Discussion
In this study, we confirmed the presence of an IFN-inducible gene expression signature (IS) in the peripheral blood of patients with SLE and developed a novel and simplified qPCR-based biomarker (ISM) test to quantitate the IS. Importantly, we showed that the magnitude of the ISM biomarker was highly similar between PBMCs and whole blood RNA samples obtained from the same subjects. At baseline, patients with SLE from four different clinical trials consistently fall into one of two subsets—ISM-High and ISM-Low and the bimodal distribution of patients with SLE by the ISM agrees with previous reports.1 ,10
The bimodal distribution expression of the IRG in ISM-Low and ISM-High patients with SLE is a distinct feature of this biomarker. High-ISM patients with SLE had markedly elevated IRG expression, while patients in the ISM-Low subset had a distribution of gene expression that overlapped with the predominantly unimodal ISM measured in healthy subjects, despite the presence of active disease. The relative distribution of patients who are ISM-High or ISM-Low varied, in part, with disease activity based on BILAG or SELENA-SLEDAI measures, but this appears to be a consequence of the trial inclusion criteria. Thus, ISM status across SLE trial populations varied from an incidence of ∼50% ISM-High scores in a mild disease population in a Phase I study16 to ∼75% in moderate-to-severe SLE disease populations in Phase II/III studies that represent patients with ERL and LN.
Despite an apparently increased incidence of ISM-High patients in Phase II/III trials in populations with moderate-to-severe disease activity compared with Phase I studies in populations with mild disease activity, the ISM status within each trial's cohorts did not correlate with differences in disease activity. Our observations are based on the consistent use of a platform—the qPCR-based ISM—across different trial populations, in contrast to reports of IS associations with increased SLE disease activity based on differing methods applied across different cohorts to measure the IS.1 ,10 ,14 For example, in the Phase II ROSE study with patients with moderate-to-severe SLE, ISM status did not correlate with differences in mean measures of disease activity assessed by BILAG or SELENA-SLEDAI.19 Further, the skewed incidence of ISM-High patients in moderate-to-severe disease was fairly similar in the ERL and LN populations, indicating that the ISM distribution is not broadly specific with end-organ activity in most patients with SLE. In the EXPLORER17 and ROSE19 studies, we did not observe consistent correlations between mucocutaneous disease manifestations and ISM status, and we observed only a modest trend in elevated musculoskeletal disease manifestations in ISM-Low compared with ISM-High patients across the four trials. Of note, we did observe increased CLASI damage scores, but not CLASI activity scores, in the ROSE trial where this instrument was used that suggests the possibility of an association of the ISM biomarker with aspects of cutaneous disease. We also observed a trend for elevated FACIT fatigue scores in ISM-High patients in the ROSE trial. The possibility still remains, however, that there are distinct clinical SLE pathotypes within the ISM subsets that have not yet been identified. We did observe that the ISM-High subset was slightly younger overall across the trials and, in at least the ERL populations, had a lower frequency of white patients, consistent with previous observations.10
ISM status does correlate with serological differences between the two subpopulations across multiple trial cohorts with extrarenal SLE. The presence of characteristic SLE autoantibodies, hypocomplementaemia, cytokine production and serum IFN bioactivity varied between ISM-High and ISM-Low patients with lupus. These findings are consistent with previous reports1 ,10 ,16 and underscore the role of type I IFNs in systemic immune activation, including extrafollicular differentiation of autoreactive B cells to antibody-secreting plasmablasts via regulation of toll-like receptors (TLRs) and increased survival via BAFF-induced signalling.6 Thus, the ISM could be a biomarker of systemic immune activation that comanifests with serological abnormalities. Indeed, our multivariate analysis confirmed independent association of ISM magnitude with T cell counts, autoantibodies, cytokine levels and disease duration parameters.
We found no significant differences between the ISM-Low and ISM-High subpopulations in autoantibody presence or titre in patients with LN, and observed comparable levels of hypocomplementaemia in these subpopulations. So, although immune-serology differences exist between ISM-High and ISM-Low patients with ERL, additional drivers of the IS likely exist in patients with LN, giving rise to a bimodal ISM distribution. This suggests that the regulation of the IRGs could be due in part to an amplification loop where, upon reaching a threshold, genes could be upregulated by additional upstream signals. These signals may include, but are not limited to, nucleic acid-containing IC activation of endosomal TLR7 and TLR9 nucleic acid sensors22 ,23 or activation of cytosolic nucleic acid sensors such as the retinoic acid-inducible gene I-like receptors, nucleotide oligomerisation domain-like receptors, and cyclic GMP-AMP synthase, which lead to type I IFN expression.6 ,24 ,25
Although our analysis was not designed to address the stability of the ISM on a particular background of immunosuppressants, it appears that the most frequently used immunosuppressants (ie, hydroxychloroquine and oral corticosteroids) in lupus management do not significantly impact the ISM status of patients with SLE. In contrast, others have shown that high-dose, pulse intravenous corticosteroids markedly and transiently attenuate the IS in patients with SLE.13 ,26 Further, blockade of IFN-α with rontalizumab in patients with ERL19 only leads to a partial reduction in ISM score by 2–4 weeks of treatment. This is consistent with observations in patients with SLE treated with a different anti-IFN-α antibody, sifalimumab.27
One potential contributing factor to this partial ISM decrease could include suboptimal dosing of patients with anti-IFN-α antibodies, and treatment of patients with higher amounts of this therapy could result in a greater magnitude of decrease of the ISM score. It is also important to consider the complexity of the genome-wide gene expression profile, and the potential limitations of our approach to derive a simplified biomarker that integrates the magnitude of a large number of genes. Indeed, a recent study using modular transcriptional analysis has demonstrated that different a priori-defined IFN-inducible modules have differential associations with disease activity such as SELENA-SLEDAI scores, serology and presence of flare.28 Further, the different IFN-inducible modules had varying degrees of longitudinal stability, and likely reflected differential contributions of IFN-β and IFN-γ in addition to IFN-α. Clearly, therefore, the IS in patients with SLE is driven by factors beyond IFN-α signalling and may include overlapping Jak-Stat signalling mechanisms used by other cytokines29 and/or sustained IC derived from pre-existing plasma cells. In this regard, it has recently been reported that blockade of the Lymphotoxin/LIGHT pathway using baminercept, a lymphotoxin-β receptor-immunoglobulin fusion protein, reduced the IFN signature in the blood of patients with rheumatoid arthritis with elevated baseline signatures30 suggesting a role of the Lymphotoxin/LIGHT pathway as an upstream modulator of IFNs in RA and potentially SLE. Furthermore in this study, patients with RA and SLE had lymphopenia associated with elevation of blood IFN signature, and treatment of patients with RA with baminercept subsequently increased lymphocyte counts. Consistent with this observation, it was noteworthy that the ISM-High patients within the trials examined in this study also had reduced circulating leucocyte counts including lymphocytes, monocytes, neutrophils and eosinophils as compared with ISM-Low patients. These data are consistent with a role for IFN signalling in retaining immune cells in tissues during inflammatory disease and therefore contributing to lymphopenia. Additional mechanistic studies are warranted to investigate upstream signals that can specifically induce genes also induced by type I IFNs.
In conclusion, we have characterised the ISM as a surrogate of the peripheral blood global IS, first identified in patients with SLE.10–12 The method identifies distinct ISM-Low and ISM-High SLE populations, and has the potential to provide a useful approach to identify and stratify the heterogeneous SLE population in the context of randomised control clinical trials and clinical practice. The correlation of baseline ISM status and clinical response has recently been evaluated in the ROSE Phase II randomised, controlled trial of the safety and efficacy of rontalizumab in extrarenal lupus.19