Linkage and association studies of single-nucleotide polymorphism–tagged tumor necrosis factor haplotypes in juvenile oligoarthritis


  • Eleftheria Zeggini,

    Corresponding author
    1. Arthritis Research Campaign Epidemiology Unit, University of Manchester, Manchester, UK
    • ARC/EU, Stopford Building, Oxford Road, Manchester M13 9PT, UK
    Search for more papers by this author
    • Eleftheria Zeggini's work was supported by the Arthritis Research Campaign.

  • Wendy Thomson,

    1. Arthritis Research Campaign Epidemiology Unit, University of Manchester, Manchester, UK
    Search for more papers by this author
  • Dominic Kwiatkowski,

    1. Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, UK
    Search for more papers by this author
  • Anna Richardson,

    1. Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, UK
    Search for more papers by this author
  • William Ollier,

    1. Arthritis Research Campaign Epidemiology Unit, University of Manchester, Manchester, UK
    Search for more papers by this author
  • Rachelle Donn,

    1. Arthritis Research Campaign Epidemiology Unit, University of Manchester, Manchester, UK
    Search for more papers by this author
  • The British Paediatric Rheumatology Study Group

    Search for more papers by this author
    • Contributors to The British Paediatric Rheumatology Study Group are as follows: M. Abinun, MD, M. Becker, MD, A. Bell, MD, A. Craft, PhD, E. Crawley, MD, J. David, MD, H. Foster, MD, J. Gardener-Medwin, MD, J. Griffin, MD, A. Hall, MD, M. Hall, MD, A. Herrick, MD, P. Hollingworth, MD, L. Holt, MD, S. Jones, MD, G. Pountain, MD, C. Ryder, MD, T. Southwood, PhD, I. Stewart, MD, P. Woo, PhD, S. Wyatt, MD, H. Venning, MD.



The presence of increased levels of tumor necrosis factor (TNF) in serum and synovial fluid of patients and the encouraging outcome of anti-TNF therapy have implicated TNFα in the etiopathogenesis of juvenile oligoarthritis. Although the locus is polymorphic, no study has investigated all TNF single-nucleotide polymorphisms (SNPs) with respect to disease. The aim of this study was to examine the association of multiple TNF SNPs with juvenile oligoarthritis and to construct and analyze SNP-tagged TNF haplotypes.


A total of 144 simplex families consisting of parent and affected child, as well as 88 healthy, unrelated control subjects were available for study. In these individuals, 9 polymorphic positions of TNF were typed by a high-throughput genotyping method based on the SNaPshot assay. The chi-square and extended transmission disequilibrium tests were used to test for association and linkage, respectively. Odds ratios (ORs) with 95% confidence intervals (95% CIs) were also calculated. Haplotype-tagging SNPs (htSNPs) for the locus were identified by ordering the haplotypes according to their frequencies.


The study detected association of several TNF SNPs and established linkage of the locus to juvenile oligoarthritis. The most significant association observed was between the intronic +851 TNF SNP and the persistent oligoarthritis subgroup (OR 3.86, 95% CI 1.6–9.2). Haplotype data mining showed that only 4 of the 9 SNPs need to be typed in order to capture the most frequent TNF haplotypes.


The TNF locus is linked and associated with juvenile oligoarthritis. Information on the htSNPs can be useful in genetic studies of diseases in which TNF may be of relevance.

Juvenile idiopathic arthritis (JIA) is a family of childhood arthropathies characterized by chronic synovial inflammation. The commonest JIA subgroup is oligoarthritis (previously termed pauciarticular disease). The most recent JIA classification criteria proposed by the International League of Associations for Rheumatology (1) divides juvenile oligoarthritis into persistent and extended disease. Persistent oligoarthritis manifests as persistent inflammation of 4 or fewer joints throughout the disease course. Extended oligoarthritis affects 4 or fewer joints during the first 6 months of disease and then affects a cumulative total of 5 or more joints.

Juvenile oligoarthritis is a complex disease of unknown etiology (2). The immunopathogenic mechanisms underlying it have been the subject of intensive study (3, 4). Genetic investigations have concentrated on the relationship between the HLA loci and juvenile oligoarthritis and have consistently shown association of alleles HLA–A*02, HLA–DRB1*08, HLA–DRB1*11, and HLA–DPB1*0201 with disease (5, 6). Linkage of the major histocompatibility complex (MHC) region on chromosome 6p21.3 has also been demonstrated (6). The tumor necrosis factor α (TNFα) gene resides in this chromosomal region and is situated between HLA–A and HLA–DRB1. It codes for a proinflammatory cytokine with an inherent role in programmed cell death. Local expression of TNF can modulate autoimmune responses, but production level imbalances can lead to pathogenic inflammation.

The fact that administration of anti-TNF treatment to patients with juvenile arthritis has yielded encouraging results may indicate a role for TNF in the pathogenesis of the disease (7, 8). Several studies have evaluated the levels of TNF in the sera and synovial fluid of patients with juvenile oligoarthritis in order to investigate the role of TNF as a mediator of chronic inflammation in this disease. The consensus observation has been a pronounced elevation of TNF levels, in both the sera and the synovial fluid of children with oligoarthritis (9–11).

Individual levels of production of TNF may be dictated by genetic variation. The TNF locus has been screened for sequence diversity, and at least 12 single-nucleotide polymorphisms (SNPs) have been identified within the promoter, exonic, intronic, and 3′-untranslated regions of the gene (12). In an attempt to determine the role of specific SNPs in transcriptional regulation, functional studies have focused mainly on the promoter region (13). However, conflicting findings of distinct alleles that affect TNF expression levels have been reported (14–17). Conversely, associations of TNF SNPs with disease are being unraveled and replicated. Indeed, several of the TNF SNPs have been implicated in diverse disorders, such as malaria, rheumatoid arthritis, systemic lupus erythematosus, and ankylosing spondylitis (18–21).

However, little research investigating the association of TNF polymorphisms with juvenile oligoarthritis has been performed. Epplen et al (22) observed no difference in the frequencies of the −308 and −238 TNF SNP alleles between German patients with juvenile oligoarthritis and controls. Date et al (23) showed an association of TNF promoter SNPs −1031, −863, and −857 with systemic JIA, but not with oligoarthritis, in a Japanese population. Finally, Ozen et al (24) found no association of the −308 and −238 polymorphisms in Turkish and Czech patients with juvenile oligoarthritis. To date, no study has investigated the association of TNF SNPs with juvenile oligoarthritis in UK Caucasians. Furthermore, no study has previously examined all known TNF SNPs or investigated linkage of the TNF locus to juvenile oligoarthritis.

The importance of family-based linkage studies becomes evident when considering the deleterious effects of population stratification in association studies (25). Moreover, when studying SNPs within a disease candidate gene such as TNF, it is useful to examine haplotypes, because SNPs may exert their effects in concert with each other. The value of constructing TNF SNP haplotypes in a disease linked to the MHC, such as juvenile oligoarthritis, is also exemplified by the need to elucidate the contribution of different susceptibility genes within the region. Finally, the availability of TNF haplotypes allows identification of haplotype-tagging SNPs (htSNPs; the minimum number of SNPs that must be typed in order to capture the most frequently occurring haplotypes) for the gene (26).

A principal aim of this study was, therefore, to examine all known TNF SNPs in juvenile oligoarthritis and to perform association and linkage studies. An additional objective was to determine linkage disequilibrium (LD) patterns between the TNF SNPs, to construct haplotypes, and to dissect them with respect to disease.



DNA samples obtained from 144 UK Caucasian simplex families (consisting of 1 affected child with juvenile oligoarthritis and healthy parents) were available through the British Paediatric Rheumatology Group National Repository for JIA, held in the Arthritis Research Campaign Epidemiology Unit in Manchester. These individuals had previously been typed for HLA–A and HLA–DRB1 (27). In 45 of the families (24 in the persistent oligoarthritis subgroup and 21 in the extended oligoarthritis subgroup), 1 parental DNA sample was available; in the other 99 families, DNA samples from both parents were available. Ninety-two of the children had persistent oligoarthritis, and 52 had extended disease. Ninety-nine of the patients were female, and 45 were male. For the purposes of the association study, the affected children from these families were used as cases, and 88 healthy, unrelated UK Caucasian blood donors were used as controls.


A genotyping assay based on the SNaPshot (PE Applied Biosystems, Warrington, UK) ddNTP primer extension method was designed for 14 TNF SNPs. This assay involves an amplification step to derive a polymerase chain reaction (PCR) product containing the polymorphic residue, followed by an allele-specific extension step. The region was amplified using primers as described in Table 1. The assay was multiplexed at the PCR level by designing amplimers that contained >1 polymorphic position. The allele-specific extension step was performed using the primers listed in Table 2. The location of each SNP within the TNF gene and the nucleotide substitution it represents are also described in Table 2.

Table 1. Length of tumor necrosis factor PCR fragments, and sequences of the primers*
SNPPCR product length, bpPrimer
  • *

    Information on the +472 single-nucleotide polymorphism (SNP) was obtained by personal communication with Anna Richardson (Oxford, UK). Information on the +1253 SNP was obtained from the NCBI dbSNP cluster (ID rs1800620). PCR = polymerase chain reaction.

Table 2. Location of single-nucleotide polymorphisms (SNPs) in the tumor necrosis factor gene, and probe sequences used for genotyping*
SNPLocationNucleotide substitutionInternal probe sequence
  • *

    The assay was not affected by which allele was present at the SNP within the probe, in cases of internal probe sequences that spanned other SNPs (e.g., the −863/−857, −244/−238, and +472/+489 SNP pairs).


The design of primers and probes as well as the positioning of the SNPs relative to the transcriptional start site of TNF were based on the GenBank sequence (accession no. M16441) (28). Probes with lengths differing by at least 4 basepairs could be multiplexed at the SNaPshot genotyping level. To amplify the TNF fragments, 10-μl reactions were set up by mixing 10 ng of DNA with 10 pmoles of each primer (except for the −1073/−1031 and −863/−857 PCRs, which used 5 pmoles of each primer), 0.2 nmoles of each dNTP, 1× NH4 buffer, MgCl2 at 2mM, and 0.5 units of Taq polymerase (Bioline, London, UK). All of the reactions took place in 384-well microtiter plates on a Tetrad thermal cycler (MJ Research, Waltham, MA). The PCR conditions consisted of an initial denaturation step at 95°C for 40 seconds, an annealing step at 58°C for 40 seconds (except for the −863/−857 and −1073/−1031 PCRs, in which the annealing temperatures were 59°C and 62°C, respectively), and an extension step at 72°C for 40 seconds. These steps were repeated 40 times in total, followed by a final extension step at 72°C for 5 minutes.

The PCR products were subsequently purified through incubation with 1 unit of Exo I (New England Biolabs, Beverly, MA) and 1 unit of shrimp alkaline phosphatase (Amersham, Buckinghamshire, UK) at 37°C for 1 hour and then at 72°C for 15 minutes. Each extension reaction, using 1 μl of purified PCR product and 0.25 pmoles of internal probe, was carried out by repeating the following cycle 25 times: 96°C for 10 seconds, 50°C for 5 seconds, and 60°C for 30 seconds. Six microliters of the extension product was incubated with 1 unit of calf intestine alkaline phosphatase (Amersham) at 37°C for 1 hour and then at 72°C for 15 minutes. Five microliters of deionized formamide was mixed with 1 μl of the purified extension product and electrophoresed on an ABI Prism 3100 genetic analyzer (Applied Biosystems, Foster City, CA). If the original DNA template had one or the other polymorphic residue, the extension step would have incorporated the complementary nucleotide that could be detected as one or another color in the extended product. The results were analyzed using GeneScan analysis and the Genotyper, version 3.7, software package (both from Applied Biosystems).

Statistical analysis.

The allele frequencies of each SNP were compared between cases and controls using the chi-square test (29); the odds ratios (OR) with 95% confidence intervals (95% CI) were also calculated. The extended transmission disequilibrium test (ETDT) (30) was used to test for linkage in the presence of association for each position separately and for the TNF SNP haplotypes, in juvenile oligoarthritis as a whole group and stratified by the persistent and extended subgroups of disease. Empirical ETDT P values were calculated by running 10,000 Monte Carlo simulations. The statistical program EHPLUS (31) was used to discern LD patterns between the TNF SNPs and HLA alleles, in order to determine whether the observed contributions of the HLA–A, HLA–DRB1, and TNF loci to oligoarthritis are independent of each other. Finally, the GENEHUNTER software package (32) was used to identify the TNF SNP haplotypes that occurred in the families studied. Ordering the fully typed haplotypes that occurred with a frequency of >1% according to their frequency led to identification of the haplotype-tagging SNPs for TNF. The polymorphic positions that could capture at least 80% of the observed haplotypes were selected as the htSNPs for the locus.


Population-based association study.

The genotypes of the 14 TNF SNPs studied were initially analyzed as part of the association study. Five SNPs (−1073, −244, +472, +943, and +1253) were found to be monomorphic in our sample group of UK Caucasians and were therefore omitted from further study. When the allele frequencies at each position were compared between cases and controls, SNPs −1031, −863, −857, −376, and +1304 showed no evidence of association with the disease. Allele frequencies and ORs with 95% CIs for the TNF SNPs that were associated with juvenile oligoarthritis or one of its subsets are detailed in Table 3. The promoter SNP allele −238G (P = 0.032) and intronic SNP alleles +489A (P = 0.021) and +851A (P = 0.024) were found to be associated with juvenile oligoarthritis. The −238G (P = 0.013), +489A (P = 0.026), and +851A (P = 0.001) alleles were also shown to be associated with the persistent oligoarthritis subgroup. The −308A SNP allele was shown to be associated with oligoarthritis as a whole group (P = 0.007), as well as with the persistent (P = 0.011) and extended (P = 0.027) subsets separately. To account for the number of tests performed, these P values were conservatively corrected by multiplying by 27, because 9 SNPs had been examined in 3 patient groups. The only significant association that was observed after correcting for multiple tests was between the +851A allele and persistent oligoarthritis (P = 0.027).

Table 3. Frequency (%) and odds ratios of tumor necrosis factor (TNF) single-nucleotide polymorphisms in patients and controls*
TNF alleleControlsJuvenile oligoarthritis
Whole groupPersistentExtended
  • *

    Data are from the total number of individuals successfully genotyped for each locus. OR = odds ratio; 95% CI = 95% confidence interval.

−308G131 (87.3)211 (76.4)136 (76.4)75 (76.5)
−308A19 (12.7)65 (23.6)42 (23.6)23 (23.5)
OR (95% CI) for A2.12 (1.2–3.7)3.11 (1.6–6.1)2.11 (1.1–4.1)
−238G139 (91.4)264 (96.3)170 (97.7)94 (94.0)
−238A13 (8.6)10 (3.7)4 (2.3)6 (6.0)
OR (95% CI) for G2.47 (1.1–5.8)3.97 (1.3–12.5)1.47 (0.5–4.0)
+489G160 (94.1)234 (87.3)148 (87.1)86 (87.8)
+489A10 (5.9)34 (12.7)22 (12.9)12 (12.2)
OR (95% CI) for A2.32 (1.1–4.8)2.38 (1.1–5.2)2.23 (0.93–5.4)
+851A152 (86.4)258 (92.8)171 (96.1)87 (87.0)
+851G24 (13.6)20 (7.2)7 (3.9)13 (13.0)
OR (95% CI) for A2.04 (1.1–3.8)3.86 (1.6–9.2)1.06 (0.5–2.2)

Single-point ETDT.

The ETDT was used to separately test each of the 9 polymorphic TNF positions in the juvenile oligoarthritis families. Table 4 provides an overview of the results obtained. The TNF locus was found to be linked and associated with juvenile oligoarthritis and with persistent oligoarthritis. The intronic +851A SNP allele was found to be linked and associated with both oligoarthritis (transmitted [T] 20 times:not transmitted [NT] 7 times) and persistent disease (T 14:NT 1), while the −238G SNP allele exhibited a trend toward deviation from random segregation in the persistent oligoarthritis group (T 7:NT 1). The lack of a positive linkage result for the extended disease subgroup may be attributable to the low number of informative transmissions. Overall, the identification of linkage and association may have been hampered by reduced power, which is inherent to the stratification process, and the examination of biallelic loci.

Table 4. Results of the tumor necrosis factor single-nucleotide polymorphism (SNP) single-point and haplotype extended transmission disequilibrium test*
 Juvenile oligoarthritis
Whole groupPersistentExtended
No. informative transmissionsPNo. informative transmissionsPNo. informative transmissionsP
  • *

    P values are empirical.


LD analysis.

LD patterns were estimated in order to determine whether the observed TNF SNP effects are attributable to the known HLA associations with disease or are independent contributors to juvenile oligoarthritis. All of the associated TNF SNPs were tested against all HLA–A and HLA–DRB1 alleles. The results showed that none of the polymorphic positions in TNF are in LD with any of the associated HLA–A or HLA–DRB1 alleles (HLA–A*02, HLA–DRB1*08, or HLA–DRB1*11) (5, 6), since the lowest P value obtained was 0.4. This indicated that the TNF SNPs play an independent role in the etiopathogenesis of juvenile oligoarthritis.

Haplotype ETDT.

When the family genotype data were entered into the GENEHUNTER program, the set of 37 TNF SNP haplotypes that occurred in this sample group was obtained. Each of these was assigned a name as though they were alleles, and the ETDT was carried out for the TNF locus haplotypes (Table 4). Linkage and association of the TNF locus was, again, demonstrated in the persistent oligoarthritis subgroup, although none of the individual haplotypes exhibited significantly distorted transmission. However, the group of haplotypes that were transmitted in excess, thus contributing to the overall P value, all contained the +851*A allele.

Haplotype-tagging SNPs.

Haplotype-tagging SNPs were determined for the total set of haplotypes, as well as for the transmitted and nontransmitted haplotype groups separately. In all cases, by considering only the −1031, −863, −308, and +489 TNF SNPs, all haplotypes with a frequency of at least 4% could be successfully captured. Specifically, in the total group, 86.9% of the haplotypes could be captured. Similarly, in the transmitted and nontransmitted haplotype groups, the aforementioned 4 SNPs could account for 90.7% (Table 5) and 83.3% (Table 6) of the haplotypes, respectively.

Table 5. Transmitted haplotypes (n = 214) occurring with a frequency of >1% in the juvenile oligoarthritis families*
−1031−863−857−376−308−238+489+851+1304Frequency, %
  • *

    Boldface indicates the haplotypes captured by the 4 haplotype-tagging single-nucleotide polymorphisms.

Table 6. Nontransmitted haplotypes (n = 162) occurring with a frequency of >1% in the juvenile oligoarthritis families*
−1031−863−857−376−308−238+489+851+1304Frequency, %
  • *

    Boldface indicates the haplotypes captured by the 4 haplotype-tagging single-nucleotide polymorphisms.



TNF is a proinflammatory cytokine implicated in the etiopathogenesis of a broad range of diseases. Different lines of evidence, such as the increased protein levels detected in synovial fluid of patients and the encouraging results of anti-TNF treatment, suggest its involvement in juvenile oligoarthritis (8, 9). Genetic investigations of TNF have until now been limited to a few SNPs. This study exhaustively utilizes the list of all known TNF SNPs and examines linkage of the locus to juvenile oligoarthritis. The observed associations of TNF SNPs with juvenile oligoarthritis suggest that this proinflammatory cytokine may be important in the activation of the disease-related processes. Furthermore, the results of the family-based study confirm linkage of the locus and the surrounding region to disease.

The TNF molecule participates in a wide array of complex pathways that involve signaling cascades and regulatory feedback loops. TNF is an important mediator of inflammatory responses, and its expression levels must be fine-tuned, because imbalances can have detrimental effects and trigger pathogenic phenotypes. Transcriptional and translational up-regulation can be determined by variation at the gene level. It is, therefore, possible that the results of our association and linkage study mirror functional effects dictated by genetic polymorphism.

Functional studies have concentrated on the promoter SNPs of the gene but have produced conflicting results. The −308 TNF polymorphism has been studied by several groups and was found by some investigators to have functional effects on transcriptional activity (14, 15), but other investigators did not report similar results (16, 17). The −238 TNF polymorphism has been associated with susceptibility to various other diseases (33, 34), but there is currently little evidence that it has functional properties (35). The intronic +489 and +851 SNPs do not appear to have been studied at the functional level thus far.

Of the markers studied, the strongest predictor of disease susceptibility was the +851 TNF polymorphism, which was highly significantly associated with the persistent oligoarthritis subgroup in both the case–control and the family analyses. At present, little is known about the potential functional properties of SNPs within introns of the TNFα gene, and this is an area that deserves further investigation. This lack of information drives the need for conclusive and in-depth studies that dissect the functional relevance of each polymorphism individually and that characterize possible orchestrated regulatory roles mediated by TNF SNP haplotypes.

Although polymorphisms of TNF are likely to contribute to juvenile oligoarthritis, the complex pattern of associations that has been revealed could also be attributable to LD with another susceptibility locus in the vicinity of the gene. By examining LD patterns, we determined that the effect of TNF is independent of the known HLA–A and HLA–DRB1 associations. The chromosomal region surrounding TNF, however, is abundant in genes of immunologic relevance. In order to identify true susceptibility genes, the genetic variation of the region must be studied, and extended haplotypes must be constructed and analyzed. We are currently applying a systematic fine-mapping approach to this chromosomal segment, in an attempt to determine whether the resident disease-causative locus is TNF itself or another gene in LD with it.

TNF haplotypes have been constructed, and htSNPs for the locus have been identified. These results will be useful in genetic studies of other diseases in which TNF may be of relevance. Results of the analysis of nontransmitted haplotypes reflect SNP patterns observed in the normal population and are applicable to association studies focusing on Caucasian cohorts. However, data stemming from the examination of transmitted haplotypes represent findings in individuals with juvenile oligoarthritis and can be constructive in research concentrating on the same disease.

In conclusion, the genetic polymorphism of TNF appears to play a significant role in juvenile oligoarthritis. It is important to examine the relationship between TNF SNPs and outcome of anti-TNF treatment, because these polymorphic markers may serve as predictors of response to therapy and may be used in targeting children affected by oligoarthritis on the basis of their TNF haplotype. Furthermore, patients in all subgroups of JIA should be screened for TNF polymorphisms that may confer susceptibility to disease. This study has introduced a system that can facilitate genetic research into TNF in JIA to a level previously not undertaken and that can be used to investigate the genetic contribution of TNF in a wide range of other conditions in which TNF is thought to play an active role.