Article Text
PDF
Abstract
Objective: To determine whether tumour necrosis factor (TNF) gene polymorphisms and/or the shared epitope are genetic predictors of the response to adalimumab (ADA) in rheumatoid arthritis (RA).
Methods: This ancillary study to the Research in Active Rheumatoid Arthritis (ReAct) Phase IIIb study included a large cohort of Caucasian patients with RA from France (n = 388) treated with ADA plus methotrexate (MTX) (n = 182), ADA plus any other DMARD (n = 98) or ADA alone (n = 108). The primary outcome was ACR50 at 12 weeks. Patients underwent genotyping for HLA-DRB1 and three TNF gene polymorphisms (–238A/G,–308A/G and–857C/T). Extended haplotypes involving HLA-DRB1 and TNF loci were reconstructed using the PHASE program.
Results: A total of 151 patients (40%) had an ACR50 response at week 12. Neither the number of HLA-DRB1 shared epitope copies nor presence of the three TNF polymorphisms tested separately was significantly associated with ACR50 response at week 12. However, haplotype reconstruction of the TNF locus revealed that the GGC haplotype (–238G/–308G/–857C) in a homozygous form (i.e. present in more than half of the patients) was significantly associated with a lower ACR50 response to ADA at 12 weeks (34% vs. 50% in patients without the haplotype) (p = 0.003; pa = 0.015). This effect was more important in the subgroup of patients concomitantly treated with MTX.
Conclusion: This large pharmacogenetic study provides preliminary data indicating that a single TNF locus haplotype (–238G/–308G/–857C), present on both chromosomes is associated with a lower response to ADA, mainly in patients treated with ADA and MTX.
Statistics from Altmetric.com
The last 10 years have led to the recognition of tumour necrosis factor (TNF)-α as one of the cornerstone cytokines involved in the pathogenesis of rheumatoid arthritis (RA). Such results have provided the basis for the development of TNF blockers for treating RA, which are efficient in about 70% of patients, but 30% are resistant. Taking into account the cost of these treatments in RA, the persisting questions about potential long-term adverse events (infections and cancers) and the availability of other efficient biotherapies, the identification of predictive factors of response to treatment is a key issue. In this field, pharmacogenetic approaches give promising hope. Nevertheless, few studies have been performed to date and some have led to contradictory results, especially those concerning the role of the shared epitope (SE) and/or TNF–308A/G polymorphism1–12 (table 1).
The aim of this study was to determine whether the SE and/or TNF gene polymorphisms, analysed separately or after haplotype reconstruction, are genetic predictors of response to adalimumab (ADA), a fully humanised monoclonal anti-TNF antibody, in patients with RA.
METHODS
Patients
This pharmacogenetic study was ancillary to the Research in Active Rheumatoid Arthritis (ReAct) protocol performed at various sites in Europe and Australia.13 In the parent ReAct study, 6610 patients were included to assess the safety and effectiveness of ADA, in combination with a variety of disease-modifying antirheumatic drugs (DMARDs).
This pharmacogenetic study includes a large cohort of 396 French patients included in the ReAct study. All the patients provided written informed consent. The study was approved by the Bicêtre Hospital Ethics Committee. Eight patients were excluded owing to their Asian or African descent. Thus, 388 patients were eligible for the study and underwent treatment with ADA plus MTX (n = 182), ADA plus any other DMARD (n = 98) or ADA alone (n = 108). Twelve other patients were excluded from the main analyses concerning association with the American College of Rheumatology (ACR)5014 response because of missing ACR50 data at 12 weeks (see table 2).
Collection of clinical and biological data and outcome measures
The clinical and biological data were from the original ReAct protocol. At baseline and weeks 2, 6 and 12, scores for variables necessary to assess Modified Disease Activity Score (MDAS) and ACR response were recorded,15 as were scores on the Health Assessment Questionnaire-Disability Index (HAQ). Outcomes were measured after 12 weeks of treatment. The primary outcome for the genetic study was ACR50 response; secondary outcomes were ACR20 and ACR70 responses.
Genetic polymorphisms in TNF and HLA genes
We chose to analyse the TNF gene polymorphisms located in the promoter region on the basis of a previous report showing four main haplotypes consisting of TNF +488,–238 and–308 single-nucleotide polymorphisms (SNPs) among a Caucasian population from the UK.16 However, TNF +488 has been reported to have strong linkage disequilibrium (LD) with TNF–857 (LD value, Somer’s D′ = 0.92 in a UK Caucasian population17). Moreover, TNF–857 was recently reported to influence clinical response to etanercept,9 so we chose to genotype TNF–857 instead of TNF +488. Haplotype reconstructions with the three studied SNPs have led, as expected, to the four main haplotypes (cited below) in our French Caucasian population of patients with RA.
Because definite HLA-DRB1 alleles have been previously reported to play an important part in RA susceptibility—the SE hypothesis18—and severity,19 we genotyped our RA patients for HLA-DRB1 alleles. The alleles considered to have the SE were HLA-DRB1*0101, *0102, *0401, *0404, *0405, *0408, *0413, *1001 and *1402.18 To analyse the SE contribution in response to ADA treatment, patients were classified as having 0, 1 or 2 copies of the SE, or as being SE carriers or not. Extended haplotypes comprising HLA-DRB1 alleles and TNF SNPs were reconstructed as well.
Genotyping methods
Patients were genotyped for HLA-DRB1 and three TNF gene polymorphisms (–857C/T,–308A/G and–238A/G). HLA-DRB1 alleles were determined by polymerase chain reaction (PCR) amplification and DNA sequencing using an ABI 3700 sequencer (PE Applied Biosystems, Courtaboeuf, France).
TNF–238A/G gene polymorphisms were genotyped by mismatch PCR–restriction fragment full length polymorphism using the MspI restriction enzyme.
TNF–308A/G polymorphism was genotyped by allelic discriminating TaqMan PCR by use of the PreDeveloped TaqMan assay kit C_7514879 (PE Applied Biosystems).
TNF–857C/T genotyping was performed by a TaqMan 5′-nuclease assay (PE Applied Biosystems) with allele-specific fluorogenic oligonucleotide probes (5′CCCTGTCTTCATTAAG and 5′CCCTGTCTTCGTTAAG) using an ABI 7000 sequence detection system (PE Applied Biosystems).
Statistical analysis
All quantitative data are expressed as the mean (SD) and all qualitative data as frequencies and percentages. All genotyped SNPs were in Hardy–Weinberg equilibrium. Within the TNF gene, a measure of the LD between the different SNPs was estimated using Somer’s D′. As the LD was significantly different among all TNF SNPs and between TNF SNPs and HLA-DRB1 alleles, we also considered the haplotypes for TNF and extended haplotypes comprising HLA DRB1 alleles and TNF. We used the software PHASE (version 2.1) to perform haplotype reconstructions.20 21 The average probability of PHASE certainty in haplotype inference was 99% for TNF haplotypes and 83% for SE-TNF extended haplotypes.
For each gene, genotypes and haplotypes were tested for the association with the ACR50 response to ADA at week 12. Differences in genotype distribution for efficacy were tested using contingency tables with the two-sided χ2 test. A Bonferroni correction was applied for five multiple comparisons (three SNPs, the SE and TNF haplotype). Both unadjusted and adjusted (denoted by pa) p values are presented. p<0.05 was considered significant.
The factors influencing the probability of achieving an ACR50 response at week 12 were studied through multivariate logistic regression. First, univariate logistic regressions were performed to screen covariates using log-likelihood ratio tests. We kept all the covariates for which p<0.15. A multivariate regression model was then built, including all candidate covariates selected in the previous step, and the final model was selected through backward selection, using log-likelihood ratio tests. At this stage, covariates were kept in the model if p<0.05.
RESULTS
Description of the cohort
The baseline characteristics of patients are presented in table 2. The profile of clinical response of the 388 patients included in this pharmacogenetic study was the same as that of the entire ReAct population (6610 patients).13 At 12 weeks, 40% of the patients showed an ACR50 response, 71% an ACR20 response and 16% an ACR70 response. MTX adjunction to ADA therapy was significantly associated with a better ACR50 response both in univariate (p = 0.005) and multivariate analyses (odds ratio (OR), 1.68; 95% confidence interval (CI), 1.07–2.66; p = 0.026). In the subgroup of patients not receiving methotrexate, there was a non-significant trend for a better response when other DMARDs were administered in addition to ADA (38% of ACR50 responders versus 29% in patients given only ADA, p = 0.21). Only 28 (7%) of our patients were previously treated with anti-TNF stopped either for intolerance (10 patients) or for absence of efficacy (18 patients). Moreover, as in the whole ReAct study, the rate of ACR50 response was the same in patients previously treated with anti-TNF than in others: 32% vs 41% (p = 0.43).
Genotype distributions, shown in table 3, were consistent with data from public databases for Caucasians (http://www.ncbi.nlm.nih.gov/projects/SNP). One patient had the rare AA genotype for TNF–238 G>A, and three patients had the rare TT genotype for TNF–857 C>T.
SE could be determined in 333 patients. The distribution of the SE among patients was as follows: 26% had no copies, 48% one copy, and 26% two copies, for a total of 74% carriers.
In 354 patients in whom we determined the three SNPs of the TNF promoter (238A/G,–308A/G and–857C/T), we identified four main haplotypes (eg, the haplotype GGC consisted of–238G,–308G and–857C). These most frequent haplotypes (GGC, GAC, GGT and AGC) accounted for nearly 93% of the total number of haplotypes, with frequencies of occurrence of 54%, 22%, 13% and 4%, respectively. The rare AAC haplotype was found in only one patient. Phenotype frequencies of these haplotypes were in accordance with those observed among healthy Caucasian controls from UK16 but some differences with other Caucasian populations cannot be excluded.
Univariate and multivariate analysis of factors influencing ACR50 response to adalimumab
We used binary logistic regression to study the probability of achieving the ACR50 response to ADA therapy at 12 weeks. Univariate regression to screen for variables (with P<0.15) to enter in the model in the second step revealed weight, body mass index, height, associated therapy with MTX, associated therapy with other DMARDs, presence of rheumatoid factor, TNF polymorphisms–857 and–238, TNF haplotype (–238G,–308G and–857C), number of swollen joints and HAQ score at baseline. These covariates were retained as candidate covariates to build a multivariate model.
In the multivariate analysis, we found the following variables associated with ACR50 response: associated therapy with MTX (OR, 1.68; 95% CI, 1.07 to 2.66; p = 0.026), presence of rheumatoid factor (OR, 1.91; 95% CI, 1.13 to 3.21; p = 0.015), body mass index (OR, 1.06; 95% CI, 1.01 to 1.11; p = 0.01), baseline HAQ score (OR, 0.55; 95% CI, 0.37 to 0.82; p = 0.003), and presence of TNF haplotypes other than the GGC homozygous haplotype (OR, 1.92; 95% CI, 1.21–3.02; p = 0.005).
Influence of the shared epitope and individual TNF genotypes on adalimumab response
The SE was not identified in uni- or multivariate analysis as a factor influencing ACR50 response to ADA therapy at week 12. In fact, we found no association between ACR50 response to ADA and SE copy number or carrier status (table 3) or any of the three TNF gene polymorphisms–238A/G,–308A/G and–857C/T genotypes (table 3) tested individually. However, we found a trend of an association between a lower response to ADA therapy at 12 weeks and–238GG,–308GG and–857CC genotypes (table 3).
Influence of the TNF haplotype on adalimumab response
In analysing the influence of TNF haplotype on response to ADA therapy, GGC haplotype carrier status showed significant results (table 4). In the first analysis, we discarded rare haplotype combinations (n = 27) that were represented less than 10 times: AGC/AAC (n = 1), GAC/AGC (n = 2), GAC/GAC (n = 9), GGT/AGC (n = 3), GGT/GAC (n = 9), GGT/GGT (n = 3). For subjects with the remaining four main haplotype combinations, those homozygous for the GGC haplotype (n = 190) had a significantly lower ACR50 response rate (34%) than subjects with the three other main combined haplotypes: GGC/GAC (47%; n = 78), GGC/GGT (53%; n = 45) and GGC/AGC (71%; n = 14) (p = 0.004) (table 4). This observation highly suggested a recessive effect of the GGC haplotype on response to treatment. In fact, the response rate for homozygous GGC haplotype carriers (34%; n = 190) was significantly lower than the response rate observed when pooling all other haplotype combinations, including the rare ones (50%; n = 164) (p = 0.003, after Bonferroni correction pa = 0.015) (table 4).
A similar but non-significant trend was observed for ACR20 and ACR70 responses to ADA therapy at week 12: the GGC/GGC group showed a 69% ACR20 response versus 76% for the other haplotype carriers (p = 0.18) and a 14% ACR70 response versus 18% for the other haplotype carriers (p = 0.5). The lack of significant difference regarding ACR20 and ACR70 responses can be explained by a lack of statistical power as responders and non-responders are unbalanced with these criteria. Conversely, ACR50 response at week 12 provides the best statistical power to demonstrate an effect with a distribution of responders and non-responders approaching 50% of the whole population.
As well as our finding of the homozygous effect of the GGC haplotype, we noted a “trans” effect of certain haplotypes present on the opposite chromosome. Alone, the “trans” association of the AGC haplotype with the GGC haplotype appeared to be associated with a better response to ADA therapy: 71% for the GGC/AGC combination (table 4) versus 40% for all the other haplotype combinations, including the rare ones. The difference between those groups was not significant after Bonferroni correction (p = 0.019; pa = 0.095). Nevertheless, few patients were AGT haplotype carriers (10 ACR50 responders and four ACR50 non-responders), so this analysis was therefore underpowered.
Baseline characteristics did not significantly differ between GGC homozygous patients and other haplotype carriers, especially in DAS28 RA disease-activity criteria (table 2).
As about half of our sample received MTX therapy with ADA, we further analysed the GGC homozygosity effect according to treatment: ADA plus MTX, ADA plus any other DMARD, and ADA alone. Surprisingly, the lower ACR50 response rate was present mainly in the group of patients receiving ADA plus MTX (n = 182): 38% ACR50 response versus 60% for the other haplotype carriers (p = 0.01) (fig 1). For the ADA plus other DMARD group (n = 98), we observed a similar, although not significant, trend: 32% ACR50 response versus 48% for the other haplotype carriers (p = 0.2) (fig 1). For the ADA-alone group (n = 108), the GGC homozygous haplotype had no significant effect (29% vs. 32% for the other haplotype carriers) (p = 0.94) (fig 1). In fine, MTX co administration with ADA seems to be more beneficial for patients not carrying the homozygous GGC haplotype.
In analysing the time-course evolution of ACR50 response among treatment groups, for patients receiving ADA plus MTX, the response differed between GGC haplotype homozygous and other haplotype carriers as soon as week 2 after treatment initiation and increased until week 12 (fig 2). For patients receiving ADA plus any other DMARD, the response showed a trend in favour of a difference between haplotype groups over time, which might have become significant with longer follow-up, which was not available in the ReAct protocol (fig 2). For patients receiving ADA alone, the response did not significantly differ between haplotype groups over time but follow-up may be not long enough to see such differences (fig 2).
Influence of the shared epitopes on the negative effect of GGC homozygosity on adalimumab response
The rate of ACR50 responders among the GGC haplotype carriers was independent of the number of SE copies (40% with two SE copies; 45% with one SE copy; 35% without SE copy). Accordingly, in the univariate regression analysis, the SE was not retained as candidate covariate due to a corresponding p value above the level of significance retained to build the multivariate model (p = 0.31). Thus, SE does not play a part in the lower response rate to ADA at week 12 observed in patients carrying the TNF homozygous GGC haplotype.
The TNF locus is located near HLA DRB1, so we analysed the extent to which the GGC haplotype was in LD with the SE. Although some alleles belonging to the SE were in LD with some TNF gene alleles, the overall LD between the GGC haplotype and the SE alleles was weak, with a Somers D′ equal to 0.24.
Our extended haplotype reconstructions resulted in 82 SE-TNF haplotype combinations. Careful examination of these haplotypes showed the GGC haplotype mainly associated with HLA DRB1*0101 and *0401 alleles of the SE, which is not surprising because GGC is the most frequent TNF haplotype and DRB1*0101 and *0401 are the most frequent HLA DRB1 alleles among Caucasians. Most of these extended haplotypes were associated with an approximate 40% ACR50 response (range 36–42%), except for the 0701-GGC haplotype (n = 29; 24% ACR50 response) and the 0405-GGC haplotype (n = 23; 30% ACR50 response).
DISCUSSION
This pharmacogenetic study has investigated the effect of TNF gene polymorphisms on response to ADA in patients with RA. Although the presence of individual TNF SNPs was not associated with a specific pattern of response to ADA, haplotype analysis provided convincing data suggesting that the ancestral haplotype of the TNF promoter, GGC, the most frequent among Caucasians,16 was associated with a lower response to ADA when present in the homozygous form.
The effect of the SE on response to TNF blockers in RA led to contradictory results in the literature (table 1). One study reported increased response to etanercept in patients with two copies of the SE (OR, 4.3; 95% CI, 1.8 to 10.3) (n = 151)2 and a second did not replicate such results (n = 198).11 Regardless, our results among 322 patients do not show any effect of SE copy number on response to ADA.
Pharmacogenetic studies of the influence of TNF genes on response to TNF-blocker treatment in RA have given discordant results (table 1). Two studies reported an association of the TNF–308GG genotype with better response to infliximab3 or etanercept8 but concerned a low number of patients: 59 and 22, respectively. A recent study involving 88 patients with various rheumatic diseases of whom 54 RA suggested that the TNF–308GG genotype was associated with a better response to TNF blockers (ADA, infliximab or etanercept) at week 24.12 Studies involving larger sample sizes (n = 151 with etanercept (two) and n = 198 with infliximab (11)) did not confirm these results, nor did ours, with 374 patients genotyped for TNF–308A>G polymorphism and receiving ADA therapy.
Many explanations may account for such discrepancies. In fact, parameters such as the response criteria chosen as primary endpoint (DAS28 decrease, ACR20 or ACR50 response) and the delay of efficacy for the primary endpoint (range from 6 to 24 months from baseline) differed among studies.2 4 7 8 12 Otherwise, the small number (less than 80 patients) of subjects in the positive studies could be another explanation, as the three negative studies each included more than 150 patients.
Interestingly, few of these studies focused on TNF haplotypes as a factor that could influence response to TNF blockers. One study analysed extended haplotypes comprising HLA DRB1 alleles, TNF +488/–238/–308 polymorphisms and lymphotoxin α 720/365/249 polymorphisms. This study involved a US Caucasian population and reported an association of two extended haplotypes with better response to etanercept in RA.2 Unfortunately, the authors did not analyse the proper effect of TNF polymorphism haplotypes on response to etanercept, which preclude any comparison with our results. The functional effect of TNF promoter haplotypes has not been reported to date. In fact, most of the functional studies of the TNF promoter focused on a separate analysis of SNPs and mainly on the–308A/G polymorphism. Some published studies reported high TNF production, among various cell types and stimulation conditions, in the A allele carriers.22–25 But these results were not confirmed by others26–28 and are still a matter of debate in the scientific community. We found that the TNF (–238G,–308G and–857C) homozygous haplotype was associated with a lower response to ADA therapy than were other haplotypes. The effect of other functional variants in high LD with the three TNF promoter’s polymorphisms studied cannot be excluded. Functional studies have to be realised to correlate this haplotype with the level of TNF production. Interestingly, recent studies suggest that subjects with high response to TNF blockers could also have high TNF-α bioactivity 29 or high TNF-α synovial levels at baseline.30
The effect of the TNF (–238G,–308G and–857C) homozygous haplotype on ADA response was restricted to patients also receiving MTX, with a trend to association with ADA plus any other DMARD treatment, and was absent in patients receiving ADA alone. The reason for such observation remains unclear. A specific effect of TNF promoter polymorphisms on response to MTX itself cannot be excluded. But in the study from Criswell et al,2 no effect of individuals TNF SNPs was associated with response in the group of patients treated with MTX alone. Moreover, the trend to a similar effect in patients treated with ADA associated with other DMARDs argues against a specific effect of MTX. Thus, the absence of effect of the GGC haplotype homozygous carriers in patients treated with ADA alone remains puzzling. The rate of high response (ACR50) could be delayed in patients with ADA monotherapy, which could explain why there is no difference at week 12 with or without the GGC haplotype. However, response data are available after week 12 in only a minority of patients of the ReAct study, which was a 12-week open label study and this hypothesis cannot be assessed.
These results, even if obtained on a large cohort of patients, are preliminary and have to be replicated. They could have some importance to better understand the mechanism of action of ADA and the better efficacy of ADA associated with MTX or with other DMARDs observed in the ReAct study. However, the difference in ACR50 response being from 34 to 50%, they will not lead to exclude patients from the drug. Moreover, MTX co-administration to ADA seems to most benefit patients not carrying the homozygous GGC haplotype.
In conclusion, this pharmacogenetic study is remarkable because of the size of the population studied as well as the quality of the clinical data recorded within the ReAct study. It provides preliminary data indicating that a single TNF locus haplotype (–238G/–308G/–857C), when carried on both chromosomes, is associated with a lower response to ADA in patients with RA, than patients with other haplotypes, mainly in those treated with a combination of ADA and MTX. The mechanism of action of this genetic predisposition needs further investigations.
Acknowledgments
We thank Axelle Belaube and Laura Contreras, Abbott France, for help in conduct of this pharmacogenomic study. We thank the following collaborators for their participation in DNA collection: Pr Pierre Bourgeois (Hôpital La Pitié-Salpétrière, Paris), Pr Alain Cantagrel (Hôpital Rangueil, Toulouse), Pr Bernard Combe (Hôpital Lapeyronie, Montpellier), Pr Thierry Schaeverbeke (Hôpital Pellegrin, Bordeaux), Pr Maxime Dougados (Hôpital Cochin, Paris), Dr Jean-Paul Eschard (Hôpital Sébastopol, Reims), Pr Liana Euller-Ziegler (Hôpital de L’Archet 1, Nice), Pr Patrice Fardellone (Hôpital Nord, Amiens), Pr René-Marc Flipo (Hôpital B Roger Salengro, Lille), Pr Pascal Hilliquin (Centre Hospitalier Sud Francilien, Corbeil Essonnes), Pr Robert Juvin (Hôpital Albert Michallon, La Tronche), Pr André Kahan (Hôpital Cochin, Paris), Pr Pierre Lafforgue (Hôpital de la Conception, Marseille), Pr Christian Marcelli (Hôpital de la Côte de Nacre, Caen), Pr Isabelle Chary-Valckenaere (Hôpital du Brabois, Vandoeuvre les Nancy), Dr Xavier Puechal (Centre hospitalier, Le Mans), Pr Jean-Michel Ristori (Hôpital Gabriel Montpied, Clermont-Ferrand), Pr Jean Sibilia (Hôpital de Hautepierre, Strasbourg), Pr Philippe Goupille (Hôpital Trousseau, Tours), Pr Aleth Perdriger (Hôpital Sud Fontenoy, Rennes), Pr Eric Houvenagel (Centre hospitalier Saint-Philibert, Lomme), Pr Christian Alexandre (Hopital de Bellevue, Saint-Etienne) and Dr Alain Heraud (Centre hospitalier Robert Boulin, Libourne).
REFERENCES
Footnotes
Funding: This work was promoted by the Club Rhumatismes et Inflammation with a grant from Abbott France.
Competing interests: None.