores, radiographs were reviewed and an agreement was reached. Wrist, metacarpophalangeal joints 2–5 and proximal interphalangeal joints 2–5 were scored on a 5-point scale (0 = no abnormalities; 1 = slight abnormalities, such as joint space narrowing or band-like osteoporosis; 2 = small but definite erosions; 3 = medium erosions; 4 = severe destructive abnormalities; 5 = mutilating abnormalities). The score of the wrist was then multiplied by 2. The total score ranged from 0 to 100.
Uncoagulated blood was taken from patients and controls and stored frozen at −20°C until DNA extraction. DNA was extracted by a modification of the salt-out technique (Nucleon TM, Scotlab, UK) and stored at a final concentration of 200 μl/ml until used for genotyping. All genotyping at the IL-1 cluster was performed by polymerase chain reaction-based methods, which have been published and validated extensively (17).
According to previously used nomenclature, alleles were named according to their frequency in the general population, i.e., allele 1 being the frequent or common allele, allele 2 the rarer form.
Methods used for IL-4R and IL-4 VNTR are detailed in Appendix A. Briefly, the IL-4 VNTR was genotyped by agarose gel electrophoresis sizing, and the IL-4R snp was typed by 5′ nuclease assay (TaqMan) (28) using a ABI-PE 7200 scanning fluorimeter.
To test whether genetic polymorphisms were risk factors for the occurrence of RA, we compared the distributions of 6 polymorphisms (frequent homozygote, heterozygote, and rare homozygote) in patients with RA and in 2 control groups: patients with PMR and healthy controls. A linear trend test (chi-square, one degree of freedom) was performed on each comparison; this test assumes that the risk of RA will be progressively higher (or lower) in heterozygotes and rare type homozygotes, compared with frequent homozygotes.
Then we studied the role of genetic polymorphisms in the progression of RA, measured by the radiologic severity of joint erosions. Because only 157 of the 233 RA patients had radiographs available, we first compared the subgroups with and without radiographs in terms of age, sex, age at onset of RA, RF, presence of nodules, and HLA–DRB1 genotype. Differences between mean values of continuous variables were tested by the Student's t-test, and differences in categorical variables by chi-square tests.
Among the 157 patients with available radiographs, we examined how the Larsen score evolved as a function of disease duration, overall and in subgroups of patients stratified by polymorphism status (dichotomized as frequent homozygotes versus one or two rare alleles). This exploratory analysis was done using nonparametric locally weighted scatter plot smoothing regression (lowess) (29). This method does not impose a functional form (e.g., straight line) to the relationship between the variables. It is closely related to a moving average: The mean Larsen score is plotted as a function of time elapsed since diagnosis. Statistical significance of slopes (Larsen units per year) was based on linear regression models. When the nonparametric regression suggested a change in the slope, we used linear models where the slope was allowed to change at a point in time suggested by the nonparametric regression plot. Differences in slope changes between patient subgroups were tested using an interaction test between slope change and polymorphism indicator. All analyses were conducted using SPSS software version 8.0 (SPSS, Chicago, IL).
Each polymorphism was analyzed in RA patients and compared with healthy controls and PMR patients (Table 2). The distribution of alleles was equivalent in the 3 populations for the 6 polymorphisms analyzed. Only 2 of 12 trend tests hinted at a possible relationship: IL-1A (+4845) rare allele was weakly positively associated with the risk of RA (P = 0.078 when compared with healthy controls) and IL-4 VNTR rare allele was negatively associated with the risk of RA (P = 0.11 when compared with healthy controls). However, in the context of multiple tests, such results are also entirely compatible with random sampling differences.
|Genotype 1: frequent 2: rare||Healthy controls n (%)||P (RA vs. controls, linear trend)||RA n (%)||P (RA vs. PMR, linear trend)||PMR n (%)|
|IL-1A (+4845)||11||76 (53)||0.08||105 (46)||0.27||53 (54)|
|12||60 (42)||101 (44)||38 (39)|
|22||8 (6)||24 (10)||7 (7)|
|IL-1B (+3954)||11||89 (62)||0.33||128 (58)||0.65||61 (64)|
|12||48 (33)||86 (37)||29 (30)|
|22||7 (5)||16 (7)||6 (6.3)|
|IL-1B (−511)||11||64 (46)||0.88||106 (46)||0.60||55 (56)|
|12||61 (44)||99 (43)||34 (34)|
|22||15 (11)||26 (11)||10 (10)|
|IL-1RN (+2016)||11||73 (51)||0.74||129 (57)||0.44||53 (54)|
|12||57 (40)||79 (35)||40 (41)|
|22||12 (9)||18 (8)||5 (5)|
|IL-4 VNTR||11||96 (72)||0.11||174 (76)||0.83||73 (75)|
|12||31 (23)||51 (22)||22 (23)|
|22||6 (5)||4 (2)||2 (2)|
|IL-4R (+1902)||11||97 (70)||0.67||150 (68)||0.45||63 (66)|
|12||38 (28)||66 (30)||31 (33)|
|22||3 (2)||6 (3)||1 (1)|
The two groups were similar in sex distribution, duration of RA, prevalence of nodules, and RF positivity (Table 1). Patients with radiographs were slightly younger both at disease onset and at inclusion in the study, and their HLA–DRB1 profile was different: The proportion of heterozygotes was greater than in patients without radiographs, but proportions of both types of homozygotes were lower. Nevertheless, more patients among those with radiographs carried 1 or 2 copies of HLA–DRB1 (P = 0.029). In aggregate, these analyses suggest that patients with radiographs were reasonably representative of the whole population of RA patients.
Graphical exploration analysis suggested that the Larsen score increased approximately linearly with disease duration. The nonparametric regression curve was almost straight, and in a simple linear regression model, the slope was 1.03 point of Larsen score per year of disease duration (95% confidence interval 0.78–1.28). Duration of disease explained 30% of the variance in Larsen scores (adjusted r2 = 0.30).
We observed a significant difference between the Larsen progression of frequent and rare allele subgroups for 2 polymorphism sites: IL-1B (−511) and IL-4 VNTR. Because patients homozygous for rare alleles of these 2 polymorphisms were not frequent enough to allow a separate analysis [n = 16 for IL-1B (−511) and 2 for IL-4 VNTR] and were similar to heterozygotes, rare homozygous and heterozygous patients were pooled together for each gene variant (patients with 2 frequent alleles were compared with patients carrying either 1 or 2 copies of rare alleles).
For IL-1B (−511) (Figure 1A), the Larsen score slopes were very similar for both patient groups with less than 20 years of disease duration, but an important difference appeared beyond 20 years: In patients who were carriers of the rare allele, the slope dramatically decreased and came close to 0, whereas it continued to rise for frequent homogyzotes. In a linear regression spline model allowing a change in slope at 20 years, this change of slope was statistically significant (Table 3). This model explained 39% of variance in Larsen scores. A similar pattern was found for IL-4 VNTR (Figure 1B). The Larsen score slopes of the patients positive and negative for the rare allele also diverged markedly for patients with more than 20 years disease duration. The difference between the 2 slopes was again statistically significant (Table 3). These significance tests on changes in slopes should be interpreted with caution because the statistical model we used was not stated a priori, but was based on the exploratory analysis of the scatterplots. This approach increases the likelihood that a chance result may be labeled as statistically significant.
|Polymorphism||Slope in frequent homozygotes at all times and in carriers of rare allele before 20 years||Difference in slopes between frequent homozygotes and carriers of rare allele, after 20 years||Slope in carriers of rare allele, after 20 years|
|Larsen points/year||95% CI||Larsen points/year||95% CI||Larsen points/year||95% CI|
|IL-1B (−511)||1.57||1.25 to 1.87||−1.71||−1.02 to −2.41||−0.14||−1.08 to 0.79|
|IL-4 VNTR||1.37||1.09 to 1.64||−1.72||−1.02 to −2.41||−0.35||−1.22 to 0.52|
|Both IL-1B (−511) and IL-4 VNTR||1.36||1.08 to 1.64||−1.73||−1.04 to −2.42||−0.37||−1.23 to 0.49|
For all other tested polymorphisms, i.e., IL-1A (+4845), IL-1B (−3954), IL-1RN (+2016), and IL-4R (+1902), no detectable differences between the Larsen score curves of the carriers of rare and frequent alleles were observed. Analysis for patients with or without DR1/DR4 subtypes was attempted but interpretation proved to be impossible because of the small sample sizes.
The relationship between rare alleles of IL-1B (−511) and IL-4 VNTR was then investigated. We separated patients into 4 groups who carried one of these rare alleles, the other, neither, or both (Figure 1C). The change of slope after 20 years appeared to concern only patients who had mutations for both IL-1B (−511) and IL-4 VNTR. The change of slope in this subgroup, compared with all other patients, was statistically significant (Table 3). However, there were virtually no patients positive for the IL-4 VNTR mutation but negative for the IL-1B (−511) mutation after 20 years. Hence it is equally plausible that both mutations in conjunction are associated with fewer erosions after 20 years of disease, or that mutation in IL-4 VNTR alone is sufficient.
Although none of the 6 polymorphisms studied in our Swiss population of RA patients appeared to protect from or to confer a trend for developing RA, our data show a significant correlation between the carriage of IL-1B (−511) or IL-4 VNTR #2 alleles and lower joint destruction in RA.
The differences between the #1 and #2 alleles were evident only for patients with more than 15–20 years of disease duration. This can be explained by the fact that the positive effects associated with the presence of these 2 rare alleles may not be of a sufficient magnitude to be detectable earlier with the number of patients recruited in our study. Indeed, IL-1B and IL-4 are only 2 of the numerous genes potentially involved in the determination of RA severity, and the magnitude of their involvement has yet to be determined. Additionally, patients with more recent onset of RA (<15 years disease duration) have progressively benefited from the recent progress in treatment interventions, e.g., low dose prednisone (30), methotrexate at higher doses, leflunomide, and anti-tumor necrosis factor agents (31, 32). All these treatments have been shown to markedly reduce joint destruction evaluated through Larsen score, thus reducing the power of studies based on radiograph analysis (33). This could have prevented detection of IL-1B (−511) and IL-4 VNTR effects in patients with more recent disease onset, thus delaying the point of divergence between the curves. Correspondingly, we observed a similar delayed drift apart in another study, where RF was analyzed in the same population with the same statistical method. Even though the presence of RF has been well recognized to be a strong independent marker of bone damage, the divergence of the 2 curves was visible only at 12 years (34).
A decrease in reliability of the Larsen score in long-duration RA would be another possible explanation. However, Larsen score was recently used to show that the progress of bone damage in RA follows a mathematical function of time for up to 30 years after disease onset (35).
Nevertheless, the possibility still remains that the observed association is merely a chance finding. Only replication of similar results in different patient populations will clarify this issue.
Both IL-1 and IL-4 have been involved in bone destruction in opposite ways. In vitro and mouse studies indicate a direct and crucial role of IL-1 in cartilage destruction (7–11), and anti-IL-1 therapy was shown to slow progression of erosions in human clinical studies (13). On the other hand, IL-4 is an antiinflammatory cytokine with protective effects on bone and cartilage both in animal models (14) and in ex vivo experiments (15).
Many cytokine polymorphisms have been correlated with inflammatory diseases and several hypotheses made about possible mechanisms. No specific change in gene regulation associated with IL-1 (−511) has been described. However, this marker has very recently been shown to be associated with another promoter base change in the IL-1 promoter (F.S. di Giovine and C.C. Campbell, unpublished observations). A single base change in IL-1 (−31) that changes the TATA box and is associated with differential binding of transcription factors and enhanced susceptibility to Helicobacter pylori infection and gastric cancer (36, 37). Likewise, in a study of scleroderma, IL-4 VNTR polymorphism has been hypothesized to affect transcriptional activity, inducing IL-4 overexpression and leading to a shift in the Th1/Th2 balance towards Th2 (23). Increased expression of Th2 cytokines may protect against the destructive actions of Th1 and other proinflammatory cytokines. Our results could also be explained by linkage disequilibrium: IL-4 is located in the sensitive 5q31.1 area in the 5th chromosome, which codes for several Th2 cytokines; uncharacterized polymorphisms in this region have already been linked with altered IgE production and atopy (38, 39).
Cantagrel et al (40) found that the combination of an RA related HLA–DR shared epitope and allele E2 in IL-1B exon 5 was associated with erosive disease, but observed no correlation between IL-1B (−511) or IL-4 VNTR and joint destruction in their study, based on a population with early RA. In contrast, our study analyzed possible long-term effects of these polymorphisms: Our RA patients had a mean of 11 years of disease activity and the effects of IL-1B (−511) or IL-4 VNTR were observed only for patients with longer disease duration.
Cantagrel at al observed an increased frequency of the IL-4 VNTR rare allele (called RP1 in their work) in RA patients (40). Differences in the population genetic backgrounds as well as in recruitment, including the mean severity of the cases, could explain this variation. The carriage rate found in our populations is in agreement with other studies and was consistent in the 3 populations studied. In contrast, a recent study examining IL-4 VNTR in RA showed results fully consistent with our observations (41).
Our study was based on a cross-sectional analysis with the nonparametric lowess regression, considering joint damage progression as a function of time. This approach takes into account the fact that the Larsen score is widely dependent on disease duration and thus avoids a major drawback frequently weakening transversal studies. We think this method represents a good alternative approach in so far as a prospective study addressing such questions is very difficult to design with the availability of new and more effective treatments, adjustable to disease severity. At this point, however, we cannot conclude whether the genetic markers we examined determine and/or predict, or merely accompany, a more or less favorable clinical course. Prospective as well as in vitro studies are in progress, aiming at resolving these issues.
The respective protective influence of IL-1B (−511) and IL-4 VNTR rare alleles on RA severity found in this study surely require further confirmation in other populations. In any case, their participation will be, at best, of moderate amplitude and probably of little use to clinicians. Nevertheless they may one day become part of a more comprehensive picture of the genetics of RA, which may eventually contribute to more accurate and advanced prediction of disease outcome.
The technical assistance of Lydia Bertrand, Françoise Mezin, Janine Timms, Ursula Spenato, and Madeleine Vuillet is gratefully acknowledged.
Forward primer: 5′-GTA AAT AGG CTG AAA GGG GGA AA 3′
Reverse primer: 5′-CAT CTT TTC CTC CCC TGT ATC TT 3′
Polymerase chain reaction (PCR) cycles: (95°C, 2 minutes) × 1; (95°C, 1 minute; 56°C, 1 minute; 72°C, 30 seconds) × 40; (72°C, 5 minutes) × 1.
Allele 1 = 3 repeats = 342 bp PCR product.
Allele 2 = 2 repeats = 272 bp.
Alelle 3 = 4 repeats = 412 bp.
PCR conditions: Genomic DNA at 200 ng/25 μl reaction. MgCl2 at 2 mM and primers at 0.5 mM final concentration.
Probe 1: 5′-C ( · FAM) AT GTA CAA ACT CCT GAT AGC CAC TGG TG ( · TAMRA)-3′
Probe 2: 5′-C ( · TET) CAT GTA CAA ACT CCC GAT AGC CAC TGG ( · TAMRA)-3′
Forward: 5′-AGG CTT GAG AAG GCC TTG TAA-3′
Reverse: 5′-CCG AAA TGT CCT CCA GCA T-3′
Cycling: (50°C, 2 minutes) × 1; (95°C, 1 minute; 95°C, 15 seconds; 61°C, 1 minute) × 40.