Article Text

Genetic variation in the nuclear factor κB pathway in relation to susceptibility to rheumatoid arthritis
  1. R Dieguez-Gonzalez1,
  2. S Akar1,
  3. M Calaza1,
  4. E Perez-Pampin1,
  5. J Costas2,
  6. M Torres2,
  7. J L Vicario3,
  8. M L Velloso4,
  9. F Navarro5,
  10. J Narvaez6,
  11. B Joven7,
  12. G Herrero-Beaumont8,
  13. I Gonzalez-Alvaro9,
  14. B Fernandez-Gutierrez10,
  15. A R de la Serna11,
  16. L Carreño12,
  17. J Lopez-Longo12,
  18. R Caliz13,
  19. M D Collado-Escobar13,
  20. F J Blanco14,
  21. C Fernandez-Lopez14,
  22. A Balsa15,
  23. D Pascual-Salcedo16,
  24. J J Gomez-Reino1,
  25. A Gonzalez1
  1. 1
    Laboratorio de Investigacion 2 and Rheumatology Unit, Hospital Clinico Universitario de Santiago, Santiago de Compostela, Spain
  2. 2
    National Genotyping Center, Hospital Clinico Universitario de Santiago, Santiago de Compostela, Spain
  3. 3
    Regional Transfusion Center, Madrid, Spain
  4. 4
    Rheumatology Unit, Hospital Universitario de Valme. Sevilla, Spain
  5. 5
    Rheumatology Unit, Hospital Universitario Virgen Macarena, Sevilla, Spain
  6. 6
    Rheumatology Unit, Hospital Universitario de Bellvitge, Barcelona, Spain
  7. 7
    Rheumatology Unit, Hospital 12 de Octubre, Madrid, Spain
  8. 8
    Rheumatology Unit, Fundacion Jimenez Diaz, Madrid, Spain
  9. 9
    Rheumatology Unit, Hospital Universitario de la Princesa, Madrid, Spain
  10. 10
    Rheumatology Unit, Hospital Clinico San Carlos, Madrid, Spain
  11. 11
    Rheumatology Unit, Hospital Santa Creu e San Pau, Barcelona, Spain
  12. 12
    Rheumatology Unit, Hospital Universitario Gregorio Marañon, Madrid, Spain
  13. 13
    Rheumatology Unit, Hospital Universitario Virgen de las Nieves, Granada, Spain
  14. 14
    Laboratorio de Investigación Osteoarticular y del Envejecimiento, Servicio de Reumatología, Hospital Universitario Juan Canalejo, A Coruña, Spain
  15. 15
    Rheumatology Unit, Hospital La Paz, Madrid, Spain
  16. 16
    Immunology, Hospital La Paz, Madrid, Spain
  1. A Gonzalez, Laboratorio de Investigacion 2, Hospital Clinico Universitario de Santiago, Travesia de Choupana sn., 15706-Santiago de Compostela, Spain; Antonio.Gonzalez.Martinez.Pedrayo{at}


Objective: To examine genetic association between rheumatoid arthritis (RA) and known polymorphisms in core genes of the nuclear factor (NF)κB pathway, the major intracellular pathway in RA pathogenesis.

Methods: Discovery and replication sample sets of Spanish patients with RA and controls were studied. A total of 181 single nucleotide polymorphisms (SNPs) uniformly spaced along the genomic sequences of 17 core genes of the NFκB pathway (REL, RELA, RELB, NFKB1, NFKB2, NFKBIA, NFKBIB, NFKBIE, IKBKA, IKBKB, IKBKE, IKBKAP, KBRAS1, KBRAS2, MAP3K1, MAP3K14, TAX1BP1) were studied by mass spectrometry analysis complemented with 5′-nuclease fluorescence assays in the discovery set, 458 patients with RA and 657 controls. SNPs showing nominal significant differences were further investigated in the replication set of 1189 patients with RA and 1092 controls.

Results: No clear reproducible association was found, although 12 SNPs in IKBKB, IKBKE and REL genes showed significant association in the discovery set. Interestingly, two of the SNPs in the IKBKE gene, weakly associated in the discovery phase, showed a trend to significant association in the replication phase. Pooling both sample sets together, the association with these two SNPs was significant.

Conclusion: We did not find any major effect among the explored members of the NFκB pathway in RA susceptibility. However, it is possible that variation in the IKBKE gene could have a small effect that requires replication in additional studies.

Statistics from

The complex aetiology of rheumatoid arthritis (RA) includes a genetic component that accounts for about 50% of disease liability1 The main genetic factor is known to reside in specific human leukocyte antigen (HLA)-DRB1 alleles, collectively called the shared epitope (SE). A second major genetic factor is a non-synonymous polymorphism in the lymphocyte thyrosine phosphatase coded by PTPN22.2 Other genetic factors, in STAT4,3 the TRAF1/C5 4 and the OLIG3/TNFAIP3 5 6 loci have been discovered in recent months.

Nuclear factor (NF)κB plays a key role in a wide range of RA-related mechanisms and is downregulated by several drugs used in RA treatments.79 NFκB activates proinflammatory cytokines, the matrix metalloproteinases, the adhesion molecule intercellular adhesion molecule 1 (ICAM1) that favours recruitment of lymphocytes, and cyclo-oxygenase 2 that is responsible for the formation of prostanoids. Surprisingly, no systematic investigation of the role of genetic variation in the NFκB pathway has been performed. Only a couple of polymorphisms in NFKB1 and NFKBIA have been ever explored.1012 Herein, we have attempted to fill this hole by studying the NFκB core players.

There are two major signalling pathways that activate NFκB.1315 The canonical pathway is triggered by proinflammatory cytokines and pathogen-associated molecular patterns. These signals cause activation of the inhibitor of κB (IκB) kinase (IKK; note that different symbols are used for proteins and the genes encoding them) complex, which includes the IKKα and IKKβ catalytic subunits and the IKKγ regulatory subunit (encoded by the IKBKA, IKBKB and IKBKG genes, respectively). The activated complex phosphorylates the inhibitors of NFκB or IκBs. These proteins, IκBα, IκBβ, IκBϵ and IκBγ (corresponding to the NFKBIA, NFKBIB, NFKBIE and a fragment of the NFKB1 genes, respectively), in their unphosphorylated form bind to NFκB dimers and retain them in the cytoplasm. After phosphorylation, the IκBs are polyubiquitinated and degraded. In this way, NFκB dimers become free to move to the nucleus where they bind the DNA promoters of their regulated genes. The NFκB dimers are formed by any of the five NFκB proteins: p50/105, p52/100, RelA, RelB and c-Rel (encoded by NFKB1, NFKB2, RELA, RELB and REL genes, respectively). Phosphorylation and degradation of the IκBs is the critical step in the regulation of this pathway. Two proteins, IκB-interacting Ras-like protein (KBRAS)1 and KBRAS2, bind to IκBα and IκBβ delaying their degradation and contributing to inhibit NFκB induction. The alternative pathway is activated by engagement of a specific subset of receptors from the tumour necrosis factor (TNF) receptor superfamily. This pathway is absolutely dependent on the kinases NFκB-inducing kinase (NIK; also known as MAP3K14) and IKKα and independent of IKKβ and IKKγ. The two kinases are assembled into an active complex trough binding to separate domains of inhibitor of κ light polypeptide gene enhancer in B cells, kinase complex-associated protein (IKBKAP). The target of phosphorylation by IKKα in this pathway is p100. It is possible that other minor pathways lead to NFκB activation. For example, T cell receptor signals activate NFκB by a pathway dependent on IKKϵ (encoded by IKBKE). Also there is crosstalk with other signalling pathways induced by signals in Toll-like receptors involving the kinase mitogen-activated protein kinase kinase kinase 1 (MAP3K1) that phosphorylates members of the IKK complex.



DNA samples were obtained from patients with RA and healthy controls of Spanish ancestry in 13 hospitals from Spain (see supplementary material). The study was divided into a first discovery phase, 458 patients and 657 controls, and a second phase of replication with 1189 patients and 1092 controls. Patients with RA had an established disease and were classified according the 1987 American College of Rheumatology (ACR) criteria.16 Clinical data are provided in table 1. Patients in the discovery phase were from a single centre and were actively recruited trying to obtain samples for each patient available. Patients in the replication phase were recruited consecutively as they attended the participating rheumatology units. It is likely that this difference in recruiting has caused the differences in clinical characteristics. Patients with longer follow-up or with seronegative RA were more prevalent in the discovery phase (table 1). Controls were obtained in three ways in different centres: from the general population (all the controls in the discovery phase and in some centres from the replication phase), blood donors and spouses of patients. The Ethical Committee for Clinical Research of Galicia approved this study and all participants gave their written informed consent.

Table 1 Clinical data of patients with rheumatoid arthritis (RA) in the discovery and replication studies

Single nucleotide polymorphism (SNP) genotyping

A total of 181 SNPs were selected to cover genetic variation in 17 genes of the NFκB pathway (table 2 and supplementary material). The SNPs were selected to be evenly spaced and to have a minor allele frequency higher than 10% in the coding sequences and their neighbouring 10 Kb upstream and downstream. Genotype determination was performed with the MassARRAY SNP genotyping system (Sequenom, San Diego, California, USA) as described previously.17

Table 2 List of the 17 genes that have been screened with their genomic position, size, number of single nucleotide polymorphisms (SNPs) selected in each of them and the density obtained expressed as mean distance between SNPs in Kb (20 Kb of flanking sequence were included for each gene)

In the replication phase, we genotyped 11 SNPs in three genes with the same technology but different assays (new multiplexes), with three SNPs failing. These three SNPs were genotyped with TaqMan SNP genotyping Assays (Applied Biosystems, Foster City, California, USA).

Statistical analysis

Analysis of results relied on the Haploview and Statistica V. 7.0 (StatSoft, Tulsa, Oklahoma, USA), programs. Hardy–Weinberg equilibrium (HWE) was tested in control samples with a p value threshold of 0.01. χ2 Association tests were performed to compare allele frequencies in 2×2 contingency tables. Multivariate logistic regression analysis was used to evaluate the effect of each associated SNP in a gene conditional on the remaining. Likelihood ratio tests for the additive, dominant and recessive genetic models were obtained relative to the codominant model. Likelihoods for the fit of each model were calculated with univariate logistic regression. Allele frequencies in the two phases were considered together with the Mantel–Haenszel test to account for differences between the two sample collections, discovery and replication. Homogeneity of effect size across sample sets was assessed with the Breslow–Day test. Cocaphase software18 was used to infer haplotypes by the expectation-maximisation algorithm and to compare haplotype frequency distribution by the homogeneity likelihood ratio test.


Discovery phase

We selected evenly spaced SNPs across the 17 core NFκB genes (table 2 and supplementary material) and their flanking sequences (10 Kb in each direction). The 181 selected SNPs were examined in the discovery phase. They provided a mean coverage of one SNP every 5.5 Kb that is comparable to the coverage of the current whole genome association (WGA) commercial panels. A total of 32 SNPs were discarded because of diverse problems. The remaining 149 SNPs were successfully genotyped in 90.9% of the samples. The discovery set of samples was formed by 458 patients with RA and 657 controls (for clinical data, see table 1). Women accounted for 76.6% of the patients with RA and 52.1% of the controls.

Among the 149 validated SNPs there was nominal evidence of significant association (p<0.05) for 12 SNPs in 3 genes (table 3).

Table 3 Single nucleotide polymorphisms (SNPs) showing a nominal significant difference in allele frequencies between patients with rheumatoid arthritis (RA) and controls from the discovery set

The strength of association was more marked in the IKBKB gene where five of the seven studied SNPs showed p values <0.05. Two SNPs, rs12676482 and rs17875732, showed minor allele frequencies in patients with RA that were half the frequencies of the controls (odds ratios (ORs) of 0.46 and 0.52, respectively; p<0.001). Association in this gene could be due to rs17875732 as suggested by conditional logistic regression analysis (see supplementary material). Of the 22 validated IKBKE SNPs, 4 showed significant allelic differences in the discovery phase. They were towards the 3′ side of the gene in a region of 17.7 Kb. Linkage disequilibrium (LD), frequency data and conditional logistic regression analysis suggested two independent genetic factors in this gene (see supplementary material). Association in the REL gene was weak and included three of the five validated SNPs (table 3). This could be accounted for only by the rs6545835 SNP (see supplementary material).

In addition to the 12 SNPs with nominal significant differences, there were 8 other SNPs showing allelic differences with p values <0.1 (see supplementary material). Genotype analysis showed that the additive model fitted data as well as, or better than, other genetic models. Results with the additive model were almost identical to the observed with allele frequencies (data not shown). Haplotype analysis of each of the 17 genes did not show any new significant association.

Replication phase

Of the 12 SNPs found at the p<0.05 level in the discovery phase (rs842647 in REL was not tested because it was well represented by rs3732179, r2 = 0.97), 11 were genotyped in a replication phase with a 97.4% call rate. All of them showed genotype frequencies that were in accordance with HWE. The replication set of samples included 1189 patients with RA and 1092 controls. In both groups, females were predominating in similar proportion, 73.94% and 70.67%, respectively. Allelic frequencies were compared and none of the differences that have been found in the discovery phase were confirmed (table 4). Genotype frequencies gave similar results with the additive genetic model, which fitted data better or equally well than other genetic models (data not shown).

Table 4 Allele frequencies of the 11 single nucleotide polymorphisms (SNPs) analysed in the replication phase of the study

The largest differences were observed in the two weakly correlated IKBKE SNPs that have larger minor allele frequency (MAF), rs2151222 and rs3748022 (p = 0.07 for each of them). All the others were very similar in patients with RA and controls of this set of samples. The sample size of the replication set was enough to allow for confirmation of the discovery phase results. It provided more than 95% power for an effect similar to the four IKBKB SNPs, the two IKBKE SNPs with low MAF, rs17433909 and rs17434047, and the REL SNP able to account for association, rs6545835.

Trying to find an explanation for the lack of replication, we compared, separately, the allelic frequencies in controls and in patients with RA between the discovery and the replication phases (see supplementary material). There was only one SNP that was slightly different between the two sets of controls in spite of differences in their recruitment in the two phases (population vs a mixture of three different approaches). In contrast, 7 of the 11 SNPs showed different allele frequencies in patients with RA from the 2 phases. The five IKBKB SNPs and the two IKBKE SNPs with low MAF were significantly less frequent in patients with RA from the discovery phase than in patients from the replication phase. Although there were differences in some clinical characteristics between patients with RA in the two phases of the study (table 1) that were indicative of a more severe disease in the replication phase, there were not any significant correlation between the available clinical data and these SNPs. Specifically, there were no differences in the allele frequencies between patients with rheumatoid factor (RF), anti-cyclic citrullinated peptides (CCPs) (only available in the discovery phase) or rheumatoid nodules and those lacking these RA features; nor differences in the strength of association (evaluated as ORs) that could explain the results observed when the patients were stratified by these clinical features; nor significant changes in the differences between genotype frequencies in patients with RA from the two phases of the study when data were adjusted for time of follow-up as covariant.

We also considered the two phases jointly with a Mantel–Haenszel analysis (table 5).

Table 5 Pooled analysis of the allelic frequencies in the discovery and replication phases

This analysis showed that 9 of the 11 SNPs were still associated with RA with PM-H values <0.05, but only 3 of them showed p values that were lower than the obtained in the discovery phase: 2 in IKBKE (rs2151222 and rs3748022) and 1 in REL (rs3732179). There was significant heterogeneity of effects between the two phases in six of the SNPs by the Breslow–Day homogeneity test (PB–D in table 5), further showing that most results were not comparable between the two sets of samples. Results were very similar if the pooling of data was performed considering towns of origin of the samples (data not shown). No significant influence of gender was present in any of the analyses.


We have taken in this study a variant of the candidate gene approach by exploring genetic variation in a functional unit in RA pathogenesis. The selected NFκB signalling pathway has an important regulatory role in RA.79 It has two clearly defined branches and many alternative players at different levels.13 15 Therefore, it seemed interesting to investigate if genetic association with RA susceptibility will identify any of them as critical. The 17 selected genes included the core elements and other members of the pathway related to either minor routes of activation, as IKBKE and MAP3K1, or critical regulatory steps, as KBRAS1 and KBRAS2. Density of SNP coverage was slightly larger than the available in the WGA panels but was not complete at r2>0.8.

The study was conducted in two phases because replication of the findings is critical to differentiate between true associations and random frequency changes. However, it should be noted that alleles with modest effects could have been missed in the discovery phase. More specifically, for an OR = 1.3 and minor allele frequency of 0.2, the power of the discovery phase was 72% and power of the replication set was >95%.

Results of the first phase were promising with SNPs in three genes showing nominal evidence of association and some of them, in IKBKB and IKBKE, a relatively strong effect. However, the evidence for association would be lower if an adjustment was performed for multiple testing. None of these associations was replicated in the second phase of the study. Allele frequency differences between the two phases were present only among the patients with RA, but we did not find correlation with available clinical data even with the clinical characteristics that were different between the two phases of the study. Therefore, the lack of replication of the original findings remains unexplained. Only two SNPs in IKBKE showed a trend to association in the replication sample set. These two SNPs, rs2151222 and rs3748022, also showed an increased statistical significance when the two sample sets were considered together. Interestingly, they are correlated and thus the possible association would be to a single genetic factor. Another possibility is that they are related with an interleukin cluster, 270 Kb telomeric to IKBKE. This cluster includes IL10, a gene with several regulatory polymorphisms that influence RA susceptibility and severity according to some studies, 1921 but not according to others.22 23 If a direct IKBKE association is confirmed it will add new perspectives into RA pathology because IKKϵ is not a component of the IKK complex and is not involved in the best-known NFκB activation pathways.24 IKKϵ activates the interferon regulatory factor (IRF)3 and IRF7 transcription factors of the innate immune response, and the NFκB pathway in T cells, by directly phosphorylating RelA.25 Also of relevance for RA pathology could be that IKKϵ is expressed in fibroblast-like synoviocytes and in the synovial intimal lining where it participates in the induction of matrix metalloproteinases through phosphorylation of c-Jun.26

It is also possible that the differences in IKBKB SNPs in the discovery set were related with the nearby PLAT gene, coding for the tissue type plasminogen activator, that is only 63 Kb telomeric to IKBKB and that we have previously found associated with RA susceptibility.27 Nevertheless, we have not detected significant LD between the regulatory SNP in PLAT and the SNPs of IKBKB (data not shown).

Results from the first large WGA in RA susceptibility have been published very recently.28 This study can provide important independent confirmation for our results because it is very powerful and exhaustive, including 3000 controls and 2000 patients with RA of British descent genotyped at 500 000 SNPs. None of the SNPs that are highlighted in the WGA publication as different in patients with RA and controls is near any of the NFκB genes explored by us. However, a deeper analysis is possible as there are many other differences in the WGA study that are not significant at the genome level but they can be used for confirmation of independent studies. For example, the WGA includes the same number of SNPs in the IKBKB locus as our study. Only one shows some difference between cases and controls (rs5029748, p = 0.01). This SNP was also included in our discovery phase but we did not find any difference. By contrast, two SNPs with significant differences in our discovery phase (rs12676482 and rs17875732) are represented in the WGA by highly correlated SNPs (r2 = 1) that are not different between patients with RA and controls. In the same way, there are two SNPs in the WGA study informative about the REL SNPs we have analysed. None of them are associated with RA. Unfortunately, it is impossible to obtain this kind of information for IKBKE because the WGA panel does not include any SNPs in the 40 Kb surrounding the IKBKE gene. In addition, flanking SNPs do not show significant correlation with IKBKE SNPs. These comparisons show a way in which WGA studies will contribute to progress in the genetic investigation of RA.

In conclusion, our study has not found important genetic factors for RA susceptibility in core members of the NFκB signalling pathway. It should be noted, however, that SNP coverage was not complete and that power of the discovery phase did not allow excluding modest effects. Additionally, it is possible that a more complex or restricted involvement, because of interaction between genetic factors or association limited to a subgroup of patients, has eluded our detection. Our initial findings of RA association with IKBKB and REL SNPs were most likely due to random fluctuations in allelic frequencies that can be expected in any study where multiple SNPs are analysed. Lack of replication in the second set of samples and of confirmatory evidence in the published WGA study support this conclusion. Nevertheless, variation in IKBKE or nearby sequences could have a role in RA susceptibility although we were unable to obtain clear confirmatory evidence. This effect may be small, in an OR range of 1.1 to 1.2, and this will make difficult future attempts at replication.


We thank Cristina Fernandez-Lopez for her excellent technical assistance.


Supplementary materials

  • Web only appendix 68:4;579


  • Competing interests: None.

  • Funding: This project was supported by grant PI04/1513 from the Instituto de Salud Carlos III (Spain) with participation of funds from FEDER (European Union). SA was the recipient of a Research Fellowship of the Fundation Articulum.

  • Ethics approval: The Ethical Committee for Clinical Research of Galicia approved this study and all participants gave their written informed consent.

  • ▸ Additional data (supplementary tables 1–6) are published online only at

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.