Coordinate overexpression of interferon-α–induced genes in systemic lupus erythematosus

Authors

  • Kyriakos A. Kirou,

    1. Hospital for Special Surgery, and Weill Medical College of Cornell University, New York, New York
    Search for more papers by this author
  • Christina Lee,

    1. Hospital for Special Surgery, and Weill Medical College of Cornell University, New York, New York
    Search for more papers by this author
  • Sandhya George,

    1. Hospital for Special Surgery, and Weill Medical College of Cornell University, New York, New York
    Search for more papers by this author
  • Kyriakos Louca,

    1. Hospital for Special Surgery, and Weill Medical College of Cornell University, New York, New York
    Search for more papers by this author
  • Ioannis G. Papagiannis,

    1. Hospital for Special Surgery, and Weill Medical College of Cornell University, New York, New York
    Search for more papers by this author
  • Margaret G. E. Peterson,

    1. Hospital for Special Surgery, and Weill Medical College of Cornell University, New York, New York
    Search for more papers by this author
  • Ngoc Ly,

    1. Expression Diagnostics, Inc., South San Francisco, California
    Search for more papers by this author
    • Drs. Woodward, Fry, and Wohlgemuth and Mr. Ly, Ms Lau, and Mr. Prentice hold stock in Expression Diagnostics, Inc. and are listed as inventors on a patent application submitted by Expression Diagnostics, Inc. for the use of peripheral blood gene expression for diagnosis and monitoring of autoimmune diseases.

  • Robert N. Woodward,

    1. Expression Diagnostics, Inc., South San Francisco, California
    Search for more papers by this author
    • Drs. Woodward, Fry, and Wohlgemuth and Mr. Ly, Ms Lau, and Mr. Prentice hold stock in Expression Diagnostics, Inc. and are listed as inventors on a patent application submitted by Expression Diagnostics, Inc. for the use of peripheral blood gene expression for diagnosis and monitoring of autoimmune diseases.

  • Kirk E. Fry,

    1. Expression Diagnostics, Inc., South San Francisco, California
    Search for more papers by this author
    • Drs. Woodward, Fry, and Wohlgemuth and Mr. Ly, Ms Lau, and Mr. Prentice hold stock in Expression Diagnostics, Inc. and are listed as inventors on a patent application submitted by Expression Diagnostics, Inc. for the use of peripheral blood gene expression for diagnosis and monitoring of autoimmune diseases.

  • Anna Yin-Har Lau,

    1. Expression Diagnostics, Inc., South San Francisco, California
    Search for more papers by this author
    • Drs. Woodward, Fry, and Wohlgemuth and Mr. Ly, Ms Lau, and Mr. Prentice hold stock in Expression Diagnostics, Inc. and are listed as inventors on a patent application submitted by Expression Diagnostics, Inc. for the use of peripheral blood gene expression for diagnosis and monitoring of autoimmune diseases.

  • James G. Prentice,

    1. Expression Diagnostics, Inc., South San Francisco, California
    Search for more papers by this author
    • Drs. Woodward, Fry, and Wohlgemuth and Mr. Ly, Ms Lau, and Mr. Prentice hold stock in Expression Diagnostics, Inc. and are listed as inventors on a patent application submitted by Expression Diagnostics, Inc. for the use of peripheral blood gene expression for diagnosis and monitoring of autoimmune diseases.

  • Jay G. Wohlgemuth,

    1. Expression Diagnostics, Inc., South San Francisco, California
    Search for more papers by this author
    • Drs. Woodward, Fry, and Wohlgemuth and Mr. Ly, Ms Lau, and Mr. Prentice hold stock in Expression Diagnostics, Inc. and are listed as inventors on a patent application submitted by Expression Diagnostics, Inc. for the use of peripheral blood gene expression for diagnosis and monitoring of autoimmune diseases.

  • Mary K. Crow

    Corresponding author
    1. Hospital for Special Surgery, and Weill Medical College of Cornell University, New York, New York
    • Department of Medicine, Hospital for Special Surgery, 535 East 70th Street, New York, NY 10021
    Search for more papers by this author
    • Dr. Crow serves on the Scientific Advisory Board of, and holds stock options in, Expression Diagnostics, Inc.


Abstract

Objective

To study the contribution of interferon-α (IFNα) and IFNγ to the IFN gene expression signature that has been observed in microarray screens of peripheral blood mononuclear cells (PBMCs) from patients with systemic lupus erythematosus (SLE).

Methods

Quantitative real-time polymerase chain reaction analysis of healthy control PBMCs was used to determine the relative induction of a panel of IFN-inducible genes (IFIGs) by IFNα and IFNγ. PBMCs from 77 SLE patients were compared with those from 22 disease controls and 28 healthy donors for expression of IFIGs.

Results

Expression of IFNα-inducible genes was significantly higher in SLE PBMCs than in those from disease controls or healthy donors. The level of expression of all IFIGs in PBMCs from SLE patients with IFNα pathway activation correlated highly with the inherent responsiveness of those genes to IFNα, suggesting coordinate activation of that cytokine pathway. Expression of genes preferentially induced by IFNγ was not significantly increased in SLE PBMCs compared with control PBMCs. IFNα-regulated gene-inducing activity was detected in some SLE plasma samples.

Conclusion

The coordinate activation of IFNα-induced genes is a characteristic of PBMCs from many SLE patients, supporting the hypothesis that IFNα is the predominant stimulus for IFIG expression in lupus.

Genetic factors, environmental triggers, and stochastic events act together to produce the myriad immunologic alterations that characterize systemic lupus erythematosus (SLE). In lupus, T cells provide help for the production of antibodies reactive with intracellular particles composed of nucleic acids and nucleic acid binding proteins. Those autoantibodies deposit in tissue and induce an inflammatory response that results in tissue damage and disease. Current models of disease pathogenesis implicate the presentation to CD4+ T cells of antigens derived from apoptotic cell fragments by activated dendritic cells (1). Recent studies have focused on mechanisms of dendritic cell activation to explain the generation of systemic autoimmunity in some individuals but not others (2, 3). Interferons (IFNs) are important immune system mediators that could impact the initiation or amplification of autoimmunity and tissue damage through their diverse actions on dendritic cells, T and B lymphocytes and natural killer cells, and mononuclear phagocytes (4).

Several lines of evidence link the IFNs to SLE. Type I IFNs, including IFNα, as well as IFNβ, IFNτ, IFNω, and IFNκ, play a critical role in innate immunity and defense against viruses. Type II IFN is represented by IFNγ, a critical factor in effector mechanisms of adaptive immune responses. Serum levels of IFNα are elevated in active SLE, and therapeutic IFNα has occasionally been noted to induce lupus autoantibodies and even clinical SLE (5–7). The interpretation of data regarding the role of IFNγ in SLE has posed challenges. Production of IFNγ by SLE mononuclear cells stimulated in vitro has been reported to be decreased, while serum levels of IFNγ and expression of IFNγ messenger RNA (mRNA) in urinary sediment are increased in patients with active lupus nephritis compared with controls (8–10). IFNγ expression is increased in murine lupus and has been implicated in murine lupus nephritis (11–18). Deficiency of either type I or type II IFN receptors abrogates murine lupus disease, and an IFN-induced gene, Ifi202, has been implicated as a candidate lupus gene in murine studies (19–23).

The prominent role of IFNs in the hierarchy of immune system mediators involved in SLE has been demonstrated recently by data from large-scale analyses of gene expression in peripheral blood mononuclear cells (PBMCs) from patients with active SLE and healthy control subjects (24–28). Expression of a spectrum of genes that constitutes an “IFN signature” is the most significant result from these studies and indicates that either type I or type II IFNs may be dominant among the pathogenic mediators involved in lupus. IFN gene expression has been linked to high disease severity (25) or activity (26). These conclusions relied on microarray data, and the reports did not provide more quantitative data to analyze the specificity of the results for SLE or the roles of type I (α) versus type II (γ) IFNs.

Important differences in the relative contributions of the IFNs to lupus in the various murine models indicate that a definitive assessment of the pathogenic roles of IFNs in human SLE requires the collection and analysis of data from human patients (20, 21, 23). To determine whether expression of the IFN pathway genes in SLE is attributable to type I or type II IFNs, we measured the capacity of a panel of IFN-inducible genes (IFIGs) to be induced in healthy control PBMCs by either IFNα or IFNγ and then quantified those IFIGs in PBMCs from SLE patients, patients with rheumatoid arthritis (RA) or inflammatory uveitis, and healthy donors. Our quantitative data identify the type I (α) IFN pathway as coordinately activated in many SLE patients, in contrast to patients with RA.

PATIENTS AND METHODS

Patients and controls.

Our cohort included 77 SLE patients, 22 disease controls (20 with RA and 2 with active idiopathic inflammatory uveitis), and 28 healthy donors. SLE and RA patients were followed up at the Hospital for Special Surgery and met the American College of Rheumatology (formerly, the American Rheumatism Association) criteria for SLE or RA (29, 30). The 2 patients with uveitis were recruited from a uveitis clinic in New York City. Study subjects signed an informed consent form approved by the Hospital for Special Surgery's Institutional Review Board. The majority of SLE patients (n = 54) were treated with only mild immunosuppression, including prednisone at ≤20 mg/day or hydroxychloroquine. Twenty-three of the SLE patients were treated with pulse glucocorticoids, prednisone at ≥40 mg/day, cyclophosphamide, azathioprine, mycophenolate mofetil, methotrexate, tumor necrosis factor inhibitors, or intravenous IgG. The mean disease activity score of SLE patients, based on the SLE Disease Activity Index (SLEDAI)–2K instrument (31), was 5 (range 0–26).

PBMCs and plasma samples.

Twenty milliliters of heparinized blood was centrifuged and the plasma was removed and stored at −70°C. The blood was then further centrifuged over a Ficoll gradient to obtain PBMCs. These were suspended in 10% fetal bovine serum–containing culture medium and distributed in aliquots of 2 × 106 each, lysed using the RNeasy Mini Kit (Qiagen, Valencia, CA), and stored at −70°C for future RNA isolation. PBMCs were characterized for percentages of monocytes, lymphocytes, and polymorphonuclear leukocytes (PMNs) based on cell scatter on fluorescence-activated cell sorting analysis. The mean percentages of monocytes and PMNs did not differ between SLE patients and disease controls or between SLE patients and healthy donors. The percentage of lymphocytes was significantly lower in the SLE patients than in the healthy donors (65% versus 76%; P = 0.007), but was similar in SLE patients and disease controls (65% versus 70%; P = 0.28).

PBMC stimulation with IFNs or patient plasma.

To identify IFIGs predominantly induced by either IFNα or IFNγ, 2 × 106 PBMCs from healthy donors were cultured in 24-well plates with medium, 1,000 units/ml of IFNα (recombinant human IFNαA [transcribed from the IFNA2 gene]; BioSource International, Camarillo, CA), 1,000 units/ml of IFNγ (R&D Systems, Minneapolis, MN), or 50 μg/ml of polyinosinic–polycytidylic acid (Calbiochem, San Diego, CA) for either 6 hours or 24 hours. To study the effect of donor plasma on gene expression, healthy donor PBMCs were stimulated for 24 hours with 600 units/ml of either IFNα or IFNγ or with 5–10% plasma from SLE patients or controls. Neutralizing antibodies to IFNα (Lot 2007; PBL Biomedical Laboratories, Piscataway, NJ), IFNγ (R&D Systems), or both, as well as isotype control antibodies, were included in some wells. Cells were lysed and stored at −70°C.

Real-time reverse transcriptase–polymerase chain reaction (RT-PCR).

RNA was extracted from each lysate using the RNeasy Mini Kit. One microgram of this RNA was reverse-transcribed to complementary DNA (cDNA) in a 20-μl reaction using SuperScript III RNase H− Reverse Transcriptase, 10 mM dNTP Mix, Oligo(dT)12-18 Primer, Ribonuclease H, and RNaseOUT Recombinant RNase Inhibitor (all from Invitrogen, Carlsbad, CA). Complementary DNA obtained from each sample was diluted 1:200, and 10 μl was amplified in a 25-μl real-time PCR using 0.4 μM of sense and antisense primer and the 2X iQ SYBR Green Supermix (Bio-Rad, Hercules, CA). Primers for all target genes and hypoxanthine guanine phosphoribosyltransferase 1 (HPRT1), a housekeeping gene, were designed using Beacon Designer 2.06 software (Premier Biosoft International, Palo Alto, CA) in conjunction with the DNA mfold 3.1 program to exclude sequences with significant secondary structure (see Table 1 in ref. 32).

Table 1. IFN-inducible genes*
Gene (no. of experiments)Mean relative expression at 24 hours (P)
  • *

    Peripheral blood mononuclear cells from healthy donors were cultured with 1,000 units/ml interferon-α (IFNα) or IFNγ for 24 hours, and RNA was analyzed by quantitative real-time polymerase chain reaction as described in Patients and Methods. Relative expression of the indicated genes compared with their expression in unstimulated cultures (medium [M]) is shown. The mean of 3–6 experiments performed for each gene is indicated.

IFIT1 (5)IFNα/M 10.4 (<0.001)
 IFNγ/M 2.4 (0.08)
 IFNα/IFNγ 4.4 (0.002)
IFI44 (5)IFNα/M 6.8 (<0.001)
 IFNγ/M 3.8 (0.003)
 IFNα/IFNγ 1.8 (0.03)
PRKR (5)IFNα/M 10.9 (0.001)
 IFNγ/M 5.4 (0.07)
 IFNα/IFNγ 2 (0.02)
OAS3 (3)IFNα/M 8.0 (0.02)
 IFNγ/M 2.6 (0.06)
 IFNα/IFNγ 3.1 (0.11)
GBP1 (5)IFNα/M 3.7 (0.008)
 IFNγ/M 16.6 (0.003)
 IFNγ/IFNα 4.5 (0.006)
IRF1 (5)IFNα/M 2.5 (0.01)
 IFNγ/M 6.4 (<0.001)
 IFNγ/IFNα 2.5 (0.001)
SERPING1 (5)IFNα/M 2.3 (0.002)
 IFNγ/M 7 (0.001)
 IFNγ/IFNα 3 (0.003)
CXCL9 (6)IFNα/M 0.7 (0.13)
 IFNγ/M 1,195 (0.004)
 IFNγ/IFNα 1,738 (<0.001)
CXCL10 (5)IFNα/M 8.3 (0.29)
 IFNγ/M 269 (0.002)
 IFNγ/IFNα 32 (<0.001)
PSMB8 (3)IFNα/M 1.74 (0.11)
 IFNγ/M 3.3 (0.01)
 IFNγ/IFNα 1.9 (0.05)
PSMB10 (3)IFNα/M 2.5 (0.001)
 IFNγ/M 4.3 (0.07)
 IFNγ/IFNα 1.7 (0.34)
GPR105 (3)IFNα/M 5.85 (0.4)
 IFNγ/M 6.79 (0.4)
 IFNγ/IFNα 1.2 (0.5)
FCGR1A (3)IFNα/M 1.2 (0.3)
 IFNγ/M 1.3 (0.8)
 IFNγ/IFNα 1.0 (0.7)

Gene amplifications for each sample were performed in duplicate using the iCycler IQ Real-Time Detection System (Bio-Rad). Standard curves for each gene were generated in all experiments using a cDNA from a sample known to contain high levels of that gene, and the efficiencies of each reaction were obtained. Melting curve analysis was performed after all PCR procedures to ensure specificity of amplification (33). In order to calculate relative expression of a PBMC lysate sample for any given target gene, each sample was amplified with primers for both the target gene and HPRT1 in separate wells. In addition, a reference sample was included in each PCR plate to provide a basis for normalization across experiments. Each sample's threshold cycle values for the target and housekeeping genes were subtracted from the corresponding values of the reference sample. The differences were then used as exponents with the base equal to 1 plus the value of the efficiency of that PCR expressed as a decimal (e.g., 1.98 for efficiency of 98%) (34). Finally, the target gene values were divided by the housekeeping gene values for each sample, and the result was the relative expression value for each unknown sample. A detailed description of the method has been reported (35).

Calculation of the IFN scores.

The mean (M) and SD of each IFIG for the group of healthy donors (MHD and SDHD) were used to calculate that gene's expression score for each study subject, defined as the number of SDHD above the MHD. Cumulative IFNα and IFNγ scores, representing the sum of the scores for each of 3 genes preferentially induced by IFNα and for each of 3 genes preferentially induced by IFNγ (see Results), were derived for each subject. We considered an IFNα score to be high if it fulfilled 1 of the 2 following criteria: 1) expression of at least 2 of the 3 IFNα genes at a level ≥2 SDHD above the MHD; 2) expression of a single IFNα gene at a level ≥4 SDHD above the MHD. The IFNγ score was considered high or low based on analogous criteria for the 3 IFNγ genes. The 3 IFNγ-regulated genes selected for inclusion in the IFNγ score (IRF1, GBP1, and SERPING1) had levels of in vitro induction by IFNγ (6.4–16.6×) and IFNα (2.3–3.7×) that were similar to those of the 3 IFNα-regulated genes selected for the IFNα score (IFIT1, IFI44, and PRKR) (2.4–5.4× and 6.8–10.9×, respectively), except in the inverse order, permitting a comparison of the relative exposure to these two cytokine classes in vivo (Table 1).

Statistical analysis.

The data distributions of all variables were examined. Two-group comparisons of continuous data were assessed using t-tests. One-way analysis-of-variance was used to compare 3 or more groups, with Dunnett's test used to do post hoc comparisons with a control. In some cases, variables were log-transformed to convert the data to a Gaussian distribution. Groups of ordinal data were compared using Kruskal-Wallis, paired Wilcoxon, and Mann-Whitney analyses.

RESULTS

Identification of genes preferentially induced by IFNα or IFNγ.

Our previous studies, along with those of others, used microarray analysis of SLE PBMCs to identify the pattern of gene expression characteristic of that disease (24–28). Statistical analysis showed that PRKR, a classic IFNα-induced gene, was most significantly overexpressed in SLE, while genes inducible by IFNγ were also identified by the statistical algorithms as being increased in SLE (25, 36, 37). In view of the abundant data implicating both IFNα and IFNγ in murine and human lupus pathogenesis, it was important to provide quantitative data to determine the relative roles of these distinct but interacting cytokine pathways.

To begin to address this issue, we first characterized a panel of IFIGs for their relative induction by IFNα compared with IFNγ. Quantitative real-time PCR was used to test 13 candidate IFIGs, the selection of which was based on reported microarray data, for preferential induction in healthy donor PBMCs by either IFNα or IFNγ (Table 1 and Figure 1) (25, 38, 39). Induction levels for both the 6-hour and 24-hour time points were measured, showing similar trends, and those at the 24-hour time point were used for our calculations. Among those genes tested were 4 IFIGs (of many good candidate genes) that were predicted to be preferentially induced by IFNα. Three of those showed significantly greater induction by IFNα than by IFNγ (IFIT1, IFI44, and PRKR, all shown in microarray studies to be overexpressed in SLE [24–28]) and were selected for the IFNα score calculation, while the fourth (OAS3) also showed a preference for induction by IFNα. The ratio of the expression level induced by IFNα to the expression level induced by IFNγ ranged from 1.8 to 4.4 for these 4 genes.

Figure 1.

Identification of genes preferentially induced by interferon-α (IFNα) or IFNγ. Two million peripheral blood mononuclear cells from 5 healthy donors (for PRKR and SERPING1) or 6 healthy donors (for CXCL9) were incubated for 6 hours and 24 hours with medium (M), 1,000 units/ml of IFNα, 1,000 units/ml of IFNγ, or 50 μg/ml of polyinosinic–polycytidylic acid (P[IC]). Cells were then lysed and used for RNA isolation, reverse transcription, and amplification by quantitative real-time polymerase chain reaction. Mean and SD relative expression (R.E.) compared with cells cultured with medium is shown for a representative gene for each of 3 patterns of IFN induction: PRKR, predominantly induced by IFNα (A); SERPING1, predominantly induced by IFNγ, with some IFNα effects (B); and CXCL9, highly induced by IFNγ (C). P values are shown for comparisons between IFNα and IFNγ stimulation at 24 hours. Each panel represents 5–6 separate experiments.

Of 9 IFIG candidates for preferential regulation by IFNγ, 6 were confirmed: guanylate binding protein 1 (GBP1); serine proteinase inhibitor, clade G, member 1 (SERPING1); interferon regulatory factor 1 (IRF1); chemokine (C-X-C motif) ligand 9 (CXCL9); chemokine (C-X-C motif) ligand 10 (CXCL10); and proteasome (prosome, macropain) subunit, beta type, 8 (PSMB8). The ratio of the expression level induced by IFNγ to the expression level induced by IFNα was highly variable, ranging from 1.9 to 1,738 for these 6 genes. CXCL9 was strongly induced by IFNγ and not at all induced by IFNα at the 24-hour time point (Table 1 and Figure 1), while the other 5 genes preferentially induced by IFNγ had an additional, although more modest, response to IFNα. Of these, GBP1 and SERPING1 have been shown to be overexpressed in SLE PBMCs in reported microarray data (25, 26) and were selected for the IFNγ score calculation, along with IRF1, because they had levels of induction by the 2 IFN types that were comparable with those of the IFIGs selected for the IFNα score, except in the inverse order (Table 1).

SLE PBMCs express IFIGs preferentially induced by IFNα, rather than by IFNγ.

To determine which IFN type is responsible for the IFN signature that has been observed in microarray studies, and to determine whether that signature is specific for SLE, quantitative gene expression data were collected from all 77 SLE patients, all 22 disease controls (20 with RA and 2 with autoimmune uveitis), and all 28 healthy donors. As a group, the SLE population had low-to-moderate disease activity, with a mean SLEDAI-2K score of 5. At the time of the study, 54 of the SLE patients were being treated with prednisone at ≤20 mg/day and/or hydroxychloroquine, and only 23 patients were receiving high doses of immunosuppressive therapy (see Patients and Methods).

Quantitative real-time PCR was used to compare the expression of 7 IFIGs in all SLE patients and controls. These included 3 IFIGs that were preferentially induced by IFNα (PRKR, IFIT1, and IFI44), 3 that were preferentially induced by IFNγ but that also showed an IFNα response (IRF1, GBP1, and SERPING1), and CXCL9, the gene that was nearly exclusively induced by IFNγ. The expression of 2 of the IFNα-induced genes (IFI44 and PRKR) was increased in SLE PBMCs compared with both control PBMC populations, while the expression of IFIT1 showed a similar trend (Table 2). In contrast, the expression of all 3 genes preferentially induced by IFNγ (IRF1, GBP1, and SERPING1) was equivalent in the 2 groups of patients and in the healthy donors. Expression of CXCL9, the gene exclusively induced by IFNγ at 24 hours, was also equivalent in SLE patients and healthy donors, but was actually lower in the SLE patients than in the disease controls (P < 0.01) (Table 2).

Table 2. Relative expression of IFN-inducible genes in SLE and control PBMCs*
 Source of PBMCs
SLE patients (n = 77)Disease controls (n = 22)Healthy donors (n = 28)
  • *

    Values are the mean ± SD. Two million peripheral blood mononuclear cells (PBMCs) freshly isolated from patients with systemic lupus erythematosus (SLE), disease controls, and healthy donors were lysed and used for RNA isolation, reverse transcription, and amplification by quantitative real-time polymerase chain reaction. Relative expression is shown for 3 genes preferentially induced by interferon-α (IFNα) (PRKR, IFIT1, and IFI44), 3 genes preferentially induced by IFNγ (IRF1, GBP1, and SERPING1), and a gene highly induced by IFNγ without a discernable effect of IFNα at 24 hours (CXCL9).

  • P < 0.01 versus disease controls and healthy donors.

  • P < 0.05 versus disease controls and healthy donors.

  • §

    P < 0.01 versus disease controls.

Genes preferentially induced by IFNα   
 PRKR3.46 ± 2.371.93 ± 0.942.11 ± 0.99
 IFIT13.78 ± 3.552.42 ± 1.872.52 ± 1.28
 IFI444.38 ± 4.382.17 ± 1.052.42 ± 1.08
Genes preferentially induced by IFNγ   
 IRF11.54 ± 0.771.84 ± 0.781.62 ± 0.69
 GBP12.43 ± 1.832.52 ± 1.352.37 ± 1.13
 SERPING11.53 ± 1.401.48 ± 0.761.63 ± 0.83
Gene highly induced by IFNγ   
 CXCL91.07 ± 0.76§2.11 ± 2.341.22 ± 0.80

Notably, PRKR mRNA expression in all SLE patients correlated with expression of the other 2 predominantly IFNα-induced genes, but not with any of the IFNγ-induced genes, suggesting coordinate activation of genes in the type I IFN pathway (Figure 2A). Correlation of expression of 2 IFNα-induced genes (IFI44 and IFIT1) in SLE PBMCs is shown in Figure 2B. CXCL9 expression did not correlate with expression of any of the other IFIGs, consistent with its distinct pattern of induction by IFNγ alone (Figure 2A). With the exception of IRF1, which only showed a weak correlation with the other IFIGs, relative expression levels of the remaining IFIGs correlated with one another (Figure 2A). Possible interpretations of these data are that with the exception of CXCL9, each of the so-called IFNγ-induced genes is variably induced by IFNα present in vivo, or that exposure to IFNα in vivo has primed SLE cells for production of and response to IFNγ (40).

Figure 2.

Correlations in expression of IFN-inducible genes (IFIGs) in peripheral blood mononuclear cells from patients with systemic lupus erythematosus (SLE). A, In the total SLE population, the relative expression of each of 7 IFIGs was plotted against the relative expression of every other IFIG, and the degree of correlation was determined by linear regression. Data are shown as R2 values. B, The relative expression of IFI44 in all SLE patients was plotted against that of IFIT1. See Figure 1 for other definitions.

IFNα and IFNγ IFIG expression scores were calculated for all SLE patients and disease controls to derive an indication of the extent of type I or type II IFN pathway activation. CXCL9 was not included among the IFNγ-induced genes selected for calculation of the IFNγ score, since its induction level by IFNγ was far higher than the induction levels of all the other 6 IFIGs by the corresponding IFN and would have unduly influenced the score. SLE patients had higher IFNα scores, but not higher IFNγ scores, than either control group (Figures 3A and B). Within the SLE population, IFNα scores were higher than IFNγ scores (mean ± SD 4.21 ± 6.96 versus 0.95 ± 2.50; P < 0.0001). More specifically, 32 of the 77 SLE patients had high IFNα scores, while only 8 had high IFNγ scores. Of those, none showed a level of CXCL9 expression that was ≥2 SD above the mean of the healthy donors. Paired analysis of IFN scores showed that even for the 8 SLE patients with high IFNγ scores, the IFNα scores were higher (Figure 3C). In view of the low CXCL9 expression in these SLE patients, we conclude that even the high IFNγ scores are likely to reflect effects of IFNα, since GBP1, SERPING1, and IRF1 are all modestly induced by IFNα, while CXCL9 expression is relatively independent of IFNα effects.

Figure 3.

Interferon-α (IFNα) and IFNγ scores in SLE patients, disease ontrols (DC), and healthy donors (HD). A and B, After IFN scores were calculated for all individuals in the study, IFNα and IFNγ scores, respectively, were plotted for the 3 groups. Horizontal bars indicate mean values for each group. P values shown are for comparisons between SLE patients and either disease controls or healthy donors. C, Paired IFNα and IFNγ scores are shown for the 8 SLE patients with high IFNγ scores. Even in these individuals, IFNα scores were significantly higher than the corresponding IFNγ scores. D, The mean expression of each of 8 IFIGs tested (PRKR, IFIT1, IFI44, IRF1, GBP1, SERPING1, CXCL9, and CXCL10) in the SLE subgroup with high IFNα scores correlated highly with the inherent capacity of those IFIGs to be induced by IFNα in peripheral blood mononuclear cells (PBMCs) from healthy donors, as measured by the mean expression induced by IFNα at the 24-hour time point. Each square represents an IFIG. See Figure 2 for other definitions.

Coordinate activation of the IFNα pathway.

To further support the hypothesis that the pattern of IFIG expression observed in SLE patients reflects coordinate activation of the IFNα pathway in vivo, we focused our analysis on those patients with high IFNα scores. The mean expression of each of 8 IFIGs tested (PRKR, IFIT1, IFI44, IRF1, GBP1, SERPING1, CXCL9, and CXCL10) in the SLE subgroup with high IFNα scores correlated highly with the inherent capacity of those IFIGs to be induced by IFNα in PBMCs from healthy donors, as measured by the mean expression induced by IFNα at the 24-hour time point (Figure 3D). Taken together, these data indicate that the type I IFN pathway, with IFNα as the prototype inducer, is coordinately activated in many SLE patients, in contrast to the disease controls.

Activation of IFNα-induced genes by SLE plasma.

Studies by others have indicated that in spite of the IFN gene expression signature seen on microarray analysis, it is difficult to detect IFNα protein in sera from most SLE patients (25, 26). Nonetheless, 3 of 5 SLE plasma samples, when incubated with healthy donor PBMCs at a concentration of ≤1/10 (volume/volume), induced IFNα target gene expression, while 2 RA plasma samples and 2 healthy donor plasma samples did not induce IFIGs (results of a representative experiment are shown in Figure 4). Neutralizing antibodies to IFNα, but not to IFNγ, fully inhibited the SLE plasma activity. One of the 2 SLE plasma samples that did not stimulate IFIG expression was from a patient who had recently been treated with intravenous pulse glucocorticoids, a treatment that may deplete IFNα-producing cells, while the other inactive plasma sample was from a patient with inactive disease and a SLEDAI score of 0. Additional analyses of larger numbers of plasma or serum samples, in experiments that include higher plasma concentrations and with full consideration of SLE clinical parameters, will be required to determine the characteristics of SLE patients whose plasma can activate type I IFIGs.

Figure 4.

Demonstration of IFNα activity in plasma from patients with systemic lupus erythematosus (SLE). To test plasma for functional IFNα activity, 3 million peripheral blood mononuclear cells from healthy donors were incubated with medium, 600 units/ml of IFNα, or 5–10% plasma for 24 hours in the presence or absence of neutralizing antibodies to one or both IFNs. Cells were then lysed and used for RNA isolation, reverse transcription, and amplification by quantitative real-time polymerase chain reaction. Relative expression of PRKR, compared with no stimulation, in a representative experiment is shown. Of 4 additional SLE plasma samples tested, 2 showed induction of PRKR and/or IFIT1 similar to that demonstrated here. None of 2 plasma samples from patients with rheumatoid arthritis (RA) or 2 plasma samples from healthy donors induced either of those genes. iso = isotype control antibody (see Figure 1 for other definitions).

DISCUSSION

In this study we have investigated the basis for the dominant gene expression signature that has been observed in microarray analyses of SLE PBMCs by our group and others. We confirmed, using the quantitative real-time RT-PCR approach, that nearly half of our SLE cohort, representing a patient group with low-to-moderate disease activity, has increased expression of a set of IFIGs in their unmanipulated PBMCs. Increased expression of these IFIGs is not a general characteristic of patients with immune-mediated inflammatory diseases, since a group of disease controls (mostly RA patients) did not have increased expression of the IFIGs compared with the group of healthy donors. CXCL9, an IFIG almost exclusively induced by IFNγ, was increased in the group of disease controls compared with both SLE patients and healthy donors (P < 0.01), suggestive of a distinct in vivo cytokine microenvironment in RA patients. It is unlikely that activation of the IFN pathway is unique to SLE. Patients with type I autoimmune diabetes are among those who may show a similar pattern of gene activation, perhaps providing important clues to common proximal stimuli for disease pathogenesis in certain autoimmune disorders (4).

Importantly, our analysis of gene expression in a large group of SLE patients demonstrates that the IFIGs overexpressed in SLE are those that are predominantly induced by IFNα. PRKR and IFI44 were significantly increased in expression in the SLE patients and were also preferentially induced by IFNα in comparison with IFNγ. PRKR was also identified in our microarray studies as differentially expressed in SLE PBMCs compared with PBMCs from disease controls (24, 36, 37). Experiments in which IFIG expression in healthy donor PBMCs cultured with SLE patient plasma was specifically inhibited by neutralizing antibody to IFNα strongly support the role of that particular cytokine in the IFIG expression signature. However, we cannot rule out a contribution of other type I cytokines, including IFNβ or IFNλ, to the activation of this pathway in some patients (41). Experiments in progress in our laboratory are quantifying the relative expression of the various type I IFN isoforms in SLE patients. In addition, infectious agents or endogenous factors, such as viruses, unmethylated CpG DNA, or single- or double-stranded RNA, could be more proximal mediators of type I IFN production and IFIG activation.

The approach used in measuring IFIG expression, quantitative real-time PCR, allowed quantitative comparisons among the SLE patients and controls. Forty-two percent of SLE patients showed activation of the IFNα pathway, as assessed by a high calculated IFNα score, compared with 9% (2 of 22) in the group of disease controls and none in the group of healthy donors. Additional analysis of extensive clinical and laboratory data has determined that those SLE patients with type I IFN pathway activation can be distinguished from those who do not demonstrate activation of those IFIGs in terms of disease activity and serologic profile (Kirou KA, Lee C, Crow MK: unpublished observations). Of particular interest, our data indicate that expression of IFNα-induced genes may identify those patients who produce autoantibodies specific for RNA binding proteins. The effect of medical therapy on activation of the IFNα pathway is also of great interest. Those patients who received intravenous high-dose glucocorticoid therapy showed low-level IFNα-induced gene expression (Kirou KA, Lee C, Crow MK: unpublished observations), as has been previously reported by others (26).

In contrast to the activation of IFNα genes in SLE, most patients did not demonstrate increased expression of a panel of genes preferentially induced by IFNγ. High IFNγ scores were seen in only 8 SLE patients, and in general were lower than the corresponding IFNα scores (Figure 3C). Moreover, the gene most dramatically induced by IFNγ in preference to IFNα, CXCL9 (MIG), was comparable in expression in the SLE patient and healthy donor groups and actually higher in the group of disease controls. Additionally, the good correlation between some IFNγ genes such as GBP1 and SERPING1 and some IFNα genes (IFIT1 and IFI44), as shown in Figure 2A, might reflect a similar, but weaker, IFNα induction effect in the so-called IFNγ genes (Table 1 and Figure 1B). This interpretation is supported by the strong correlation between the inherent capacity for a given IFIG to be induced by IFNα in PBMCs from healthy donors and the mean expression of that gene in SLE patients with activation of the IFNα pathway (Figure 3D).

In other words, the entire panel of IFNα-responsive IFIGs studied is coordinately activated in our SLE patients, supporting the hypothesis of a broad in vivo effect of IFNα (or an endogenous mediator with downstream targets identical to those of IFNα) on the immune system in SLE. This interpretation is made with full recognition of the complex panoply of factors that will affect IFIG expression in vivo, including variable requirements and sensitivity for induction of one IFIG compared with another, differential kinetics of IFIG expression, and expression of different IFIGs by distinct cell types with variable distribution among body compartments.

Our results show that approximately half of the SLE patients have evidence of full activation of the IFNα pathway. It is unclear whether genetic, environmental, immunologic, or stochastic factors determine which patients will have full IFNα pathway activation, but it is likely that all of these contribute. Once induced, the IFNα pathway could contribute to disease by several mechanisms, including promotion of apoptosis leading to increased availability of autoantigens, maturation of antigen-presenting cells, development of Th1 responses, and inhibition of activation-induced T cell death (2, 3, 41–48). IFNα may also promote production of pathogenic autoantibodies by direct and indirect effects on B cells, resulting in differentiation and immunoglobulin class switching to IgG and IgA isotypes (49, 50).

Although our data do not support the hypothesis that the IFNγ pathway is activated in SLE PBMCs, we cannot exclude a role for IFNγ-induced gene expression in certain SLE patients or at sites of inflammation, such as the kidney. IFNγ has been implicated in lupus pathogenesis, in both murine and human studies. Most convincingly, it has been found in kidneys with active lupus nephritis (9, 10, 13, 15, 18). We suggest that IFNα is produced early in the evolution of SLE, with activation of the IFNα pathway associated with clinical disease. Induction of Th1 responses and IFNγ production, with activation of IFNγ target genes occurring in a subset of patients, may culminate in more advanced organ inflammation. Analysis of gene expression in additional patients, with parallel analysis of PBMCs and tissue biopsy material, will be required to address this hypothesis.

Research using murine lupus models has focused on IFNγ as a pathogenic cytokine that contributes to target organ inflammation and damage (11–19). In contrast, the results of our study of SLE patients support the hypothesis that type I interferon, IFNα, plays a dominant role in SLE and suggest that assay of IFNα pathway genes may ultimately prove to be a useful approach for measuring lupus disease activity in some patients. Further characterization of immune function in those with activation of the IFNα pathway should provide new insights into the proximal triggers of lupus initiation or exacerbation. Finally, in view of the pleiotropic effects of type I IFN on immune functions that are consistent with immune system alterations characteristic of SLE, it is plausible that therapeutic targeting of the IFNα pathway might benefit those SLE patients with high expression of IFNα-induced genes.

Acknowledgements

The authors thank the patients, healthy donors, study coordinators, and rheumatologists at the Hospital for Special Surgery who participated in this study. We particularly appreciate the contributions of Adrienne Davis and Drs. Jane Salmon, Stephen Paget, Michael Lockshin, Lisa Sammaritano, and Michael Samson.

Ancillary