Large Scale Analysis of Tumor Necrosis Factor Alpha Levels in Systemic Lupus Erythematosus

Abstract

Background

SLE disease manifestations are highly variable between patients, and the prevalence of individual clinical features differs significantly by ancestry. Serum tumor necrosis factor alpha (TNF-α) is elevated in some SLE patients, and may play a role in disease pathogenesis. We detected associations between serum TNF-α, clinical manifestations, autoantibodies, and serum IFN-α in a large multi-ancestral SLE cohort.

Methods

We studied serum TNF-α in 653 SLE patients, including 214 African-American, 298 European-Americans and 141 Hispanic-American subjects. TNF-α was measured using ELISA, and IFN-α was measured with a functional reporter cell assay. Stratified and multivariate analyses were used to detect associations in each ancestral background separately, with meta-analysis when appropriate.

Results

Serum TNF-α levels were significantly higher in SLE patients than in nonautoimmune controls (p<5.0×10⁻³ for each ancestral background). High serum TNF-α was positively correlated with high serum IFN-α when tested in the same sample across all ancestral backgrounds (meta-analysis OR=1.8, p=1.2×10⁻³). While serum TNF-α levels alone did not differ significantly between SLE patients of different ancestral backgrounds, the proportion of patients with concurrently high TNF-α and high IFN-α was highest in African-Americans and lowest in European-Americans (p=5.0×10⁻³). Serum TNF-α was not associated with autoantibodies, clinical criteria for the diagnosis of SLE, or age at time of sample.

Conclusions

Serum TNF-α levels are high in many SLE patients, and we observed a positive correlation between serum TNF-α and IFN-α. These data support a role for TNF-α in SLE pathogenesis across all ancestral backgrounds, and suggest important cytokine subgroups within the disease.

Systemic lupus erythematosus (SLE) is characterized by a wide variety of clinical manifestations, including inflammation of the skin, renal, hematologic and musculoskeletal organ systems. Differences in the prevalence of particular clinical and serologic manifestations of disease by ancestral background have long been appreciated (1). Some of the clinical differences observed between ancestral backgrounds likely represent differences in biological pathways related to disease pathogenesis, although little is currently known about the molecular mediators of these differences.

Previous studies have documented elevated serum tumor necrosis factor alpha (TNF-α) levels in some patients with SLE, and these levels have been correlated with clinical disease activity and anti-dsDNA antibodies (2, 3), and TNF-α is over-expressed in renal tissue in lupus nephritis (4). Although TNF-α is present at sites of inflammation, the role TNF-α plays in human SLE pathogenesis remains controversial. The role of TNF-α in murine models of SLE has been similarly controversial. In some models TNF-α improved disease features, while in others TNF-α blockade has been beneficial (4).

Interferon alpha (IFN-α) and TNF-α appear to cross-regulate each other in vitro (5). TNF-α inhibits peripheral dendritic cell generation and secretion of IFN-α by these cells (5). In healthy PBMCs, culture with etanercept led to an increase of IFN-α and IFN-α-inducible genes, and IFN-α inhibits secretion of TNF-α (5, 6). Many lines of evidence support the idea that IFN-α is a primary pathogenic factor in SLE, including the development of SLE in patients given recombinant IFN-α to treat viral infections and malignancy, and familial aggregation of high serum IFN-α in SLE families (7, 8). Thus, there is some reasonable concern that SLE-like features which have arisen during anti-TNF-α therapy may relate to increased IFN-α (9), and that this increase in IFN-α could catalyze a change of the clinical syndrome from rheumatoid arthritis to SLE. Clinical trials in human SLE suggest that short-term TNF-α blockade may have benefit in lupus nephritis, as well as transient benefit in SLE arthritis (4), but some significant side effects have been reported in a small group of patients who have received long term anti-TNF-α therapy (10).

In the present study, we examine relationships between serum TNF-α levels and simultaneous IFN-α measurements, serologic, and clinical parameters in SLE. Given the interrelated nature of many of the clinical and serologic characteristics in SLE and the potential for a relationship between TNF-α and IFN-α, we used multivariate regression models to account for between-variable relationships.

PATIENTS AND METHODS

Patients, Samples and Data

Serum samples were obtained from 653 SLE patients (214 African Americans, 298 European American and 141 Hispanic American patients) from the Lupus Family Registry and Repository at the Oklahoma Medical Research Foundation (OMRF). All subjects met the American College of Rheumatology (ACR) criteria for the diagnosis of SLE, and the presence or absence of these criteria as well as of SLE-associated autoantibodies (antinuclear antibody, anti-Ro, anti-La, anti-Sm, anti-RNP and anti-dsDNA) were available for all subjects. 62 unrelated individuals who were screened by medical record review for the absence of autoimmune disease were included in the study as controls. The controls were of similar age (mean age = 45.6 years, SD = 12.9 years), gender (90.3% female), and ancestral background (39% African-American, 44% European-American, 15% Hispanic-American) as the SLE patients. All subjects provided informed consent, and the study was approved by the institutional review board.

Measurement of serum IFN-α activity

We have developed a sensitive and reproducible bioassay to detect serum IFN-α activity (8), as ELISA methods for detection of IFN-α in human serum have been complicated by low sensitivity and specificity. In this bioassay, reporter cells (WISH cells, ATCC #CCL125) are used to measure the ability of sera to cause IFN-induced gene transcription. The reporter cells are cultured with patient sera for 6 hours and then lysed, and three canonical IFN-α-induced transcripts (IFIT-1, MX-1, and PKR) are measured using rtPCR. Relative expression data from the three transcripts are then normalized to healthy donor sera (n=141) run in the same assay, and data are presented as an IFN-α activity score. For more details please see (8).

Measurement of TNF-α levels

Serum TNF-α was measured in all of our samples using a commercial ELISA kit (Pierce, Rockford, IL). According to the manufacturer, addition of TNF-α receptor types 1 and 2 (40 mg/ml) does not interfere with this assay. Patient samples were considered positive if they were more than two standard deviations above the mean of our non-autoimmune control population.

Statistical Analysis

The SLE cohort was stratified by self-reported ancestral background (African-American, European-American, and Hispanic-American), and each ancestral background was analyzed separately. Quantitative TNF-α data were compared using the non-parametric Mann-Whitney U test. Categorical analyses were done using chi-square test statistics to compare proportions between groups. In addition to stratification by ancestral background, in some analyses subjects were further stratified categorically into high vs. low TNF-α and IFN-α subgroups (high = 2SD above the mean of healthy donors). This resulted in four groups representing all four possible combinations of the two categorical variables as follows: high IFN-α/high TNF-α, low IFN-α/high TNF-α, high IFN-α/low TNF-α, and low IFN-α/low TNF-α.

Multivariate modeling was performed using logistic regression. The model included European vs. non-European ancestry, age at recruitment, each of the ACR criteria, each autoantibody, and IFN-α high vs. low as predictor variables, and TNF-α high vs. low was the outcome variable. The ACR criterion “anti-nuclear antibodies” was not included in the regression as it was almost uniformly positive in SLE patients, and the “Immunological disease” criterion was not included as its individidual components anti-dsDNA and anti-Sm were already part of the model. Results with p<0.0029 in the multivariate analysis would withstand a Bonferroni correction for multiple comparisons accounting for the number of variables tested.

RESULTS

Serum TNF-α are higher in SLE patients than non-autoimmune controls

When quantitative serum TNF-α was examined in SLE patients, levels were significantly higher in SLE patients than in non-autoimmune controls across all ancestral backgrounds (Figure 1A). There were no significant differences between SLE patients of different ancestral backgrounds (all p>0.09). The non-autoimmune controls were well matched by age, gender, and ancestral background to the SLE patients (see Methods above), and TNF-α levels were not significantly different between controls of different ancestral backgrounds.

Figure 1.

Serum TNF-α in SLE patients and healthy controls. A. shows quantitative TNF-α levels in SLE patients stratified by ancestral background. Lines indicates the median, boxes show the interquartile range, and error bars show the 10^th and 90^th percentiles. P-values calculated using the Mann-Whitney U test. B. shows the proportion of SLE patients with high vs. low TNF-α and IFN-α across the different studied ancestral backgrounds. Bars represent the percentage of subjects in each cytokine category from each ancestral background, with p-values comparing the differences in proportion of subjects in a given cytokine group by ancestral background, calculated by Chi-square test statistic. * = p<0.05, ** = p<0.005, *** = p<0.0005. AA = African-American, EA = European-American, HA = Hispanic-American.

Different combinations of high vs. low TNF-α and IFN-α are observed in different ancestral backgrounds

In Figure 1B, we divided each ancestral background into four groups, stratifying by high vs. low levels of both TNF-α and IFN-α. African-Americans have the highest proportion of individuals in the high TNF-α/high IFN-α group, which is statistically significantly different from the European-American group which has the lowest proportion (p=5.0×10⁻³). Conversely, European-Americans have the highest proportion of patients with low TNF-α/low IFN-α, while African-Americans have the lowest proportion in that group (p=2.7×10⁻⁴).

High TNF-α is associated with high IFN-α across all ancestral backgrounds, but is not associated with clinical or serologic features

When comparing clinical characteristics of the subjects across the three backgrounds, we confirm many of the previously demonstrated differences between ancestral backgrounds (Table 1). Similar to the quantitative analysis above, in categorical analysis there was no statistically significant difference in the proportion of high TNF-α subjects across the three ancestral backgrounds. In Supplemental Table 1 we show the prevalence of the various clinical characteristics and autoantibodies when stratified by high vs. low TNF-α levels in each background. In African-Americans, low TNF-α subjects were more likely to have photosensitivity than high TNF-α subjects. The only consistent trend across ancestral backgrounds was that high TNF-α subjects were more likely to have high IFN-α levels. The odds ratios for high TNF-α predicting high IFN-α were strikingly similar in each ancestral background (ORs range from 1.76 to 1.86, p = 4.8×10⁻³ by Fisher's combined probability test).

Table 1.

Prevalence of clinical characteristics, autoantibodies, TNF-α and IFN-α in SLE patients stratified by ancestral background

Patient characteristic	European Americans (%)	Hispanic Americans (%)	African Americans (%)	X2	p-value
Female	86.6	88.7	92.1	3.77	0.15
Age* +/−SD	44.0+/−14.6	39.3+/−13.8	41.8+/−13.0	-	0.0039
Photosensitivity	50.7	46.1	31.8	18.59	9.2×10−5
Discoid Rash	9.40	14.9	26.7	27.46	1.1×10−6
Malar Rash	59.4	49.6	37.4	24.16	5.7×10−6
Oral Ulcer	33.9	30.5	24.3	5.48	0.065
Serositis	42.6	40.4	33.6	4.32	0.12
Hematologic	56.7	58.9	67.3	6.08	0.048
Neurologic	12.1	9.9	15.4	2.51	0.29
Renal	35.2	46.8	42.1	5.92	0.052
Arthritis	82.9	61.7	74.3	23.47	8.0×10−6
Immunologic	80.9	80.1	84.1	1.2	0.55
ANA	98.7	96.5	99.5	5.47	0.065
Anti-dsDNA	35.6	29.8	31.3	1.83	0.40
Anti-Sm	3.36	4.26	11.2	14.47	7.2×10−4
Anti-RNP	10.4	17.0	38.8	62.02	3.4×10−14
Anti-Ro	22.1	25.5	26.2	1.27	0.53
Anti-La	6.04	9.22	6.07	1.76	0.41
High TNF-α	29.2	36.2	36.4	3.74	0.15
High IFN-α	32.6	41.8	51.4	18.43	1.0×10−4

Patient characteristics are shown as the percentage having that characteristic in each ancestral background.

^*Age is the exception, which is shown as mean years of age at time of sample, +/− standard deviation. Differences in proportions across the three ancestral backgrounds are tested using a Chi-square test statistic (X²) with 2 degrees of freedom, and difference in age is assessed by one-way ANOVA.

Multivariate analysis of associations between TNF-α, IFN-α, autoantibodies,clinical, and demographic features in SLE patients

We performed a multivariate analysis to detect independent associations with high serum TNF-α in SLE patients, including all of the available variables for each patient such as ancestry, age, clinical features and autoantibodies present vs. absent, and serum IFN-α high vs. low. As shown in Table 2, only high IFN-α was significantly associated with high TNF-α. To explore clinical associations further, we examined the prevalence of clinical manifestations in patient subgroups representing each of the four possible combinations of high vs. low TNF-α and IFN-α shown in Figure 1B, but no significant relationships were observed that were related to TNF-α levels (data not shown).

Table 2.

Results of multivanate regression to detect independent associations with high TNF-α levels across all ancestral backgrounds.

Patient characteristic	β coeff.	Standard error	p-value
Non-European ancestry	−0.270	0.191	0.15
Age at recruitment	−0.003	0.007	0.71
Photosensitivity	−0.271	0.186	0.14
Discoid rash	−0.370	0.248	0.14
Malar rash	0.368	0.186	0.048
Oral ulcers	0.090	0.197	0.65
Serositis	−0.167	0.184	0.36
Hematologic	−0.089	0.183	0.63
Neurologic	0.388	0.254	0.13
Renal	0.013	0.186	0.94
Arthritis	−0.182	0.210	0.39
Anti-dsDNA	−0.481	0.196	0.014
Anti-Sm	−0.371	0.412	0.37
Anti-RNP	0.057	0.247	0.82
Anti-Ro	0.209	0.234	0.37
Anti-La	0.346	0.382	0.37
IFN-α	0.611	0.189	0.0012

Each of the patient characteristics in the table was used as a predictor variable in the logistic model, with TNF-α high vs. low used as the outcome variable. B-coeff. = beta coefficient for each variable in the logistic model.

Complement CH50 was associated with cytokine parameters

Given that immune complexes may be related to TNF-α and IFN-α levels, we looked for a correlation between complement CH50 testing and TNF-α and IFN-α. Most patients had CH50 data available, and in European-Americans lower CH50 was associated with increased TNF-α and IFN-α (Supplemental Table 2). In African-Americans and Hispanic-Americans, CH50 was inversely correlated with IFN-α, but was not correlated with TNF-α. The lack of relationship with TNF-α in these backgrounds could be due to smaller sample size, although no strong trends were observed and this may represent a biological difference.

DISCUSSION

Our study confirms that elevated serum TNF-α is a frequent finding in human SLE, slightly less prevalent than high IFN-α in SLE using the same criteria (11). Interestingly, we found a positive correlation between TNF-α and IFN-α, but no correlation between TNF-α and any clinical or serologic features of SLE. This suggests that TNF-α may cooperate with IFN-α or other cytokine or environmental factors not measured in this study to result in an impact upon clinical features of disease. Some previous studies have documented a relationship between high TNF-α and anti-dsDNA antibodies (4), but we did not replicate this finding in our study. It is possible that the previously described association between high TNF-α and anti-dsDNA antibodies may have been secondary to high IFN-α, as high IFN-α is strongly associated with both of these characteristics. We do not have disease activity data available in this study from the time of sample. The correlation between CH50 and TNF-α in European-American ancestry could indicate a relationship between TNF-α and disease activity, and further large scale longitudinal studies would be warranted.

While in vitro studies suggest cross-regulation between TNF-α and IFN-α (5), not all studies of autoimmune disease populations fit this model. For example, upregulation of both type I IFN and TNF-α has been shown in synovial biopsies from rheumatoid arthritis (12). In juvenile dermatomyositis, the promoter polymorphism in the TNF-α gene which has been linked to higher TNF-α expression was associated with increased serum IFN-α levels (13). Thus, it is likely that cross-regulation of TNF-α and IFN-α in humans in vivo will be complex. In our cross-sectional study of SLE sera, we do not have evidence for cross-regulation, as we instead find levels of the two cytokines to be positively correlated, even when controlling for clinical and demographic variables. Mechanistically, this could relate to immune complexes triggering Toll-like receptors, which could result in production of both inflammatory cytokines such as TNF-α as well as IFN-α (14). Correlations observed between CH50, a proxy for immune complexes, and these two cytokines may relate to this phenomenon. We have not tested IFN-α levels in SLE patients receiving anti-TNF-α drugs, so we do not know if IFN-α would increase. This would be of interest, as some of the hesitation to use anti-TNF-α drugs in SLE stems from the idea that blocking TNF-α may result in a compensatory increase in IFN-α which could worsen SLE disease activity.

It is not clear whether high TNF-α predisposes to SLE, or if levels rise after disease is established. One previous study reported a difference in simultaneous TNF-α and IFN-α levels related to PTPN22 genotype (15), although the variant classically associated with SLE-risk was associated with lower TNF-α. Additionally, the promoter polymorphism in the TNF-α gene which has been linked to higher TNF-α has been associated with SLE susceptibility, although the TNF-α locus is in the HLA region which is characterized by multiple association signals that are difficult to resolve due to high linkage disequilibrium in the region. Given the frequent elevation of TNF-α in SLE sera, it seems likely that TNF-α plays a role in the disease process. Further understanding of this role will increase our knowledge of disease pathogenesis, and may allow for the definition of a subset of patients who could benefit from TNF-α blockade, which has been safe and effective in multiple other inflammatory diseases.