Low-dose IL-2 therapy for immune-related adverse events via Tfh/Treg balance modulation: a prospective cohort and murine model study Authors: Yifan Wang, Zhening Zhang, Yufei Li, Yongjing Cheng, Min Wang, Yudong Liu, Hao Li, Zhi Peng, Jing He Published: 2026 ABSTRACT AbstractBackgroundImmune-related adverse events (irAEs) are associated with anti-PD-1/PD-L1 therapy in gastrointestinal cancers. Altered phenotypes and frequencies of T cell subsets play a critical role in irAEs. This study characterized the Tfh/Treg balance in irAEs patients and assessed the efficacy of LDIL-2 therapy in mouse models.MethodsA prospective study evaluating the immunological characteristics of 26 gastrointestinal cancer patients through both cellular and transcriptomic analyses before and after anti-PD-1/PD-L1 therapy was performed. Patients were categorized into two groups: the AE group (patients who developed irAEs during the follow-up period) and the NAE group (patients who did not). Meanwhile, MRL/MpJ-Faslpr mice, a lupus mouse model, and tumor-bearing C57BL/6 J mice were used to evaluate the efficacy of LDIL-2 in treating anti-PD-L1 induced irAEs and its impact on antitumor effectiveness.ResultsOur study discovered that in AE patients, Tfh-related genes were significantly upregulated during the early stages of anti-PD-1/PD-L1 therapy (IL-21: p < 0.01; PDCD1: p < 0.05), whereas the upregulation of Treg-related genes was less pronounced compared to Tfh-related genes. This distinction was not observed in NAE patients. Flow cytometry results further confirmed a substantial increase in the Tfh/Treg ratio in AE patients during the early stages of therapy (p < 0.001), a phenomenon absent in NAE patients. In MRL/MpJ-Faslpr mice, anti-PD-L1 therapy induced a shift in the Tfh/Treg ratio, along with severe organ inflammatory lesions and elevated anti-dsDNA antibody levels. In contrast, LDIL-2 treated mice maintained a balanced Tfh/Treg ratio, exhibiting significantly fewer organ inflammatory lesions and reduced anti-dsDNA production. Furthermore, in tumor-bearing mice, LDIL-2 did not impair the antitumor efficacy of anti-PD-L1 therapy.ConclusionsOur study underscores the importance of Tfh/Treg balance in the development of irAEs and demonstrates the potential of LDIL-2 as a therapeutic option for irAEs, providing a promising alternative to traditional immunosuppressive therapies for managing irAEs.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00262-026-04452-6.Keywords: Immune checkpoint inhibitors, PD-1 inhibitors, PD-L1 inhibitors, Immune-related adverse events. T follicular helper cells, Regulatory T cells, Low-dose IL-2 Introduction Current treatment guidelines for gastrointestinal cancers, including gastric, colorectal, and esophageal cancers, emphasize the use of immune checkpoint inhibitors (ICIs) that target the programmed cell death protein 1 (PD-1) or programmed death-ligand 1 (PD-L1) axis. Drugs such as nivolumab and pembrolizumab have become central to managing advanced or metastatic gastrointestinal cancers [1–5]. However, the heightened immune activation triggered by anti-PD-1/PD-L1 therapy can result in immune-related adverse events (irAEs), a collection of autoimmune conditions impacting various organs and systems, such as the skin, heart, liver, and musculoskeletal system [2–7]. These irAEs affect up to 80% of patients, with a particularly high incidence in those receiving combination ICI therapies [8, 9]. The prevalence and complexity of irAEs present significant challenges for clinicians in managing patients receiving anti-PD-1/PD-L1 therapy. Anti-PD-1/PD-L1 monoclonal antibodies (mAbs) function by blocking T cell inhibition, thereby enhancing the antitumor immune response. However, alterations in the phenotype or frequency of various T cell subsets have also been identified as mechanisms contributing to many autoimmune conditions, including irAEs [10]. Recent studies have shown that the expansion of activated, proliferating CD4+ T cells in the early stages of ICI treatment may be a potential predictor of subsequent irAEs [11]. CD4+ T cell subsets, including follicular helper T cells (Tfh) and regulatory T cells (Treg), play a crucial role in mediating effective antitumor immunity and maintaining immune self-tolerance [12–14]. Tfh cells express surface molecules such as inducible T-cell co-stimulatory (ICOS), PD-1, and C-X-C chemokine receptor type 5 (CXCR5). The increase in Tfh cells has been implicated in various autoimmune diseases [15]. Recent studies suggest that ICIs can promote a dysregulated Tfh cell response, leading to irAEs [16]. A multi-omic analysis of over 18,000 patient samples indicates a correlation between Tfh cells and irAEs [17]. Other studies have observed that PD-1 deficiency in antigen-specific T cells increases Tfh cell frequency [18], and anti-PD-1 treatment has been shown to elevate Tfh cells, as confirmed by both cellular and transcriptomic analyses [19]. Treg cells are characterized by high levels of CD25 and Forkhead box protein P3 (Foxp3). In contrast to Tfh cells, Treg cells are essential for maintaining self-tolerance and regulating immune responses [14]. In ICI treatment, Treg cells can suppress irAEs to maintain immune homeostasis while potentially limiting antitumor immunity through immunosuppressive mechanisms, which may affect the efficacy of ICI treatment [20]. Tfh and Treg lineages are interconnected, with Tfh differentiation dependent on the transcription factor B-cell lymphoma 6 (BCL-6), while Treg cells rely on Foxp3 [14]. These transcription factors compete to maintain the Tfh/Treg balance. In many autoimmune diseases, patients show increased Tfh cells and reduced Treg cell number, leading to a Tfh/Treg imbalance skewed toward Tfh cells in circulation [21–23]. Emerging evidence suggests that PD-1 antagonism can disrupt this delicate balance, often leading to an expansion of Tfh cells and/or altered Treg activity, thereby skewing the Tfh/Treg ratio [13, 24, 25]. This resulting imbalance is increasingly recognized as a key contributor to the pathogenesis of irAEs. However, the role of this imbalance in the development of irAEs remains unclear. Modulating the Tfh/Treg balance has been beneficial in treating many autoimmune diseases [14]. Interleukin-2 (IL-2) is a critical cytokine for maintaining this balance. Our previous study showed that low-dose IL-2 (LDIL-2) promotes Treg function and suppresses Tfh cell activity, demonstrating efficacy in autoimmune diseases such as systemic lupus erythematosus (SLE) and Sjögren’s disease (SjD) [26, 27]. Compared to traditional immunosuppressants, LDIL-2 generally has fewer systemic immunosuppressive effects, leading to a lower incidence of infections and other adverse effects [28, 29]. However, the use of LDIL-2 in the context of irAEs remains unexplored. In this study, through immunophenotypic analysis, gene expression analysis, and clinical observation in gastrointestinal cancer patients receiving anti-PD-1/PD-L1 mAb, we found an imbalance of Tfh and Treg cells in patients who developed irAEs after 6 weeks of anti-PD-1/PD-L1 therapy. Subsequently, we utilized MRL/MpJ-Faslpr mice and tumor-bearing C57BL/6 J mice to investigate the efficacy of LDIL-2 in managing irAEs and its impact on antitumor therapy. Materials and methods Patient enrollment We prospectively enrolled patients with unresectable tumors at the Department of Gastrointestinal Oncology, Peking University Cancer Hospital, between December 1, 2021, and December 1, 2022. Eligible participants had a confirmed pathological diagnosis of invasive gastrointestinal carcinoma, were scheduled to receive anti-PD-1/PD-L1 monotherapy, and met the following inclusion criteria: 1) age ≥ 18; 2) Eastern Cooperative Oncology Group (ECOG) Performance Status of 0 or 1; 3) expected survival of ≥ 3 months; 4) no prior exposure to immunotherapy, including anti-PD-1/PD-L1 agents. Patients receiving concurrent systemic treatments, such as targeted therapy, radiotherapy, or cellular therapy, were excluded. Additionally, individuals with preexisting rheumatic diseases (e.g., SLE, SjD), hematological disorders (e.g., idiopathic thrombocytopenic purpura, myelodysplastic syndrome), or conditions requiring corticosteroids or immunomodulators were also excluded. Demographic data, tumor characteristics, treatment details, and irAEs were recorded. Both male and female patients were included, and sex was not considered a biological variable. Patient follow-up and irAEs grading All patients were followed up for 6 months. The evaluation of irAEs and their grading were conducted using the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. Any adverse event with an immunological origin that arose within 6 months of starting treatment was classified as an irAE. Peripheral blood mononuclear cells (PBMCs) collection For all patients, EDTA-anticoagulated whole blood was collected. PBMCs were isolated by Ficoll-Hypaque density gradient centrifugation and immediately processed for RNA extraction and flow cytometry. In terms of sample collection timing, two time points were adopted: 1) within 3 days prior to the first dose of immunotherapy (baseline) and 2) 6 weeks after initiating anti-PD-1/PD-L1 therapy (early stages) (Fig. 1A). AE patients have elevated Tfh/Treg ratios after the early stages of anti-PD-1/PD-L1 therapy. (A) Graphical representation of the experimental approach for the discovery cohort. (B-C) Volcano plot highlighting the Tfh-related genes and Treg-related genes after early stages of anti-PD-1/PD-L1 treatment in AE (B) and NAE (C) patients. (D-E) Bubble plots of mRNA abundance changes in the early stage of anti-PD-1/PD-L1 immunotherapy, analyzed in AE patients (D) and NAE patients (E). (F) Tfh cell was defined as CD4+CXCR5+ICOS+ subset; Treg cell was defined as CD4+CD25+CD127.− subset. (G-H) The dynamic of Tfh/Treg ratio between early stages and baseline in AE (G) and NAE (H) patients. Statistical significance was determined by Wilcoxon signed-rank test. (I-J) Dynamic of Tfh/Treg ratio in PBMCs of AE (I) and NAE (J) patients after cultured with anti-PD-1 mAb and anti-PD-1 mAb + LDIL-2. Statistical significance was determined by Wilcoxon signed-rank test. IrAEs, immune-related adverse events; AE, patients who developed irAEs within the follow-up period after anti-PD-1/PD-L1 treatment; NAE, patients who didn’t develop irAEs within the follow-up period after anti-PD-1/PD-L1 treatment; early stages, 6 weeks after receiving anti-PD-1/PD-L1 treatment; Anti-PD-1: PBMCs cultured with anti-PD-1 mAb; Anti-PD-1 + IL-2: PBMCs cultured with anti-PD-1 mAb combined with LD-IL-2; Blank: PBMCs treated with PBS. Tfh, follicular helper T cell; Treg, regulatory T cells. * p < 0.05, ** p < 0.01, *** p < 0.001 In vitro experiment PBMCs were isolated from 11 patients with gastrointestinal cancer who were naïve to immunotherapy using density gradient centrifugation. Cells were assigned to three treatment groups: (1) Blank group (treated with PBS); (2) Anti-PD-1 group (treated with 10 µg/mL Nivolumab; Bristol-Myers Squibb, USA); and (3) Anti-PD-1 + LDIL-2 group (treated with 10 µg/mL Nivolumab plus 10 IU/mL rhIL-2; SL PHARM, China). Following a 72-h incubation at 37 °C in a humidified 5% CO2 atmosphere, the proportions of Tfh and Treg cells were analyzed by flow cytometry. Detailed methods for RNA sequencing (RNA-seq) and flow cytometry are provided in the Supporting Information [30]. Mice Animal studies received ethical approval from the Institutional Animal Care and Use Committee (IACUC) of Peking University Health Science Center (2021PHE026). Eight-week-old female MRL/MpJ-Faslpr (C001976) mice were purchased from Cyagen Biosciences Inc. Eight-week-old male C57BL/6 J mice were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. All mice were maintained under SPF conditions, provided with sterile pellet food and water, and maintained on a normal day-night cycle. Mice were allocated into each experimental group by randomization. For the generation of tumor-bearing mice, log-phase Mouse Lewis lung cancer (LLC) cells suspended in phosphate buffer solution (PBS) were subcutaneously inoculated into the right flank of C57/BL6J mice (2.5 × 105 tumor cells per mouse) [31]. Mice experimental protocol MRL/MpJ-Faslpr mice were randomly divided into three groups (n = 7 per group). The experimental group received twice-weekly intraperitoneal injections of anti-PD-L1 antibodies (clone 10F.9G2, Bio X Cell, BE0101) at a dose of 10 mg/kg[32], combined with subcutaneous administration of rhIL-2 (SL PHARM, Beijing, China) every other day at a dose of 30,000 IU per mouse [26]. The other two control groups received either anti-PD-L1 antibodies alone or PBS. Throughout the 4-week treatment period, mice were monitored daily for activity and appearance, with weight and urinary protein levels quantified every 2 weeks. C57/BL6J LLC-bearing mice were also randomly divided into 3 groups (n = 6 per group) on the day after tumor inoculation and received the same drugs described above. Mice were analyzed at 13 weeks with blinding on experimental groups. The lung, kidney, and skin tissues of mice were stained with Hematoxylin and Eosin (H&E) and evaluated by two blinded pathologists. Immunohistochemical analysis was performed on kidney tissue. Detailed methods for histological and immunohistochemical analysis, spleen cell collection, and anti-dsDNA concentration measurement are provided in the Supporting Information. Assessment of lung injury Lung sections were evaluated in 10 random fields per slide based on immune cell infiltration, alveolar thickening, and parabronchial infiltration with the following scores: 0 (indicating no discernible findings) to 4 (indicative of diffuse inflammation with over 50 visualized lumens), as previously published [33, 34]. Assessment of kidney injury Kidney sections were scored according to the National Institutes of Health (NIH) activity and chronicity index adapted by the International Society of Nephrology/Renal Pathology Society classification for lupus nephritis [35, 36]. The scoring criteria include these findings: endocapillary hypercellularity, neutrophils/karyorrhexis, fibrinoid necrosis, wire loop lesions and/or hyaline thrombi, cellular/fibrocellular crescents, and interstitial inflammation. Software ImageJ was used to quantify IgG and C3 staining intensity. Assessment of skin injury Skin inflammation in MRL/MpJ-Faslpr mice (neck) was assessed using a 0–3 scoring system. The criteria were: 0, no lesions; 1, mild hair loss with erythema and few lesions; 2, erythema, scabbing, and limited hair loss; 3, extensive ulceration [37]. Statistical analysis In all comparisons, participants have matched demography characteristics. Descriptive analysis was employed to present demographic characteristics. Continuous variables were compared using a Student’s t-test for normally distributed data or the Mann–Whitney test for non-normally distributed data. Two-sided p-values < 0.05 were considered statistically different. All statistical analyses in this study were conducted by SPSS 24.0, Prism 9.0, and R (Version 4.3.2, https://www.r-project.org). Results Characteristics of patients A total of 26 patients were finally enrolled, including 19 (73.08%) patients with gastric cancer, 1 (3.85%) patient with colorectal cancer, 5 (19.23%) patients with neuroendocrine cancer, and 1 (3.85%) patient with esophageal cancer (Table 1). Table 1Clinical characteristics of enrolled patientsCharacteristicsPatients with irAEs onset (n = 13)Patients without irAEs onset (n = 13)Age, years, mean ± SD55.54 ± 14.0761.08 ± 15.10GenderMale910Female43Tumor typeGastric cancer910Colorectal cancer01Neuroendocrinal cancer32Esophageal cancer10Highest irAEs gradeNumber of patients293242Affected organ & systemsTotal eventsSkin7Musculoskeletal system3Cardiovascular system1Liver3Gastrointestinal tract2Thyroid5Adrenal gland1Pancreas2Bone marrow1SD, standard deviation; irAEs, immune-related adverse events; BMI, body mass index. Skin: including rash; Musculoskeletal system: including myositis; Cardiovascular system: including myocarditis; Liver: including elevated liver enzymes; Gastrointestinal tract: including colitis; Thyroid: including thyroid dysfunction; Adrenal gland: including adrenal insufficiency; Pancreatic: including pancreatitis; Bone marrow: including myelosuppression Clinical characteristics of enrolled patients SD, standard deviation; irAEs, immune-related adverse events; BMI, body mass index. Skin: including rash; Musculoskeletal system: including myositis; Cardiovascular system: including myocarditis; Liver: including elevated liver enzymes; Gastrointestinal tract: including colitis; Thyroid: including thyroid dysfunction; Adrenal gland: including adrenal insufficiency; Pancreatic: including pancreatitis; Bone marrow: including myelosuppression Thirteen patients developed irAEs, while thirteen patients did not. These irAEs affected various organs and systems, including the skin, muscles, heart, liver, gastrointestinal tract, pancreas, thyroid, adrenal gland, and bone marrow. We categorized patients who developed irAEs during the follow-up as the AE group and those who did not as the NAE group. All AE patients experienced at least grade 2 or higher grade of irAEs (Table 1). Elevated Tfh/Treg ratios are associated with the development of irAEs Using RNA-seq, we characterized the transcriptomic features of PBMCs from 6 AE patients and 6 NAE patients at both baseline and after 6 weeks (early stages) of anti-PD-1/PD-L1 therapy. We specifically focused on genes critical for the Tfh and Treg lineages. In AE patients, Tfh-related genes, including IL-21 and PDCD1, showed a significant upregulation after anti-PD-1/PD-L1 therapy (IL-21: p < 0.01; PDCD1: p < 0.05) (Fig. 1B). Notably, IL-21 levels increased substantially compared to baseline, reaching a 5.81-fold increase after the early stages of treatment (Fig. 1D). Regarding Treg-related genes, including FOXP3 and IL2RA, although they were also upregulated after anti-PD-1/PD-L1 therapy, the increase did not reach statistical significance. Moreover, the upregulation magnitude was markedly lower than that of Tfh-related genes (Fig. 1B, D). In NAE patients, after anti-PD-1/PD-L1 therapy, IL-21, FOXP3, and IL2RA expression slightly increased compared to baseline, while PDCD1 expression decreased. The changes in both Tfh-related and Treg-related genes were not statistically significant in NAE patients, and the magnitude of change between these two gene groups was not markedly different (Fig. 1C, E). Furthermore, we used flow cytometry to analyze the relative proportions of the T cell subsets in baseline and early stages PBMCs from the other 7 AE patients and 7 NAE patients. Since the surface marker PD-1 cannot be stained in PBMCs after anti-PD-1/PD-L1 therapy, we used the CD4+CXCR5+ICOS+ subset to represent Tfh cells (Fig. 1F). After the early stages of anti-PD-1/PD-L1 therapy, AE patients showed a substantial increase in the Tfh/Treg ratio (p < 0.001) (Fig. 1G), while this phenomenon was not found in NAE patients (Fig. 1H). The raw frequencies of Tfh and Treg cells are presented in Supplementary Fig. 3. These results suggest that AE patients exhibited a more robust Tfh-skewed Tfh/Treg balance following anti-PD-1/PD-L1 therapy, and this shift may be closely associated with the occurrence of irAEs. Low-dose IL-2 could reverse the Tfh/Treg imbalance associated with irAEs Previous research has indicated that LDIL-2 can modulate the immune balance between Tfh and Treg cells, alleviating various autoimmune diseases. Therefore, we investigated whether LDIL-2 could reverse the Tfh/Treg imbalance associated with irAEs, potentially preventing the Tfh-skewed phenotypic transformation and the onset of irAEs. PBMCs were isolated from patients at baseline and cultured with anti-PD-1 mAb (Anti-PD-1 group), anti-PD-1 mAb combined with LDIL-2 (Anti-PD-1 + IL-2 group) and PBS (Blank group). Subsequently, data were stratified into AE and NAE groups based on patients’ later irAE development. We observed that the Tfh/Treg ratio significantly increased in PBMCs from both AE and NAE patients after in vitro anti-PD-1 mAb treatment. Notably, PBMCs from AE patients exhibited a significantly greater increase in the Tfh/Treg ratio (p < 0.01) (Fig. 1I) compared to NAE patients (Fig. 1J). In both groups, this elevation was consistently suppressed when co-treated with LDIL-2. These in vitro results further validate our in vivo finding that AE patients exhibited a more robust Tfh-skewed Tfh/Treg balance following anti-PD-1/PD-L1 therapy. Moreover, these findings suggest that LDIL-2 may help treat irAEs by reversing the Tfh/Treg imbalance following anti-PD-1/PD-L1 therapy. Low-dose IL-2 ameliorates irAE-like pathology in murine models To explore whether LDIL-2 can assist in the treatment of irAEs, we employed MRL/MpJ-Faslpr mice as a preclinical model, which is homozygous for the lymphoproliferation spontaneous mutation (Faslpr) and shows systemic autoimmunity [38]. Mice were periodically treated with anti-PD-L1 mAb (anti-PD-L1, n = 8), anti-PD-L1 mAb combined with LDIL-2 (anti-PD-L1 + IL-2, n = 8), and PBS (blank, n = 8) (Fig. 2A). After 4 weeks of treatment, the anti-PD-L1 group showed a significant increase in urinary protein levels compared to the blank group (p < 0.05). In contrast, levels in the anti-PD-L1 + IL-2 group were significantly lower than those in the anti-PD-L1 group (p < 0.05) and showed no statistically significant difference from the blank group (Fig. 2B). LDIL-2 could alleviate irAEs while preserving tumor control in mice. (A) Schematic workflow of MRL/MpJ-Faslpr mouse models. (B) Urinary protein levels in the blank group, anti-PD-L1 group and anti-PD-L1 + IL-2 group MRL/MpJ-Fas.lpr mice at 9, 11, and 13 weeks of age, mean and SD. (C) Representative photomicrographs of skin, lung, and kidney tissues stained with hematoxylin and eosin (H&E) in the three groups of mice. The anti-PD-L1 group exhibited pathological changes in all three organs: in the lung, peribronchial infiltration of chronic inflammatory cells (black arrows); in the skin, epidermal thickening (red arrows) with scattered inflammatory cell infiltration in the dermis (black arrows); and in the kidney, glomerular capillary occlusion (green arrows), perivascular inflammatory cell infiltration (black arrows), and congested dilated interstitial vessels (blue arrows). These changes were significantly alleviated in the anti-PD-L1 + IL-2 group. (D) Injury score of skins, lungs, and kidneys in the three groups of mice, mean and SD. (E) Representative immunofluorescence staining for IgG and C3 of kidney sections and (F) data quantification in the three groups of mice. (G) Anti-dsDNA antibody levels in the three groups of mice at 13 weeks of age. Blank group, mice treated with PBS. Anti-PD-L1 group, mice treated with anti-PD-L1 mAb; anti-PD-L1 + IL-2 group, mice treated with anti-PD-L1 mAb combined with LDIL-2. SD, standard deviation. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.001 Moreover, H&E staining revealed that the anti-PD-L1 group exhibited marked epidermal thickening and more extensive inflammatory cell infiltration in the skin, as well as more pronounced lymphocytic infiltration in the lungs, compared with the blank group. In the anti-PD-L1 + IL-2 group, both the skin lesions and lymphocytic infiltration in the lungs were markedly reduced compared to those in the anti-PD-L1 group (Fig. 2C). In the kidney, sections from the anti-PD-L1 group exhibited frequent glomerular capillary occlusion, perivascular chronic inflammatory cell infiltration, and congested dilated interstitial vessels. These pathological features were not evident in the anti-PD-L1 + IL-2 group (Fig. 2C). Consistently, the injury scores for the skin (p < 0.05), lung (p < 0.05), and kidney (p < 0.001) were significantly higher in the anti-PD-L1 group than in the Blank group (Fig. 2D). In the anti-PD-L1 + IL-2 group, injury scores for the skin (p < 0.05), lung (p < 0.05), and kidney (p < 0.01) were significantly reduced compared to the anti-PD-L1 group and were comparable to those in the Blank group (Fig. 2D). Furthermore, the anti-PD-L1 group exhibited significantly more substantial IgG deposition in the glomeruli compared to the blank group (p < 0.05), along with increased C3 deposition, which did not reach statistical significance. In contrast, in the anti-PD-L1 + IL-2 group, levels of both IgG and C3 deposition were comparable to those in the blank group and lower than those in the anti-PD-L1 group, with the reduction reaching statistical significance for IgG (p < 0.0001) (Fig. 2E-F). In addition, the anti-PD-L1 group exhibited significant higher serum anti-dsDNA levels compared to the blank group (p < 0.001). Meanwhile, the anti-PD-L1 + IL-2 group showed an apparent reduction in anti-dsDNA levels compared to the anti-PD-L1 group (p < 0.05) (Fig. 2G). These results indicate that anti-PD-1/PD-L1 therapy can induce an earlier onset and exacerbation of autoimmune disease in mice with a predisposition to autoimmunity. In contrast, LDIL-2 can inhibit this trend, suggesting that LDIL-2 protects against inflammatory responses associated with irAEs. Low-dose IL-2 modulates the Tfh/Treg balance in murine models We performed splenocyte flow cytometry on MRL/MpJ-Faslpr mice after 4 weeks of treatment. The results showed a significant increase in Tfh cells in the anti-PD-L1 group compared with the blank group (p < 0.01), whereas a decrease was observed in the anti-PD-L1 + IL-2 group compared to the anti-PD-L1 group (Fig. 3A). Treg cell levels were also elevated in the anti-PD-L1 group compared with the blank group, and this increase was further enhanced in the anti-PD-L1 + IL-2 group (Fig. 3B). The Tfh/Treg balance shifted significantly toward Tfh dominance in the anti-PD-L1 group (p < 0.05), but this shift was markedly suppressed in the anti-PD-L1 + IL-2 group (p < 0.05) (Fig. 3C). The cellular alterations in the anti-PD-L1 group closely mirrored the trends observed in PBMCs from patients in the AE group. Treatment with LDIL-2 could partially mitigate these changes, which suggests that LDIL-2 may prevent and treat irAEs by modulating the balance between Tfh and Treg cells. LDIL-2 modulates the Tfh/Treg balance. Splenic lymphocyte flow cytometry revealed that the administration of LDIL-2 could counteract the expansion of Tfh cells and inhibit the reduction of Treg cells induced by anti-PD-L1 therapy, thereby modulating the balance between Tfh and Treg cells, mean and SD. Blank group, mice treated with PBS. Anti-PD-L1 group, mice treated with anti-PD-L1 mAb; anti-PD-L1 + IL-2 group, mice treated with anti-PD-L1 mAb combined with LDIL-2. Tfh, follicular helper T cell; Treg, regulatory T cells. SD, standard deviation. * p < 0.05, ** p < 0.01 Low-dose IL-2 preserves anti-tumor efficacy in murine tumor models We then employed a tumor-bearing mouse model to investigate whether LDIL-2 compromises the antitumor effect of anti-PD-1/PD-L1 therapy. LLC-inoculated C57BL/6 J mice were treated with anti-PD-L1 mAb (Anti-PD-L1 group, n = 7), anti-PD-L1 mAb combined with LDIL-2 (Anti-PD-L1 + IL-2 group, n = 7), and PBS (Blank group, n = 7) (Fig. 4A). The tumor volumes among the three groups were similar before treatment initiation. After 3 weeks of different interventions, the tumor volumes in the anti-PD-L1 + IL-2 group mice were comparable to those in the anti-PD-L1 group mice but significantly smaller than the blank group (Figs. 4B and S4). Our results suggest that LDIL-2 could alleviate irAEs in mouse models without compromising the antitumor effect of anti-PD-1/PD-L1 therapy. LDIL-2 treatment does not compromise the antitumor efficacy of PD-L1 blockade. (A) Schematic workflow of tumor-bearing mouse model. (B) Endpoint tumor weight of LLC-bearing mice in the three experimental groups after 3 weeks of treatment. Data are presented as mean ± SD.. Blank group, mice treated with PBS. Anti-PD-L1 group, mice treated with anti-PD-L1 mAb; anti-PD-L1 + IL-2 group, mice treated with anti-PD-L1 mAb combined with LDIL-2. SD, standard deviation. *** p < 0.001 Discussion Tfh and Treg cells play contrasting roles in immune regulation and maintain a dynamic balance. Their imbalance has been shown to play an essential role in various autoimmune diseases [21–23], potentially contributing to the development of irAEs. Our findings illuminate the distinct circulating immunological landscape of irAEs, emphasizing the association of T cell subsets, particularly Tfh and Treg cells, with the occurrence of irAEs. According to our results, patients who developed irAEs exhibited a significant shift in the Tfh/Treg balance toward Tfh cells after the early stages of anti-PD-1/PD-L1 therapy, as shown by both cellular and transcriptomic analyses, suggesting that this imbalance may play a pathogenic role in irAEs. Our in vitro experiments corroborated these findings, demonstrating a significantly more pronounced increase in the Tfh/Treg ratio within PBMCs from AE patients after anti-PD-1 mAb treatment. Although an elevated Tfh/Treg ratio was also observed in NAE patients, this is consistent with previous reports indicating that PD-1 blockade generally promotes Tfh activation and modulates Treg activity across individuals [18, 19, 24, 25]. Our study further demonstrates that such immune perturbations only become clinically significant and lead to irAEs when the Tfh/Treg imbalance is excessive or dysregulated. Our data indicate that the critical biomarker is the dynamic change in the Tfh/Treg ratio from baseline to the early stages of treatment, rather than a single absolute “favorable” value. An “unfavorable” profile, associated with a higher risk of developing irAEs, is characterized by a significant increase in the Tfh/Treg ratio after 6 weeks of ICI therapy. Conversely, a “favorable” profile is one where this ratio remains relatively stable. These findings confirm previous research linking Tfh cells to the occurrence of irAEs [17] and the role of Treg cells in suppressing irAEs and maintaining immune homeostasis [20], while also highlighting that irAEs share immunological characteristics with other autoimmune diseases, such as SLE and SS, particularly in T cell profiles involving Tfh and Treg cells [21–23]. IL-2 plays a vital role in maintaining the Tfh/Treg balance. LDIL-2 has been shown to enhance Treg function while inhibiting Tfh cell activity, demonstrating therapeutic benefits in various autoimmune diseases [26, 27]. Therefore, LDIL-2 may offer potential therapeutic benefits for irAEs. Our selection of 30,000 IU every other day in vivo and 10 IU/mL in vitro as the low-dose is based on our previous work demonstrating their ability to selectively modulate CD4+ T cell subsets—specifically by expanding the Treg population and enhancing its suppressive function—without compromising overall immunity [26, 27]. It is important to note that these in vitro and in vivo doses were chosen for their validated biological effects within each respective experimental system, rather than through a direct dose-equivalence conversion. While ultra-low doses (1,000 IU daily) have been reported to potentially impair anti-tumor responses by over-expanding Tregs [39], that study also indicated that doses exceeding 10,000 IU maintain anti-tumor efficacy while effectively expanding Tregs, which is consistent with our findings. In vitro experiments show that LDIL-2 treatment significantly reduces the increase in Tfh cell proportions caused by anti-PD-1 mAb exposure. In the mouse study, we used an autoimmune mouse model to evaluate the therapeutic effects of LDIL-2 and a tumor-bearing mouse model to assess its impact on antitumor efficacy. In the mouse study, we employed two distinct models to unequivocally address specific biological questions: autoimmune-prone MRL/MpJ-Faslpr mice for irAE-like pathology, and standard immunocompetent C57BL/6J mice with LLC tumors for antitumor efficacy. This dual approach was essential because C57BL/6J mice do not consistently develop robust irAEs within the experimental timeframe, and rapid tumor growth would confound irAE assessment in a single model. Our mouse study revealed that anti-PD-L1 mAb promotes a shift in the Tfh/Treg balance toward Tfh cells in circulation while also inducing inflammatory lesions in the kidneys, lungs, and skin, along with an increase in autoantibody production, suggesting the occurrence of irAEs. LDIL-2 effectively reversed this Tfh/Treg imbalance, reducing both organ inflammation and autoantibody production associated with anti-PD-L1 mAb, thereby alleviating the onset of irAEs. Notably, the use of LDIL-2 did not significantly compromise the antitumor efficacy. Current treatments for irAEs mainly rely on corticosteroids and other immunosuppressive drugs. Corticosteroids serve as the initial treatment, particularly for moderate to severe irAEs, with dosage tailored according to the severity. If corticosteroids prove inadequate, alternative immunosuppressive agents such as mycophenolate, infliximab, or tocilizumab are introduced [40]. However, excessive use of immunosuppressants can lead to various side effects, including severe infections and a reduced cancer-fighting potential of ICIs. LDIL-2, a milder biologic agent, has fewer adverse effects than conventional immunosuppressive agents and has shown long-term benefits in treating various autoimmune diseases [28, 29]. Our study demonstrated that LDIL-2 shows therapeutic potential for irAEs while preserving the antitumor efficacy of anti-PD-1/PD-L1 therapy, providing a promising new approach for managing irAEs. Although our study contributes to optimizing irAE treatment, there are several limitations. First, although we ensured the comparability of mouse groups in each analysis, the relatively small cohort size underscores the need for more extensive studies to validate our findings. Furthermore, our predefined sample allocation strategy, necessitated by the limited volume of patient PBMCs, resulted in a reduced sample size for individual assays (e.g., RNA-seq and flow cytometry). While this approach optimized data quality for each specific analysis, it inherently limited the statistical power for these individual assessments. Secondly, due to limited sample volumes and the observational nature of the clinical cohort, we did not perform in vitro functional assays on Tregs. Future research should incorporate functional evaluations to further elucidate how LDIL-2 restores immune balance. Thirdly, the MRL/MpJ-Faslpr mouse model, while effective for modeling irAEs susceptibility, possesses an inherent inflammatory genotype that may amplify immune responses, thus representing a specific limitation in generalizing the magnitude of observed effects. Fourthly, there is currently no human evidence supporting our results regarding the efficacy of LDIL-2 in treating irAEs. Further clinical trials to confirm the effectiveness of LDIL-2 in patients with irAEs could provide valuable evidence for the clinical management of irAEs and offer guidance for healthcare professionals. Fifthly, our study does not compare LDIL-2 with other immunosuppressive agents regarding efficacy, safety, and their effects on the antitumor activity of ICIs in treating irAEs. Future research should include comparative studies of LDIL-2 against other immunosuppressive therapies, such as mycophenolate mofetil, tocilizumab, and tofacitinib, to evaluate their relative efficacy and safety profiles. In conclusion, our study highlights the critical role of the Tfh/Treg balance in the development of irAEs and confirms the therapeutic potential of LDIL-2 in mice. LDIL-2 reverses the Tfh/Treg imbalance and reduces tissue inflammation and autoantibody production. While mitigating autoimmune responses, it does not significantly diminish the antitumor efficacy of anti-PD-1/PD-L1 therapy, making it a promising alternative to traditional immunosuppressive treatments for irAEs and offering a new direction for the clinical management of irAEs. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (PDF 635 KB) Acknowledgements We acknowledge all the participants and their families. Abbreviations B-cell lymphoma 6 Common Terminology Criteria for Adverse Events C-X-C chemokine receptor type 5 Forkhead box protein P3 Immune checkpoint inhibitors Inducible T-cell co-stimulatory Immune-related adverse events Low-dose interleukin-2 Monoclonal antibodies Peripheral blood lymphocytes Programmed cell death-1 Programmed cell death ligand-1 RNA sequencing Sjögren’s syndrome Systemic lupus erythematosus Follicular helper T cells Regulatory T cells Author contributions Y.W. and J.H. contributed to the conception and design of the study, initiation of the investigation, data acquisition and analysis, writing, reviewing, and editing of the manuscript, and led the clinical experiments. Z.Z., Z.P., contributed to sample collection, the study’s conception and design, and the manuscript’s writing and editing. Y.D.L. contributed to the design of the study, data analysis, and manuscript revision. H.L., Y.C., M.W. and Y.F.L. participated in the clinical experiments and reviewed and edited the manuscript. Funding This work was supported by grants from the Research Ward Excellence Clinical Research Program (BRWEP2024W034080206), National Key R&D Program of China (2022YFE0131700 and 2022YFC3602000), National Natural Science Foundation of China (82071813 and 82271835), and Beijing Nova Program (20220484206). Data availability Detailed extracted data on all included studies are available from the corresponding authors upon reasonable request. Declarations Conflicts of interest The authors declare no potential conflicts of interest in this work. Ethics approval Human study was conducted following the Declaration of Helsinki and received approval from the ethics committee of Peking University Cancer Hospital (2022KT125). Animal studies received ethical approval from the Institutional Animal Care and Use Committee (IACUC) of Peking University Health Science Center (2021PHE026). Consent to participate Patients signed written informed consent to collect biospecimens and anonymously use their clinical and molecular data for research purposes. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary file1 (PDF 635 KB) Data Availability Statement Detailed extracted data on all included studies are available from the corresponding authors upon reasonable request.