Therapeutic Responses and Predictors of Low-Dose Interleukin-2 in Systemic Lupus Erythematosus

Objective. Interleukin-2 (IL-2) is effective and well tolerated in patients with systemic lupus erythematosus (SLE). However, patient response to IL-2 therapy varies. Therefore, biomarkers are needed to eciently identify patients who may respond well to IL-2 treat ment. We investigated clinical and immunological biomarkers to predict low-dose IL-2 responses. Methods. A pooled post-hoc analysis was performed in SLE patients who received low-dose IL-2 treatment in two clinical trials. Factors predicting responses in clinical and T-cell subset changes were evaluated by logistic regression. Good response (GR) and poor response (PR) were dened according to whether patients achieved or did not achieve an SLE Responder Index-4 (SRI-4), respectively. Results. A good response at 68% of patients was achieved in the lower Treg group, compared to 0% in the higher Treg group. A good response at 68% was achieved in patients with lower Treg, compared to 0% in patients with higher Treg. In comparison to PR, GR was more strongly associated with low Treg proportions at baseline (12.85±6.07% vs. 9.43±2.82%, P<0.01). There were more patients with skin rash in the GR group than in the PR group (68.75% vs. 30.77%, P=0.042). Multivariate analysis showed that low Treg proportions and skin rash presence were both independently associated with GR to low-dose IL-2 treatment. A nomogram to identify GR probability exhibited a clear discrimination (concordance index, 0.812; 95% condence interval, 0.64-0.97). Based on the area under the receiver operating characteristic (ROC) curve (AUC) of 0.813, the specicity of a low regulatory T cells (Tregs) proportion ( ≤ 13.35%) plus skin rash to predict GR to IL-2 therapy was 100%, with a sensitivity of 68.75%. Conclusion. Treg proportion skin rash indicate GR low-dose absolute number of Tregs. IFN-α, interferon-α. IL-2, interleukin-2. IL-21, interleukin-21. IFN-γ, interferon-γ. IL-10, interleukin-10. TGF-β, transforming growth factor-β.. ANA, Anti-nuclear antibody. Sm, Anti-Sm antibody.


Introduction
Systemic lupus erythematosus (SLE) is characterized by a breakdown in immune tolerance, leading to auto-reactive immune responses and consequential tissue and organ damage. Regulatory T cells (Tregs) suppress autoimmunity and regulate innate and adaptive immune responses. Treg impairment is associated with a loss of immune tolerance and autoimmunity. Low-dose interleukin-2 (IL-2) has been studied as an immunoregulatory agent that selectively promotes Treg expansion, showing promising e cacy in the treatment of SLE [1][2][3][4]. However, these clinical studies showed that a substantial proportion of SLE patients had a sub-optimal response to low-dose IL-2 and failed to achieve an SLE Responder Index-4 (SRI-4) at the end of their IL-2 treatment period. To date, the prediction of response to IL-2 therapy is not well studied in SLE. It was found that increase of the CD25hi-expressing cells among CD3 + CD4 + Foxp3 + CD127lo Tregs might be suggestive of clinical responsiveness to low-dose IL-2 treatment [3]. However, indicators evaluated after the therapy were limited in their ability to guide therapeutic decision. In practice, baseline biomarkers are more valuable to guide treatment choices. In this study, we analyzed multiple clinical and laboratory parameters to identify baseline markers to predict response to IL-2.

Methods
Participants This study was to analyze the therapeutic responses to low-dose IL-2 in a randomized controlled trial (RCT) (NCT02465580) of 29 SLE patients as primary cohort. Then, the prediction model was validated by using an independent cohort with 23 patients (NCT02084238). Full details of study designs and inclusion/exclusion criteria for each completed study have previously been published [1,2].
Studies were conducted in accordance with the Declaration of Helsinki, the International Conference on Harmonization Guidelines for Good Clinical Practice, and local regulations.
All the patients were treated for the rst 12 weeks which included three treatment cycles with IL-2 along with standard treatment, and followed up for further 12 weeks without IL-2. Patients were evaluated at screening and every 4 weeks up to week 24. The primary e cacy endpoint was the attainment of an SRI-4 at week 12 in both trials.
The primary cohort for the development of the nomogram was the low-dose IL-2 group in trial 1, and the independent validation cohort comprised the patients with IL-2 immunological analyses in IL-2 group in trial 2. The performance of the internally validated nomogram was tested again in the validation cohort.
The logistic regression formula formed in the primary cohort was applied to all patients in the validation cohort, and the total points were calculated for each patient. In this cohort, logistic regression was then performed by using the total points as a factor. Finally, the concordance index (c-index) was derived on the basis of the regression analysis.
Statistical analysis Descriptive statistics are provided as the mean ± the standard deviation (SD), median (interquartile range (IQR) (p25 to p75)) or n (%) depending on data distribution. Differences in baseline characteristics between SRI-4 vs. non-SRI-4 groups were compared using a Student's t-test (or, if not normally distributed, the Wilcoxon rank-sum test) and the χ2 test for continuous and categorical data, respectively. Baseline items showing signi cant associations in simple (univariable analysis) models quali ed for further analysis in multiple logistic regression models for the assessment of priority, independence and confounding potentiality. We developed models for the prediction of SRI-4 responses using multiple imputation multivariable logistic regression. We tested the accuracy of the clinical model by comparison of the area under the receiver operator characteristic (ROC) curve (AUC) in both the primary cohort and the validation cohort. Statistical analyses were performed using SPSS v.22.0 or R v.3.6.3 software. Two-sided P values < 0.1 were considered statistically signi cant.

Characteristics of patients and comparisons of responses to IL-2 therapy
Characteristics of the patients in the primary and validation cohorts are shown in Table S1. Patients were divided into those with good response (GR) and those with poor response (PR) according to whether they achieved or did not achieve an SRI-4, respectively, at week 12. Sixteen patients (55.17%) had GR to lowdose IL-2 treatment, whereas 13 patients had PR.

Low Treg Proportion Predicts Clinical Response To Il-2 Treatment
We divided SLE patients who received IL-2 treatment into two groups, i.e. those with a low or high proportion of Tregs, according to a cutoff value of 13.35%, calculated by ROC with an AUC of 0.73 (Fig. 1A). The subsequent analysis suggested that patients with a low proportion of Tregs had a three times higher probability of achieving GR to IL-2 treatment than patients with a high proportion of Tregs (P = 0.011) ( Table 2). A good response at 68% of patients was achieved in the low Treg group at week 12, compared to 0% in the high Treg group. In the follow-up period, 88% of patients in the low Treg group achieved GR at week 24 at end of the follow-up, compared to 0% in the high Treg group (Fig. 1B). The proportion of patients who achieved complete renal remission (CR) was signi cantly higher in the low Treg group than in the high Treg group at week 12 (33.33% vs. 0%) and week 24 (40% vs. 0%) ( Fig. 2A). The low Treg group had a lower 24 h urine protein excretion (24 h-UPE) at week 24 than the high Treg group (0.83 ± 0.51 vs. 0.94 ± 0.49, P = 0.093) (Fig. 2B). The proportion of patients achieved 50% steroid tapering was also signi cantly higher in the low Treg group than that in the high Treg group at week 12 (44% vs. 0%) and week 24 (52% vs. 0%) (Fig. 1C).

Modeling and validation
To further clarify the application of these predictive markers, a nomogram was developed to estimate the likelihood of GR to low-dose IL-2 treatment in each individual patient (Fig. 3). The adjusted c-index for predicting GR in the primary cohort was 0.812, ranging from 0.64 to 0.97. In the independent validation procedure, the c-index was 0.776, ranging from 0.49 to 1.06 and thus showing good discrimination ability across all patients. Besides, the AUC for the combined model (i.e., the proportion of Tregs and the presence of skin rash) was 0.813, with the sensitivity of 68.75% and 100% speci city (Fig. 1A).

Discussion
CD4 + Tregs play a central role in immune tolerance, and Treg dysfunction is well described in autoimmune disorders. In vivo, Treg expansion through low-dose IL-2 treatment has been reported in a number of diseases, including chronic graft-versus-host disease (cGVHD), type 1 diabetes, SLE and several other autoimmune diseases (AIDs) [1][2][3][4][5][6][7]. These studies indicated that this treatment is well tolerated and effective, but prediction of patients who may have a better response to low-dose IL-2 is still a challenge in clinical practice.
Previous studies showed low Tregs and IL-2 in the circulation of SLE patients and the e cacy of lowdose IL-2 treatment. In this study, we proved that a low proportion of Tregs and the presence of skin rash could be predictors of responses to low-dose IL-2 therapy in SLE patients.
Increase of CD25hi-expressing Tregs was associated with clinical responsiveness to low-dose IL-2 [3]. However, post-treatment indicator would not help decision making by clinicians. Our ndings in this study support the utility of the baseline Treg and skin rash before IL-2 administration to predict SRI-4 response.
We also built a predictive model, which may help clinicians in decision making. The model indicated that Treg ≤ 13.35% plus skin rash presence predicted GR to IL-2 therapy with 100% speci city and 68.75% sensitivity. In addition, there was a trend that serum IL-2 and IL-10 levels were decreased in the SRI-4 responder group, although not statistically signi cant. This difference may become statistically signi cant with a larger sample size.
It remains unclear why there is a subgroup appearing to bene t more from low-dose IL-2. Low number of Treg cells was found in skin lesion of patients with cutaneous erythematosus [8]. Low-dose interleukin-2 speci cally expands Treg cells to confer immunosuppressive capacity to autoimmune response and in ammation. A limitation of this study is the small sample size.

Conclusion
In summary, we found in a pooled post-hoc analysis that a low Treg proportion and skin rash indicated good response to low-dose IL-2 treatment in SLE patients.    Nomogram for the prediction of clinical remission following low-dose IL-2 therapy in primary cohort. To use the nomogram, an individual patient's value is located on each variable axis, and a line is drawn upward to determine the number of points received for each variable's value. The sum of these numbers is located on the total points axis, and a line is drawn downward to the axis for the probability of disease alleviation to determine the likelihood of an SRI-4.