Vitiligo is an acquired autoimmune disease characterized by a loss of melanocytes in the skin and mucous membrane, resulting in patchy depigmentation, affecting 0.5-2% population worldwide (1–3). To date, vitiligo is still a refractory disease because of the high recurrence rate after treatment cessation and obscure pathogenesis. Although, conventionally, it was considered that the pathogenic mechanisms between segmental and non-segmental vitiligo are different, according to the latest understanding of pathogenesis in vitiligo, the onset and recurrence of two type vitiligo are both closely related to inflammatory and immunity-related mechanisms, especially T-cell immunity (4–7). The defect melanocytes were directly destructed by CD8+Tcells (6). Studies have demonstrated that the recruitment of T cells is mediated by IFN-γ via IFN-γ-regulated chemokines, such as C-X-C motif chemokine ligand (CXCL) 9 and CXCL10 (8–10). IFN-γ mainly derived from CD4+Tcell and CD8+Tcell that immune polarized to type-1-like in vitiligo (11). Then, IFN-γ activates the Janus kinase (JAK) signaling pathway, promoting the expression of its regulated chemokines, including CXCL9, CXCL10 and CXCL11 (9, 12, 13). IFN-γ also increased the expression of CXCR3, the receptor of CXCL9, CXCL10 and CXCL11, on the CD8+Tcells and melanocytes (8, 9, 14). Interplaying between IFN-γ and IL-6 is required for the expression of IFN-γ-regulated genes, which is a key part of autoimmunity (15, 16). The upregulation of IL-6 is also reported in vitiligo patients (17). The chemokines orchestrate CD8+Tcells positing in epidermis adjacent to melanocytes and mediate the apoptosis of melanocytes through specifically combining with CXCR3 (10). Recent study showed that the fibroblasts are essential for the process of aggregating CD8+Tcell through the IFN-γ-CXCL9/CXCL10-CXCR3 axis (18). Besides, as reported, curative effect of cultured melanocyte transplantation was associated with the level of CXCL9 in the skin blister fluid, that is, the patients with high level of CXCL9 preferred occurring re-depigmentation after melanocyte transplantation (19). The result indicated that CXCL9 might induce recurrence of vitiligo. Currently, tissue resident memory T cell (TRM) is regarded as the chief cause of vitiligo recurrence and persistent depigmentation (7, 20). TRM cells are incapable of inducing melanocytes apoptosis directly and more likely to recruit T cells from the circulation for exerting the attack by secreting IFN-γ and IFN-γ-regulated chemokines: CXCL9 and CXCL10 (21). A model study reported that the survival of TRM cells in lesion skin depended on IL-15. The blockage of IL-15 receptor significantly decreased TRM survival and reversed vitiligo (22). To date, a number of clinical researches have reported that IFN-γ and its induced chemokines, CXCL9 and CXCL10, are related to disease activity, severity and prognosis in vitiligo (8, 19, 23, 24). However, very few studies have investigated the cytokines and chemokines that serve as predictive biomarkers for disease recurrence in vitiligo. Based on the existed research on the pathogenesis of vitiligo, it was hypothesized in present study that IFN-γ, IFN-γ-regulated chemokines, IL-6 and IL-15 are associated with disease activity and also predictive biomarkers for disease recurrence in vitiligo. In this study, we first established a multivariate regression model to predict the recurrence of vitiligo. This research also provided clinical evidence that IFN-γ is the crucial part for vitiligo recurrence.

Determining the status of vitiligo is fundamental for treatment. In current practice, disease stage is mainly assessed by clinical symptom based on patients’ self-reporting (26). However, the clinical manifestations of vitiligo lag behind melanocyte apoptosis (27), making it challenging to detect the progression of vitiligo early and prevent timely. Hence, we utilized the Meso Scale Discovery (MSD) platform to measure plasma concentrations of some vitiligo-associated cytokines, aiming to identify potential biomarkers that could assess the condition of vitiligo more sensitively and objectively.

The circulating concentrations of IFN-γ and its regulated chemokines are increased and correlate with disease activity in a number of autoimmune diseases, such as alopecia areata, systemic lupus erythematosus, and Graves’ disease (28–33). However, due to the detection limit with conventional ELISA, there exist few researches on IFN-γ as a biomarker (24, 34). A cytokine analysis study by Ng et al., using the MSD detection technology, found that IFN-γ is a sensitive and specific cytokine for reflecting vitiligo activity (24). MSD, a technology with low detection limits that are more suitable for samples with low endogenous levels of cytokines, were also employed in our study for multi-cytokine detection (35, 36). However, our study did not find a significant difference in plasma IFN-γ levels between patients with active and stable vitiligo, though the level of plasma IFN-γ was increased in vitiligo patients, which possibly due to differences in the definitions of active and stable vitiligo, their active vitiligo definition: patients experienced progression within 3 months; our active vitiligo definition: patients experienced progression within 6 months. In agreement with the previous studies, our results also showed that the plasma levels of CXCL9 and CXCL10 are increased in vitiligo patients (8, 19, 37). However, in contrast to the previous reports that patients with active vitiligo had higher expression of CXCL9 and CXCL10 compared to those with stable vitiligo (8, 38, 39), we only found CXCL9 expression is higher in the circulation of patients with active vitiligo compared to those with stable vitiligo, which may be related to the participants in our study, including segmental and non-segmental vitiligo patients. But a clinical study that included both segmental and non-segmental vitiligo patients, similar to our study, also showed that only serum CXCL9 levels correlated with vitiligo activity (19). The relationship between the expression of CXCL11 and vitiligo remains unclear because of the limited research on CXCL11 and vitiligo (40). But given its role in other autoimmune diseases and the relationship with CXCL9 and CXCL10, we also included CXCL11 in our study as a potential pathogenic factor (28, 41, 42). Rashighi et al. found that the expression of the CXCL11 gene is upregulated in lesional skin of vitiligo patients (28, 41, 42). Similarly, our results showed that the level of CXCL11 in plasma was elevated in vitiligo patients but no association with disease activity. Besides, almost consistent with previous results, our results also showed that the expression levels of IL-6 and IL-15 were significantly higher in vitiligo patients compared to healthy controls, with no significant difference between active and stable vitiligo (39); But a clinical study reported that IL-6 expression was significantly higher in active vitiligo compared to stable vitiligo (43). To exclude the potential impact of disease duration on cytokine levels, we performed further statistical analyses to examine the relationship between disease duration and cytokine levels. Our analysis did not reveal any significant correlation between cytokine levels and disease duration. However, Mascarenhas et al. reported a significant negative correlation between CXCL10 levels and disease duration (44). This discrepancy may be attributed to differences in the patient cohorts. In the study by Mascarenhas et al., the disease duration of recruited patients was evenly distributed over a range of 0–40 years, with most patients having a disease duration exceeding 10 years; In contrast, the majority of patients in our cohort had a disease duration of less than 10 years. Such a relatively short disease duration might not fully reflect the potential relationship between disease duration and cytokine levels.

In recent years, a central role of IFN-γ in vitiligo has been established (7). IFN-γ upregulates the expression of chemokines CXCL9 and CXCL10 through the JAK/STAT pathway to orchestrate CD8+ T cells positing in epidermis adjacent to melanocytes and facilitate their cytotoxic activity against surrounding melanocytes (10, 45). The interaction between IL-6 and IFN-γ is required for the expression of IFN-γ-regulated genes in autoimmunity (15). Besides, in the treatment of melanoma, immune checkpoint inhibitors often induce vitiligo-like skin lesions and patients with these lesions survive longer than those without (46, 47). Garris et al. reported that during programmed cell death protein 1 (PD-1) antibody treatment, the secretion of IFN-γ by CD8+ T cells is crucial for activating anti-tumor T cells, and neutralizing IFN-γ in mice impairs control over melanoma progression (48). Analysis of tumor microenvironment gene expression profiles revealed that tumors with higher levels of IFN-γ mRNA respond more effectively to PD-1 antibody therapy (49, 50). These results indicate that IFN-γ signaling is essential for the autoimmune-mediated destruction of melanocytes. Remarkably, tissue resident memory T cell (TRM) and recirculating memory T cells (TCM) also express IFN-γ, IFN-γ-induced chemokines and CXCR3 (20, 21, 51, 52), suggesting that IFN-γ and its inducible proteins may be involved in the maintenance and recurrence of vitiligo. Additionally, TRM cells express the CD122 subunit of IL-15 receptor. Richmond et al. reported that short-term blocking of IL-15 using anti-CD122 inhibits TRM production of IFN-γ and reverses vitiligo. Long-term blockade of IL-15 depletes TRM cells from skin lesions (22), indicating that IL-15 is essential for the survival of TRM cells. A mice model study demonstrated that JAK inhibitors reversed vitiligo but TRM cells are not depleted with the treatment (53). Another model study showed that blocking the recruitment or depleting TCM cells leads to repigmentation of vitiligo lesions, though the number of TRM cells in the lesions remains unchanged (21). Therefore, we propose a potential recurrence hypothesis: The binding of IL-15 to its receptor on TRM sustains survival of TRM cells and promotes the secretion of IFN-γ by these cells. The interaction between IFN-γ and IL-6 upregulates the expression of CXCL9, CXCL10, and CXCL11 in TRM cells and fibroblasts at the lesion site. These chemokines recruit TCM cells from lymph tissues to the vicinity of epidermal melanocytes by forming a gradient. Then, the recruited TCM cells attack the melanocytes and secrete IFN-γ, reinforcing the positive feedback loop.

If the hypothesis is confirmed, there should be correlations among the cytokines and chemokines involved. To test this hypothesis, we performed a correlation analysis of the plasma suspects expression in the recurrent patients. Consistent with expectations, the results showed strong positive correlations among these factors, providing indirect evidence for our hypothesis. Interestingly, IL-6 showed a strong positive correlation with the other five factors, which suggest that the interaction between IL-6 and IFN-γ in the induction of IFN-γ-regulated genes may also play a role in the pathogenesis of vitiligo recurrence. To further explore whether these suspected factors are triggers for vitiligo recurrence, we compared the plasma concentrations of these factors between recurrent and persistent stable patients. Our results showed that, except for IL-15, the levels of these factors were significantly higher in recurrent patients compared to persistent stable patients, which suggest that increased levels of IFN-γ, CXCL9, CXCL10, CXCL11 and IL-6 are potential triggers for vitiligo recurrence. ROC curve indicates that these five factors can serve as predictive biomarkers for vitiligo recurrence, providing an objective reference for assessing the disease status of vitiligo patients. Although IL-15 levels in recurrent patients showed an upward trend, the difference was not statistically significant. This may be due to the sustained high expression of IL-15 in stable patients to maintain TRM cell survival in lesional skin (22). In the process of recurrence, IL-15 may primarily act in the lesional skin, resulting in only a slight increase in circulation.

To predict vitiligo recurrence more reliably. we established a multivariate binary logistic regression model using the cytokines and chemokines significantly elevated in recurrent patients as independent variables and disease recurrence as the dependent variable. This model also eliminate collinearity among factors, identifying that IFN-γ is an independent predictor for vitiligo recurrence. Remarkably, the model demonstrated excellent predictive capability, with an accuracy of 90.5% on the training dataset and 88.9% on the testing dataset, which suggest that it is possible to predict the recurrence of vitiligo by monitoring the concentrations of cytokines and chemokines in the circulation, enabling timely intervention to prevent disease progression. But it should be noted that the sample size in this study is insufficient for the construction of the model, which may limit the model’s stability and generalizability, especially the potential issue of overfitting. However, our model adheres strictly to the established principles of predictive model and can provide a foundation for future larger-scale research.

There are several limitations in this study. Firstly, this is a single-center study, which may introduce a potential selection bias in the patient cohort. Secondly, the cohort size was relatively small, particularly for training the model, even though the model is theoretically supported. Moreover, our study only detected a few cytokines and chemokines closely related to vitiligo but did not screen all the cytokines that may play a role in pathogenesis of vitiligo. Above all, future large-scale studies based on multicenter and longitudinal samples are needed to investigate the plasma cytokines associated with vitiligo recurrence and validate the feasibility of multivariate model for predicting vitiligo recurrence.

In conclusion, this study discovers that plasma IFN-γ, IFN-γ-regulated chemokines, including CXCL9, CXCL10 and CXCL11, and IL-6 might be potential biomarkers for vitiligo recurrence, with CXCL9 also associated with disease activity. Additionally, our study also demonstrated that IFN-γ is an independent predictor of vitiligo recurrence and multivariate model could be a candidate approach for predicting vitiligo recurrence.