The TNF-863A allele strongly associates with anticentromere antibody positivity in scleroderma




Scleroderma is characterized by the presence of 3 predominant, yet almost mutually exclusive, antibodies: anticentromere antibody (ACA), antitopoisomerase antibody, and anti–RNA polymerase antibody. The purpose of this study was to investigate tumor necrosis factor (TNF) polymorphisms in scleroderma, with the specific aim of determining whether TNF polymorphisms would prove to be stronger markers for ACA than class II major histocompatibility complex (MHC).


We studied 214 UK white scleroderma patients and 354 healthy controls. All subjects were investigated for 5 TNF promoter region polymorphisms by sequence-specific polymerase chain reaction.


We showed that an NF-κB binding site polymorphism (known to be functionally relevant) in the TNF promoter region was present in 51.8% of patients with ACA and 16.3% of patients without ACA (χ2 = 25.1, P = 0.000004 [corrected P = 0.00002]). Using haplotype mapping, we showed that this was a primary TNF association that could explain the previous weak links between ACA production and class II MHC alleles. In marked contrast to our ACA results, HLA class II (especially DRB1*11) appeared to be primary in that it could explain the weaker TNF association with antitopoisomerase production. Further, we observed a separate TNF haplotype to be associated with scleroderma per se, although the level of significance was much lower (χ2 = 8.7, P = 0.003 [corrected P = 0.02]).


We believe these findings may have importance both for the directional pathogenesis of scleroderma progression and for the treatment of scleroderma with anti-TNF agents.

Scleroderma is characterized by the presence of autoantibodies, with ∼95% of patients having detectable antinuclear antibodies (1). There are 3 predominant antibodies in scleroderma: anticentromere antibodies (ACAs), antitopoisomerase antibodies, and anti–RNA polymerase antibodies. These autoantibodies are almost totally mutually exclusive and define different clinical subsets of the disease with reasonable accuracy (2–4).

The tumor necrosis factor (TNF) gene codes for TNFα, a key proinflammatory cytokine that has recently gained in importance in rheumatic disease as the target of new and effective therapies for rheumatoid arthritis (5–7). Anti-TNF drugs have been tried in inflammatory bowel disease and Behçet's disease for several years, with a very good response in a cohort of patients with rheumatoid arthritis (7–10). Therefore, it is not surprising that such therapy is now being extended to other conditions ranging from sarcoidosis (11, 12) to scleroderma (13).

ACA is a marker for pulmonary vascular disease in scleroderma patients. The mechanisms underlying this association might therefore be important with regard to more focused treatment approaches.

We have recently shown that the production of another major autoantibody found in scleroderma patients, antitopoisomerase, is tightly linked both to the presence of fibrosis and to the HLA class II alleles DRB1*1101/04 and DPB1*1301 (14). However, we and other investigators have found only weak associations between DRB1*0101 and DQB1*05 and the production of ACA (15–17). These class II alleles are in linkage disequilibrium with a TNF haplotype that includes at least 1 functional polymorphism (18). Furthermore, TNFα is a potent regulator of type I collagen production in fibroblasts, through a pathway which represses transforming growth factor β induction of connective tissue growth factor and, hence, collagen production (19). TNF imparts its effect through an up-regulation of the transcription factor NF-κB, and it has already been documented that polymorphisms in the TNF gene promoter region can also quantitatively affect NF-κB production (20). Many other factors such as interferon-γ (IFNγ) exert similar effects, but usually through different mechanisms. For example, IFNγ probably exerts its influence via the induction of Smad7 (21). It is therefore feasible that subtle differences in the way that lung and skin fibroblasts regulate collagen expression make some scleroderma patients susceptible to lung fibrosis, although others are not.

The purpose of this study was to investigate TNF polymorphisms in scleroderma, with the specific aim of determining whether TNF polymorphisms would prove to be stronger markers for ACA than class II major histocompatibility complex (MHC). Such a finding would relocate the emphasis from the ACA effect being immunologically mediated (HLA association) toward a TNF effect specific to the function of the particular TNF single-nucleotide polymorphism (SNP) most closely associated.



Two hundred fourteen unrelated white scleroderma patients, mainly from the southern UK and attending the Royal Brompton Hospital and the Royal Free Hospital, were evaluated for TNF polymorphisms. All patients fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) preliminary criteria for scleroderma (22).

Eighty-five of the 189 patients (45.0%) who were tested for ACA were positive, and 47 of the 182 patients (25.8%) who were tested for antitopoisomerase were positive. Forty-nine patients were negative for both ACA and antitopoisomerase, and no patients had both antibodies.

Verbal and written consent was obtained from all patients. Clinical approval was given by the Ethics Committees of the Royal Brompton Hospital and the Royal Free Hospital.

The UK control population comprised 354 unaffected white Caucasians, mainly recruited from the southern UK. All were healthy, as judged by check-ups (including medical history, physical examination, and routine laboratory blood testing) at regular intervals during the 10-year period before blood was obtained for DNA extraction. All gave written consent.

Sequence-specific primers (SSPs) and polymerase chain reaction (PCR).

Genomic DNA from all subjects was extracted from peripheral blood cells. Polymorphisms were determined using SSPs and PCR. A total of 5 TNF promoter SNPs were identified: −1031T/C, −863C/A, −857C/T, −307G/A, and −237G/A. The details of primer sequences and PCR conditions were the same as those previously described (23). HLA genotyping was carried out by the method of Bunce and Welsh (24, 25).

Linkage disequilibrium in TNF haplotype(s) and HLA–DRB1 and DQB1 was analyzed individually both for patients and for controls. The controls in this study were the same as those used in a 30-gene high-density SNP map of the HLA region (18). Haplotype assignment had been previously performed using the sophisticated computer programs PHASE and EM (18). All rare haplotypes were reconfirmed.

Statistical analysis.

The genotype, phenotype, and allele frequencies were determined by direct counting for both control and scleroderma groups. All genotype frequencies were tested for Hardy-Weinberg equilibrium. Knowledge STUDIO (Angoss Software, Toronto, Ontario, Canada) was used both as an audit tool and for data mining. Statistical analysis was performed using chi-square contingency table analysis with the appropriate number of degrees of freedom. Fisher's exact test was used if expected cell frequencies were lower than 5. P values were corrected (Pcorr) for the number of tests performed. P values less than 0.05 were considered significant.


Table 1 summarizes the genotype and allele frequencies of the TNF promoter polymorphisms investigated in the scleroderma patients and controls. All populations were in Hardy-Weinberg equilibrium for all genotype frequencies.

Table 1. Genotype frequencies and allele carriage frequencies of tumor necrosis factor promoter polymorphisms in scleroderma patients and controls*
PositionScleroderma patients (n = 214)Controls (n = 354)
  • *

    Values are percentages.

  • P for trend = 0.01, corrected P (Pcorr) = 0.5, degrees of freedom (df) = 2.

  • P = 0.009, Pcorr = 0.05, df = 1.


We observed an increase in the TNF-1031C allele in scleroderma patients compared with controls (46.3% versus 35.3%; χ2 = 6.7, P = 0.009 [Pcorr = 0.05]). The TNF-1031 T/C genotype was increased in scleroderma patients compared with controls (40.7% versus 29.9%), and the TNF-1031 T/T genotype was decreased in scleroderma patients compared with controls (53.7% versus 64.7%), but there was no significant difference between scleroderma patients and controls after correction (P = 0.01, Pcorr = 0.5, 2 degrees of freedom [df]).

Table 2 shows the genotype and allele carriage frequencies of the TNF promoter polymorphisms in the ACA-positive (group A; n = 85) and antitopoisomerase-positive (group B; n = 47) scleroderma patients and in the scleroderma patients negative for both ACA and antitopoisomerase (group C; n = 49). Patients in these 3 groups did not overlap. The most striking associations were between ACA and both the TNF-1031C allele (64.7% versus 29.8% versus 28.6% for groups A, B, and C, respectively; χ2 = 22.9, P = 0.00001 [Pcorr = 0.00005], 2df) and the TNF-863A allele (51.8% versus 17% versus 16.3% for groups A, B, and C, respectively; χ2 = 25.1, P = 0.000004 [Pcorr = 0.00002], 2df). Furthermore, the TNF-307A allele (25.9% versus 38.3% versus 55.1% for groups A, B, and C, respectively; χ2 = 11.4, P = 0.003 [Pcorr = 0.02], 2df) was decreased in group A and increased in group C. The TNF-857T allele was slightly increased in group B (12.9% versus 27.7% versus 10.2% for groups A, B, and C, respectively; χ2 = 6.7, P = 0.04 [Pcorr not significant], 2df).

Table 2. Genotype frequencies and allele carriage frequencies of tumor necrosis factor promoter polymorphisms in scleroderma patients positive for anticentromere antibody (ACA) or antitopoisomerase antibody (antitopo) and in those negative for both ACA and antitopo*
PositionScleroderma patientsP for trendPcorr
ACA positive (n = 85)Antitopo positive (n = 47)ACA and antitopo negative (n = 49)
  • *

    Values are percentages. NS = not significant.

  • P values and corrected P (Pcorr) values that were significant are shown in boldface.

  • Four degrees of freedom.

 −1031   0.000050.0003
 −863   0.000040.0002
 −307   0.020.1

Construction of haplotypes.

We defined 6 major haplotypes with the 5 TNF promoter polymorphisms (Table 3). Haplotype 1 was decreased (χ2 = 8.7, P = 0.003 [Pcorr = 0.02]) and haplotype 6 was increased (P = 0.000002, Pcorr = 0.00001) in scleroderma patients as a group (Figure 1). However, neither haplotype discriminated between ACA and antitopoisomerase (P not significant) (Figure 2). Figure 2 shows the frequencies of the haplotypes in normal controls and in patients who were positive for ACA (group A), positive for antitopoisomerase (group B), and negative for both ACA and antitopoisomerase (group C). ACA was strongly associated with haplotype 3, which includes the C allele at position −1031 and the A allele at position −863 (χ2 = 24.2, P = 0.000005 [Pcorr = 0.00003], 2df). Antitopoisomerase was associated with haplotype 4, which includes the −857T allele (P = 0.007, Pcorr = 0.04, 2df). Patients negative for both ACA and antitopoisomerase showed associations with haplotype 2, which includes the −307A allele (P = 0.003, Pcorr = 0.02, 2df).

Table 3. Tumor necrosis factor (TNF) promoter haplotypes*
  • *

    Rarer alleles are shown in boldface. Reprinted, with permission, from ref. 23.

Figure 1.

TNF promoter haplotype frequencies (%) in scleroderma patients (solid bars) and in controls (open bars). * = P = 0.003, corrected P (Pcorr) = 0.02; ** = P = 0.000002, Pcorr = 0.00001; *** = P = 0.05, Pcorr not significant.

Figure 2.

TNF promoter haplotype frequencies in controls (blue), anticentromere antibody (ACA)–positive scleroderma patients (group A; red), antitopoisomerase antibody–positive scleroderma patients (group B; yellow), and scleroderma patients negative for both ACA and antitopoisomerase antibodies (group C; green). Haplotype 2 was significantly decreased with group A and increased with group C (P = 0.003, corrected P [Pcorr] = 0.02, 2 degrees of freedom [df]). Haplotype 3 was significantly increased with group A (P = 0.000005, Pcorr = 0.00003, 2df). Haplotype 4 was significantly increased with group B (P = 0.007, Pcorr = 0.04, 2df).

Linkage disequilibrium between TNF haplotype and HLA–DRB1 and DQB1.

We investigated the linkage disequilibrium between TNF haplotypes and HLA–DRB1 and DQB1 alleles for controls and scleroderma patients. Haplotypes were defined from the TNF haplotype as the anchor point using D′ (26). TNF haplotype 1 was in linkage disequilibrium with DRB1*15 and DQB1*0602, TNF haplotype 2 was in linkage disequilibrium with DRB1*03 and DQB1*02 (*0201), and TNF haplotype 5 was in linkage disequilibrium with DRB1*07 and DQB1*0303. TNF haplotype 3 showed the strongest association with DRB1*01 (χ2 = 11.3, P = 0.0008) and DQB1*0501 (χ2 = 9.9, P = 0.002), and TNF haplotype 4 showed the strongest association with DRB1*11 (χ2 = 10.9, P = 0.0009) and DQB1*0301(χ2 = 9.2, P = 0.002), but the D′ for both of these haplotypes was less than 0.5 (Table 4).

Table 4. Linkage disequilibrium in the tumor necrosis factor (TNF) haplotype, HLA–DRB1 and DQB1*
TNF haplotype, HLA alleles in disequilibriumScleroderma patientsControls
  • *

    Values are D′. D′ values ≥0.5 are shown in boldface. Haplotypes were defined from the TNF haplotype as the anchor point (equal to 1).


In ACA-positive patients, the strongest association with HLA was with HLA–DRB1*01 (χ2 = 6.0), and the strongest association with a TNF polymorphism was with TNF-863A (χ2 = 27.7). In contrast, in antitopoisomerase-positive patients, the strongest association with HLA was with HLA–DRB1*11 (χ2 = 15.9), and the strongest association with a TNF polymorphism was with TNF-857T (χ2 = 6.8).


In this study, we have investigated the association between 5 potentially functional TNF promoter polymorphisms and scleroderma. We detected a significant increase in the allele frequency of the TNF-1031C allele in the whole group of scleroderma patients compared with the controls. Furthermore, with the construction of TNF promoter haplotypes, only haplotype 6 was significantly increased in scleroderma patients, although haplotypes 3, 5, and 6 include TNF-1031C. This suggests that the primary association is not with TNF-1031C but with other polymorphisms present only on haplotype 6. This will be investigated in further studies.

The major finding of the present study relates to the association between autoantibodies (ACA, antitopoisomerase) and TNF polymorphisms; ACA showed very strong associations with both TNF-1031C and TNF-863A and negative associations with the TNF-307A allele. Antitopoisomerase showed a positive association with the TNF-857T allele and a negative association with both TNF-1031C and TNF-863A. Haplotype analysis showed ACA to have a very strong association (Pcorr = 0.00003) with haplotype 3, which includes the TNF-1031C and TNF-863A alleles. Since only haplotype 3 includes TNF-863A, while haplotypes 3, 5, and 6 include TNF-1031C, we believe that TNF-863A is the key player in ACA-positive scleroderma patients. However, the linkage disequilibrium in Caucasian patients and controls between alleles at −863 and −1031 is such that very large patient groups or groups from other ethnic backgrounds are needed to confirm our hypothesis. In terms of function, it is perhaps less important to prove this, since both sites are thought to be related to NF-κB binding (20).

Udalova and colleagues demonstrated a clear effect of the change from C to A at TNF-863 on the relative binding affinities of different forms of the NF-κB complex (20). It was shown that the p50/p50 homodimeric form of this complex had a significantly decreased affinity to its DNA binding site for −863A. Since the p50/p50 homodimer acts as a transcription repressor on binding to its regulatory site in the promoter region of the TNF gene, decreased binding is thought to result in an inadequate down-regulation of TNF gene expression, and thus in increased TNFα production. It is interesting to note in this context that Young and colleagues recently reported an increase in the mean TNFα expression in scleroderma patient leukocytes (27).

We then evaluated the question of previously reported class II MHC associations with scleroderma autoantibody subsets. Could the known strong linkage disequilibrium observed in this region explain our TNF findings, or were alleles at both TNF and class II important? Our group and others (16, 28) have previously reported the association between ACA and HLA–DRB1*01 and HLA–DQB1*0501, as well as the negative association between ACA and HLA–DQB1*0201 (no TNF polymorphisms were included in these studies). In the current study, we found a strong association between ACA and the TNF-863A allele, the TNF-1031C allele, and TNF haplotype 3.

Can linkage disequilibrium within the region explain this apparent disparity? In this regard, we were able to utilize the data generated in a much larger haplotype analysis across the MHC that was carried out with the same control population (28). Haplotype 3 is indeed in linkage disequilibrium with HLA–DRB1*01 and HLA–DQB1*0501, and while D′ is not strong, it is sufficient to explain the weak observed association. TNF haplotype 2 is in tight linkage disequilibrium with HLA–DRB1*03 and HLA–DQB1*0201 and can explain the reduction in frequency of HLA–DQB1*0201. Thus, although both HLA class II and TNF polymorphisms are associated with ACA, the association with the TNF polymorphism is stronger, and linkage disequilibrium can explain all of the HLA class II association (i.e., it is the NF-κB site within the TNF promoter which is the major or sole cause of the observed association with ACA). We found no evidence for a second independent marker in the class II MHC region for ACA production. This may have important mechanistic implications.

In marked contrast, the antitopoisomerase association is completely explainable by linkage disequilibrium with the known MHC markers. We conclude that, in contrast to ACA, HLA class II is the prime or sole candidate for association with antitopoisomerase production. Partial support for this comes from Kuwana et al, who reported that antibodies to HLA–DR and HLA–DQ blocked antitopoisomerase production in vivo (29). However, this leaves our HLA–DP association unexplained (14), since DP and TNF are not in linkage disequilibrium, and antibodies to DP did not block antitopoisomerase production in the system of Kuwana et al.

In summary, we have observed a series of associations between TNF intragene haplotypes and scleroderma as well as within scleroderma subsets. The association between TNF haplotype 3 and the ACA subset is highly significant (χ2 = 24.2, P = 0.000005 [Pcorr = 0.00003], 2df), making this transcription factor binding region of the TNF promoter the strongest non-MHC genetic marker so far described in this disease. Why, then, does a disease that is clearly not familial have strong genetic associations? The fact that HLA and TNF associations in scleroderma associate with different clinical subsets does not fully explain this apparent dichotomy. If scleroderma really consisted of separate diseases in terms of susceptibility, with converging final pathways, then we would expect each subset with a strong association to be familial. We therefore believe the most likely explanation is that the primary cause of scleroderma is not genetic, but the way the disease progresses most definitely is. We say “primary cause” because in one Choctaw Indian population, the disease is indeed both familial and directly genetically associated (30).