Dr H. S. Howe, Department of Rheumatology, Allergy and Immunology, Tan Tock Seng Hospital, 11, Jalan Tan Tock Seng, Singapore 308433. E-mail: email@example.com
Our objective was to investigate the serum levels of interferon-inducible protein-10 (IP-10) in systemic lupus erythematosus (SLE) and their correlation with disease activity and organ manifestations. Serum IP-10 levels were assessed in 464 SLE patients and 50 healthy donors. Disease activity was assessed by the revised SLE Activity Measure, and the concomitant active organ manifestations, anti-ds DNA antibody titres, complement levels and erythrocyte sedimentation rates recorded. Peripheral blood mononuclear cell (PBMC) synthesis of IP-10 in SLE patients and controls was determined by in vitro cultures stimulated with mitogen or lipopolysaccharide. Elevated serum IP-10 levels were observed in SLE patients, which were significantly higher in the presence of active haematological and mucocutaneous manifestations. SLE PBMCs exhibited enhanced spontaneous IP-10 production in vitro. Serial IP-10 levels correlated with longitudinal change in SLE activity, even at low levels where anti-dsDNA antibody and complement levels remain unchanged. These data demonstrate that IP-10 levels are increased in SLE and serum IP-10 may represent a more sensitive marker for monitoring disease activity than standard serological tests.
Systemic lupus erythematosus (SLE) is an autoimmune disease in which inflammation from infiltration and sequestration of immune cells results in injury and damage to multiple organs. Extensive immune system defects characterize this disease, including polyclonal B cell activation, production of pathological autoantibodies because of excessive T cell help provided to B cells, increased apoptosis and skewed cytokine production [1,2]. A characteristic gene expression profile comprising of type I interferon (IFN)-stimulated genes has been observed in SLE patients. However, as the majority of SLE patients do not have detectable serum or plasma IFN-α, the role of IFN-α as a potential biomarker remains uncertain .
The IP-10 (IFN-inducible protein-10) is a CXC chemokine first cloned in 1985 as a protein secreted by peripheral blood mononuclear cells (PBMC), fibroblasts and endothelial cells . Its biological functions include stimulation of monocytes, natural killer (NK) and T cell migration, regulation of T cell and bone marrow progenitor maturation, modulation of adhesion molecule expansion and inhibition of angiogenesis . Its receptor CXCR3 is most apparent in endothelial cells and fibroblasts, with lower density of binding in monocytes, epithelial cells, B and T lymphocytes and thymocytes. IP-10 is highly inducible in dendritic cells by IFN-α2, promotes the recruitment of monocytes, T and NK cells bearing its CXCR3 receptor into inflamed sites, and may play an important role in both the induction and perpetuation of chronic inflammatory responses . Up-regulated expression of IP-10 mRNA and protein has been described in a variety of autoimmune conditions, such as type 1 diabetes  and rheumatoid arthritis , as well as in chronic infections such as tuberculosis  and chronic hepatitis . In addition, several animal models have established a key role for IP-10 in the onset and progression of inflammatory disease, including the autoimmune pulmonary inflammation of mice susceptible to spontaneous onset SLE .
In the predominantly CD4+ T cell inflammatory infiltrates associated with damage in chronic discoid lupus erythematosus, IP-10 is expressed mainly by basal layer keratinocytes and macrophages [12,13]. Recently, elevated plasma or serum IP-10 levels were found in several studies of small numbers of SLE patients [14–18]; however, the correlation with disease activity was variable. Thus we sought to explore further the serum levels of this chemokine in SLE and their relationship with disease activity and organ manifestations in a larger cohort of 464 oriental lupus patients. In addition, we studied the in vitro response of IP-10 secretion by peripheral blood cells from both SLE patients and healthy controls.
We recruited 464 lupus patients who satisfied the 1997 American College of Rheumatology classification criteria for SLE  and 50 unrelated age- and sex-matched healthy controls (age 30·0 ± 9·8 years; female : male ratio 46:4). Blood samples were collected by venepuncture, and the serum processed immediately and stored at − 80°C until analysis. The SLE patients were from a prospective study cohort begun in 2001 and on follow-up at our hospital in Singapore. Assessments were made 4-monthly for patients whose disease duration was 3 years or fewer, and annually for those whose disease duration was more than 3 years. At each study visit current and past disease manifestations, disease activity, full blood count, erythrocyte sedimentation rate (ESR), complement C3 and C4 levels and anti-ds DNA titres were recorded according to a standard protocol. The study was approved by the hospital's institutional review board and written informed consent was obtained from all patients and controls.
The 464 SLE patients comprised 366 (78·9%) Chinese, 63 (13·6%) Malay, 35 (7·6%) Indian and other racial ethnicity and the female : male ratio was 9·5:1. The mean age was 40·3 ± 12·7 years and mean disease duration 113 ± 93 months; 92 (24·7%) patients had a disease duration of fewer than 36 months. Of 464 patients, 88·4% were on treatment with systemic steroids/immunosuppressive drugs; prednisolone in 88·4%; azathioprine in 30·4%; and cyclophosphamide in 4·8%. Some 62·6% were on hydroxychloroquine.
Measurement of disease activity
Disease activity was assessed at each study visit using the revised SLE Activity Measure (SLAM-R), which has a score ranging from 0 to 81 where a higher score signifies more severe disease . The SLAM-R reports on lupus activity in nine organs and systems including one laboratory category and one other item allowing for scoring of an unlisted feature of lupus disease activity. The nine organs and systems are mucocutaneous (oral/nasal ulcers, peri-ungual erythema, malar rash, photosensitive rash, erythematous, maculo or papular rash, discoid lupus, lupus profundus, bullous lesions, urticarial rash, livedo reticularis, leucocytoclastic vasculitis, palpable purpura, panniculitis, vasculitic ulcer, nail fold infarct or alopecia), musculoskeletal (arthritis), pulmonary (pleurisy/pleural effusion, pneumonitis), cardiac (hypertension, carditis, Raynaud's), gastrointestinal (serositis, pancreatitis, ischaemic bowel or other causes of abdominal pain because of SLE), nervous system (headache, stroke syndromes, cortical dysfunction or seizures), eye (cytoid bodies, papillitis, pseudotumour cerebri, episcleritis or retinal or choroidal haemorrhages), renal (active urinary sediments, proteinuria or renal impairment) and haematological (cytopenias or haemolytic anaemia). Each item in SLAM-R is scored for severity on a scale of 0–3, allowing manifestations to be graded. Disease manifestations in our SLE patients were rated as active when the SLAM-R score for that item was more than zero.
A SLAM-R score of 7 has been suggested as the cut-off for at least a 50% probability of initiating treatment [21,22]. A decline to a score of ≤ 5 has been used to define low levels of disease activity , while a SLAM-R score of > 10 has been used to define high disease activity . We categorized patients based on their SLAM-R scores as having inactive disease if the SLAM-R was ≤ 5, mildly active disease if > 5 to ≤ 10, moderately active disease if > 10 to ≤ 15 and severely active disease if > 15.
In vitro stimulation assays
The PBMC from SLE patients (n = 29) and healthy donors (n = 11) were isolated by gradient centrifugation over Histopaque-1077 (Sigma, St Louis, MO, USA) according to the manufacturer's recommendations and resuspended in complete RPMI-1640 medium supplemented with 2 mM L-glutamine, 100 IU penicillin and 100 mg/ml streptomycin (all obtained from Invitrogen Gibco, Carlsbad, CA, USA). PBMCs (2 × 105/well) were placed in triplicate in 96-well flat-bottomed plates (Nunc, Roskilde, Denmark) in complete RPMI-1640 medium supplemented with 10% fetal calf serum (Invitrogen Gibco) at 37°C 5% CO2 and stimulated with phytohaemagglutinin (PHA) at 10 µg/ml or lipopolysaccharide (LPS; both Sigma) at 5 µg/ml for 3 days. Culture supernatants were harvested and stored at − 20°C until estimated by enzyme-linked immunosorbent assay (ELISA). The production of IP-10 in culture supernatants was determined by ELISA and a P-value of < 0·05 was considered statistically significant.
Laboratory assessments and ELISA
Serum IP-10 levels were determined by sandwich ELISA using paired antibodies from OptEIA Systems (BD Biosciences, San Diego, CA, USA), according to the manufacturer's instructions, with a lower detection limit ≤ 10 pg/ml. Serum IFN-α levels were also determined by ELISA (Bender MedSystems, Burlingame, CA, USA; lower detection limit 7·8 pg/ml). Serum samples were tested routinely in duplicate. Anti-ds DNA (IgG) antibody levels were assessed by ELISA (Euroimmune, AG, Luebeck, Germany) using purified ds-plasmid DNA isolated from Escherichia coli cells as capture antigens and were considered positive if the antibody titre was ≥ 100 IU/ml. Serum complement C3 and C4 levels were determined by nephelometry developed by Dade Behring (Marburg, Germany), according to the manufacturer's protocol. ESR was measured using a modified Westergren method.
The data distribution of all variables was examined. Continuous data are presented herein as mean ± standard deviation when the data are distributed normally or median and interquartile range (IQR) when not. Categorical data are presented as percentages with/without the absolute counts. The IP-10 levels and SLAM-R scores were found to be not normally distributed. Non-parametric χ2 and Mann–Whitney U-tests were performed for univariate comparisons among subject groups. The Cuzick and Altman non-parametric test for trend across ordered group, an extension of the Mann–Whitney U-test for comparison of more than two groups , was used when comparing IP-10 levels between the previously defined categories of disease activity. Correlation between disease activity score, standard disease activity markers and IP-10 levels was assessed using the Spearman correlation.
Receiver operating characteristic (ROC) analysis, used commonly to quantify the accuracy of diagnostic tests that discriminate between two disease states, was performed to compare the performance of levels of serum IP-10 with anti-ds DNA antibody in identifying patients with moderate to severely active SLE. The analysis uses the ROC curve, a graph of the sensitivity versus (1-specificity) of the diagnostic test. The global performance of a diagnostic test is summarized by the area under the ROC curve (AUC), where the greater the AUC is, the better the global performance implied .
As multiple observations and measurements were made of the same individual at different time-points in our longitudinal cohort, it is reasonable to expect that correlation between these observations could exist and would have to be accounted for before valid inferences can be drawn, especially in fitting multiple regression models. This was accomplished by fitting the generalized linear model (GLM), a flexible generalization of ordinary least-squares regression that unifies various statistical models, including linear regression, logistic regression and Poisson regression, under one framework, using the generalized estimating equation approach (GEE) with robust variance estimator to identify independent factors (such as demographic, clinical, serological and therapeutic agents) associated with elevated serum IP-10 levels and high disease activity . A P-value of < 0·01 was considered statistically significant. All statistical analyses were performed using the Intercooled stata version 8·2 for Macintosh (Stata Corporation, College Station, TX, USA).
Serum IP-10 levels and IP-10 production by PBMCs
Analysis of 774 specimens from our cohort found significantly higher median serum levels of IP-10 in SLE patients than in healthy controls [73·1 (IQR: 29·0–187·7) pg/ml versus 22·1 (IQR: 9·4–38·1) pg/ml; P < 0·001]. Median serum IP-10 levels were also significantly higher (P < 0·001) in patients with inactive lupus when compared with healthy controls (Fig. 1).
Having established that SLE patients have higher circulating levels of IP-10, we cultured purified human PBMCs from healthy donors and patients with SLE, either in medium alone or with mitogenic activators such as PHA or LPS, and determined the production of IP-10 in culture supernatants by ELISA. PBMCs from both SLE and healthy donors produced significant amounts of IP-10 in response to both PHA and LPS activation when compared with medium alone (P < 0·01). Interestingly, the level of spontaneous IP-10 production from lupus patients was found to be much higher than that of normal donors (410 ± 80 versus 133 ± 38 pg/ml, P < 0·05). This observation suggests that the in vivo elevated serum IP-10 levels could, in part, be derived from peripheral blood cells.
Correlation of serum IP-10 levels and SLE disease activity
We next examined the relationship between serum IP-10 levels and disease activity as measured by SLAM-R. Based on their SLAM-R scores, 74·5% of patients had inactive disease (SLAM-R ≤ 5), 11·1% mildly active disease (SLAM-R > 5 to ≤ 10), 2·7% moderately active disease (SLAM-R > 10 to ≤ 15) and 1·9% severely active disease (SLAM-R > 15). The median SLAM-R score was 2 (IQR: 1–5). Serum IP-10 levels correlated significantly with the SLAM-R score (r = 0·227, P < 0·001). SLE patients with moderate [174·1 (IQR: 96·6–319·5) pg/ml] or severe [936·7 (IQR: 68·8–1865·2) pg/ml] disease exhibited higher serum IP-10 levels compared with those with inactive lupus [74·0 (IQR: 29·0–150·9) pg/ml; P < 0·001] (Fig. 1). Of 19 patients who were not on treatment, either newly diagnosed lupus patients prior to starting therapy or patients in remission, four had a SLAM-R score higher than 5. Spearman's correlation of serum IP-10 levels with SLAM-R in these patients was 0·513, P = 0·025.
Next, we studied how serum IP-10 levels changed over time in patients whose follow-up spanned four or more study visits. IP-10 levels over a period of 4–21 months in 44 SLE patients were analysed. As shown in Fig. 2, we observed that change in serum IP-10 levels in our SLE patients correlated significantly (r = 0·296, P < 0·001) with change in disease activity as measured by SLAM-R.
Serum IP-10 levels and SLE disease activity markers
To examine further the relationship of IP-10 and disease activity, we studied the correlation between SLAM-R scores and concurrently measured serum IP-10 levels and various laboratory parameters, including ESR, serum complement C3, C4 and titres of anti-ds DNA antibody. Spearman's correlation coefficient between serum IP-10 levels (r = 0·224, P < 0·001) and disease activity was found to be higher than other traditional disease activity markers (anti-dsDNA antibody (r = 0·142, P = 0·001), C3 (r = −0·141, P = 0·001) and C4 (r = −0·101, P = 0·005). We observed that serum IP-10 levels correlated positively with ESR (r = 0·238, P < 0·001) and anti-ds DNA antibody titres (r = 0·127, P < 0·001) and negatively with C3 (r = −0·112, P = 0·004) and C4 (r = −0·180, P < 0·001). Moreover, change in disease activity over time correlated better with change in serum IP-10 levels (r = 0·296, P < 0·001) than in other standard disease-associated parameters (C3: r = −0·149, P = 0·009; C4: r = −0·034, P > 0·1; anti-ds DNA antibody: r = 0·060, P > 0·1). In ROC analysis (Fig. 3), serum IP-10 levels performed better in identifying patients with moderate and severe disease activity than anti-ds DNA antibody titres [AUC (mean ± standard error): 0·828 ± 0·043 versus 0·640 ± 0·069, P = 0·02]. In mild disease activity, change in IP-10 levels correlated with change in SLAM-R (rho = 0·21, P < 0·001), whereas there was no correlation between change in disease activity and change in levels of anti-dsDNA antibody, C3 or C4 (Fig. 2). All the conclusions from the univariate analysis remained unchanged when data were analysed using multivariate analysis with GLM regression to account for multiple sampling of the same individual over time (result not shown).
We also evaluated the correlation of ESR, C3, C4 and anti-ds DNA antibody levels with IP-10 using the GEE. IP-10 levels correlated significantly with ESR and levels of anti-ds DNA antibody (P < 0·001) but not with complement (P > 0·1). Of the independent variables, IP-10 was found to have the strongest correlation with SLAM-R. The correlation between IP-10 and SLAM-R remained significant (P < 0·001) after adjusting for the use of corticosteroid or immunosuppressive/immunomodulatory agents. In addition, as IP-10 is highly inducible by IFN-α, we examined the relationship between serum levels of IFN-α and IP-10 in our SLE cohort. Detectable levels of IFN-α were found in only 61 (7·88%) of the samples and, when present, correlated with serum IP-10 levels (Spearman's correlation = 0·393, P < 0·01). The correlation remained significant in multivariate analysis using the GEE approach (P < 0·01). Taken together, these data suggest a direct association of increased serum levels of IP-10 with disease activity and suggest that IP-10 may be a clinically valuable marker of disease activity.
Serum IP-10 levels and active organ manifestations
We next determined whether serum IP-10 levels correlated with specific SLE organ involvement. Patients with low (≤ 120·6 pg/ml) and high (> 120·6 pg/ml) serum IP-10 levels were compared using univariate analysis for different organ manifestations (Table 1). Patients with active SLE haematological manifestations had significantly higher median serum IP-10 levels than those without [97·0 (IQR: 37·4–195·2) pg/ml versus 63·0 (IQR: 24·2–125·1) pg/ml; P < 0·002]. Serum IP-10 levels were also related to the degree of anaemia (r = 0·159, P = 0·001) and lymphopenia (r = 0·128, P = 0·004). Active mucocutaneous manifestations showed a similar trend [104·0 (IQR: 38·9–263·3) pg/ml versus 75·7 (IQR: 31·6–195·2) pg/ml; P = 0·02]. No significant difference in serum IP-10 levels was observed between those with and without renal or central nervous system manifestations, possibly because of the small number of patients with these manifestations. Active mucocutaneous (P = 0·001) and haematological (P = 0·004) manifestations remained related significantly to serum IP-10 levels after adjusting for the presence of active renal and neurological manifestations with the GEE.
Table 1. Clinical and laboratory features of systemic lupus erythematosus (SLE) patients with normal or high serum interferon (IFN)-inducible protein-10 (IP-10).†
The cut-off serum IP-10 level was defined as the 95th percentile of its levels in the healthy controls (120·6 pg/ml). Values are expressed as mean ± standard deviation. The non-parametric χ2 and Mann–Whitney U-tests were used accordingly because the data did not follow a normal distribution; n.s. = not significant, P > 0·01.
(n = 164)
(n = 300)
41·3 ± 14·4
39·8 ± 11·6
Sex (female : male ratio)
(n = 164)
(n = 300)
Haematological involvement (%)
Mucocutaneous involvement (%)
Neurological involvement (%)
Renal involvement (%)
Other disease activity markers:
(n = 266)
(n = 508)
0·796 ± 0·260
0·850 ± 0·228
0·137 ± 0·094
0·166 ± 0·081
Anti-dsDNA levels (IU/ml)
277·1 ± 378·2
184·0 ± 281·1
In multivariate analysis with the GEE, elevated IP-10 levels (> 120·6 pg/ml) were associated independently with active mucocutaneous manifestations, elevated levels of anti-ds DNA antibody and high ESR, as well as azathioprine use. Treatment with other immunosuppressive/immunomodulatory agents (such as cyclophosphamide or hydroxychloroquine) did not affect IP-10 levels significantly; neither did demographic features, other active organ manifestations or complement levels. After adjusting for possible immunosuppressive/immunomodulatory treatment effects, including the use of various combinations of these agents, IP-10 levels remained related significantly to levels of anti-dsDNA antibody and ESR, and azathioprine use.
Our results validate the finding of previous smaller studies that serum IP-10 levels correlate positively with SLE disease activity; as measured in this study with SLAM-R, anti-ds DNA antibody titres and complement C3 and C4 levels. Importantly, longitudinal change in IP-10 concentrations correlated more closely with disease activity than the standard disease parameters of complement C3, C4 and anti-ds DNA antibody titres.
The following observations support the involvement of IP-10 in the immunopathogenesis of SLE. Our in vitro data show that PBMCs from SLE patients spontaneously produce significantly higher amounts of IP-10 than healthy controls, suggesting that these cells exhibit an up-regulated response in IP-10 production. Increased expression of IP-10 has been found in the basal layers of the epidermis and also in the infiltrating leucocyte subsets clustered in perivascular sites  in the cutaneous lesions of SLE. Thus, the elevated serum IP-10 levels we observed in active cutaneous disease could be related to the enhanced synthesis of IP-10 by activated macrophages and keratinocytes in these lesions [12,13,28,29]. Significantly decreased numbers of CXCR3 expressing lymphocytes in the peripheral blood of patients with subacute cutaneous and disseminated chronic discoid LE is thought to result from recruitment of these cells from the circulating blood into the skin [28,29]. Also, it has been shown that circulating CD4+, CXCR3+ and CD4+CCR2+ T cells are reduced during SLE flares as they are recruited into inflamed tissue sites where increased amounts of IP-10 are produced . While two recent studies demonstrated correlation of serum IP-10 levels with renal disease activity [15,16], we failed to observe a significant correlation in our cohort. The discrepancy may be due to differences in severity and stage of renal disease, as well as the number of patients studied. Also, urine cytokine levels may be a more accurate indicator of localized inflammatory response in the kidneys .
In SLE, a large number of potential biomarkers have been described, but few molecules satisfy the required validation criteria . Generally, a combination of biomarkers is thought to be much more useful than a single biomarker as this is unlikely to have sufficient discriminatory ability. Recently, Behrens suggested that using a score of IFN-regulated chemokines might be clinically useful for monitoring SLE disease activity, as 12 of 23 inflammatory and/or homeostatic chemokines upregulated in SLE were IFN-inducible .
In this study we have validated, in a large oriental cohort, that serum IP-10 levels are elevated in SLE, and also observed that this enhanced expression of serum IP-10 is contributed in part from production by peripheral blood cells. We have shown that changes in serum IP-10 in individual patients correlated significantly with longitudinal change in disease activity. Although this correlation was not strong, it was better than that with both anti-dsDNA titres and complement levels which are used as standard tests of disease activity. Moreover, in contrast to anti-ds DNA titres and complement levels, this correlation remained even at low levels of disease activity. As IP-10 levels correlated with even low levels of disease activity, a rising trend in an individual patient may forewarn of impending clinical flare much earlier than levels of anti-dsDNA antibodies and complement, allowing the patient to be observed more closely or therapeutic intervention implemented. Thus it may be useful to measure serum IP-10 levels in combination with the standard measures of anti-dsDNA and complement levels, particularly in patients whose disease is relatively quiescent, to detect change in disease activity. Further studies are warranted to validate the potential of this chemokine as an additional marker of clinical disease activity.
We thank Chai Hoon Lim and Shir Lin Seet for their technical support. This study received funding support from the Biomedical Research Council of Singapore grant 01/1/28/18/016.