The study cohort comprised 25 pathologically confirmed patients (21 male; median age 52 years) with HNSCC who were being treated with pembrolizumab or tislelizumab, a biologic targeting programmed cell death 1 (PD‐1) (Figure 1A). No patient had a pre‐existing autoimmune disease. The median interval between the onset of ICI and the onset of dry mouth was 3 months (Table 1).

To quantitatively assess the impact of ICI therapy on exocrine gland function, we prospectively monitored salivary and lacrimal secretion in the 25 patients before and after treatment. After treatment, the total unstimulated salivary flow (UWSF) was significantly reduced compared with the pre-treatment levels (P < 0.05), with decreased secretion observed in 64% of patients (16/25) (Figure 1B). Notably, all treated patients exhibited UWSF rates below the diagnostic threshold for hyposalivation (1.5 mL/15 min), with a median of 0.45 mL/15 min (range: 0–2.49 mL/15 min), indicating severe resting-state salivary hypofunction. Similarly, Stimulated Whole Salivary Flow (SWSF) following exposure to 3% citric acid exhibited significant post-treatment reduction in 72% of patients (18/25) (P < 0.05) (Figure 1C). A common clinical sign among the enrolled patients was severe xerostomia, characterized by the absence of salivary pooling and an atrophic tongue surface (Figure 1D). Beyond salivary dysfunction, lacrimal gland function, assessed by the Schirmer’s test, was also significantly impaired. Post-treatment tear secretion demonstrated marked bilateral reduction compared with baseline measurements in 68% of patients (17/25) (P < 0.05), providing an objective physiological basis for xerophthalmia (Figure 1E).

From a subjective assessment perspective, we employed the Multidisciplinary Salivary Gland Society (MSGS) scale (30) as our primary evaluation tool. Despite the lack of significant changes in xerostomia scores on the MSGS scale (Supplementary Figure S1), subsequent clinical interview assessments revealed a substantial symptomatic burden reported by patients. Xerostomia was often exacerbated during physical activity or at night, and some individuals experienced nocturnal awakening due to tongue–palate adhesion, which required frequent water intake. Additional complaints included viscous saliva, dry throat with hoarseness, dysgeusia (altered taste perception), and reduced tolerance to spicy or acidic foods.

To elucidate the pathological basis of glandular hypofunction after ICI therapy, we analyzed salivary gland specimens (submandibular gland, SMG; sublingual gland, SLG; and parotid gland, PG) from 50 patients treated with and without ICI. The baseline information is shown in Table 2. The acinar marker aquaporin-5 (AQP5) (31) and the ductal marker cytokeratin 7 (CK7) were analyzed, reflecting secretory function and ductal architecture, respectively. The ICIs-treated group and the untreated group showed different patterns: In the untreated group, AQP5 and CK7 precisely delineated the acinar and ductal structures (Figures 2A, B). In contrast, glands from the ICIs group exhibited widespread atrophy and loss of AQP5-positive acini (Figures 2A, C, D). Quantitative analysis confirmed a significant downregulation of the expression of the AQP5 protein in the ICI group compared with that of controls, whereas the CK7 protein remained unchanged (Figure 2E). These findings suggest that ICI therapy causes selective damage to acinar cells, thereby compromising glandular secretory function, while the ductal structures remained relatively intact.

The destruction of the acinar and duct appeared to be the result of lymphocyte infiltration. A striking example was observed in a patient who had received five cycles of ICI therapy, and whose submandibular gland showed near-total acinar destruction, with remnant ducts engulfed by diffuse lymphocytic infiltrates (Figure 2D). Hematoxylin and Eosin (H&E) staining revealed prominent focal lymphocytic infiltrates in glands of the ICI group, predominantly in periductal areas, which were absent in controls (Figure 2F). Semi-quantitative scoring confirmed this finding (P < 0.05) (Figure 2G).

To assess tissue remodeling, Masson’s trichrome staining was performed. This revealed extensive deposition of dense collagen fibers, indicative of significant fibrosis, in the interlobular and periductal stroma of the ICI group, a feature not observed in controls (Figures 2H, I).

Collectively, these findings demonstrate that ICI therapy induces periductal-centric lymphocytic sialadenitis accompanied by severe fibrosis. This pathological process culminates in the destruction of functional acinar-ductal units, which is responsible for the progressive loss of secretory function.

The pathological profile of sialadenitis is etiology-dependent and requires a careful differential diagnosis. Therefore, to identify the underlying cause of this ICI-induced-condition, further elucidation of its pathological characteristics is required. To characterize the immune composition of ICI-induced sialadenitis, we performed immunohistochemistry and immunofluorescence staining of tissue sections in 3 major glands. The results revealed that lymphocytic foci were densely populated by CD3⁺ T cells and CD19⁺ B cells were scattered only sparsely and individually, failing to form aggregates (Figures 3A, B).

Further analysis of T cell subsets revealed that the distribution of CD4⁺ T cells within the foci closely mirrored that of CD3⁺ T cells, identifying them as the core component of the infiltrated lymphocyte. Compared with the untreated group, the glands of the ICI group showed a significant increase in the density of CD3⁺ and CD4⁺ T cells (P < 0.05). However, the densities of CD8⁺ T cells and CD19⁺ B cells did not differ significantly between groups. Moreover, although CD8⁺ cytotoxic T cells were present, they were diffusely scattered and did not form dominant clusters within the foci (Figure 3C). Quantitative analysis corroborated these observations (Figure 3D). To identify the immune cell population driving glandular damage, we conducted multiplex immunofluorescence which revealed a striking spatial association: dense CD4⁺ T cell infiltrates consistently corresponded with a profound loss of AQP5⁺ acini. This association was not apparent for CD8⁺ T cells (Figure 3E). Quantitative analysis confirmed this observation, demonstrating a significant negative correlation between CD4⁺ T cell density and AQP5-positive area (Spearman’s ρ = -0.9588, P < 0.0001) (Figure 3F). These findings pinpoint these CD4⁺ T cells as the primary mediators of acinar destruction in this disease. To support our histological findings and quantitatively define the infiltrating leukocyte populations, we performed flow cytometry on single-cell suspensions isolated from salivary gland specimens (Figure 4A). Due to the limited availability of patient samples, the samples were restricted to the submandibular gland. In particular, the submandibular gland contributes about 70% of total saliva secretion, highlighting its predominant role in salivary function and the rationale for focusing our analysis on this gland. Consistent with our IHC data, flow cytometric analysis revealed a significant increase in the percentage of CD3⁺ and CD4⁺ T cells in the ICI group compared with that of controls (P < 0.05). (Figure 4B).

Therefore, immunotherapy-induced sialadenitis presents a unique immune profile: a CD4⁺ T cell-driven, B cell-quiescent, non-myeloid adaptive immune response.

Given the predominance of CD4⁺ T cells, we further examined their main functional subsets, including Th1, Th2, Th17, and regulatory T cells (Tregs) (Supplementary Figure S2). The markers (Th1: CXCR3⁺CCR6^-; Th17: CCR6⁺ CXCR3^-; Th2: CXCR3^-CCR6^-; Treg: CD25⁺Foxp3⁺) provided critical insight. Among all the subsets of CD4⁺ T cells analyzed, only the population of Th17 cells was significantly and specifically expanded in the ICI group (P < 0.05) (Figure 4B). In contrast, the proportions of Th1, Th2, and Treg cells did not differ between the groups (Figure 4B).

To determine whether this expansion of Th17 cells was functionally relevant, we assessed the expression of their hallmark effector cytokine, IL-17A, at the transcript, protein and tissue levels. First, quantitative real-time PCR (qPCR) analysis revealed a significant upregulation of IL17A mRNA in the ICI group (P < 0.05) (Figure 4C). In contrast, the signature cytokine mRNA levels for Th1 (IFN-γ), Th2 (IL-4), and Tregs (IL-10, TGF-β) were comparable between the groups (Figure 4C). This finding was corroborated at the protein level by ELISA, which demonstrated significantly higher concentrations of secreted IL-17A in tissue homogenates from the ICI group (P < 0.05) (Figure 4D). Finally, to link IL-17A production with infiltrating lymphocytes in situ, we performed immunofluorescence staining. The imaging revealed a clear co-localization of the IL-17A protein with infiltrating CD4⁺ T cells, confirming these cells as the primary source of cytokines within the tissue (Figure 4E). Nevertheless, given the therapeutic target of ICIs, we examined the expression of programmed death-1/programmed death-ligand 1 (PD-1/PD-L1). Immunohistochemical analysis revealed the absence of PD-L1 in normal ductal and acinar epithelial cells. Within inflammatory infiltrates, although some lymphocytes showed positive PD-1 staining, the intensity was generally mild to moderate (Figure 4F). Furthermore, the staining for myeloid cell markers (e.g., macrophages and neutrophils) was negative, indicating that the infiltrate was not driven by innate immune cells (Supplementary Figure S3).

To establish a preclinical model that recapitulates our clinical findings, we administered an anti-PD-1 antibody to mice bearing tongue squamous cell carcinoma. Consistent with our observations in patient biopsies, flow cytometric analysis of SMGs from these mice revealed a significant increase in infiltrating CD45⁺ leukocytes. Within this CD45⁺ population, we observed a significant expansion of the CD3⁺ T cell compartment, with no change in B cell frequency. Furthermore, the T cell population displayed a pronounced skewing towards a CD4⁺ dominant phenotype (Figures 5A, B). Histopathological examination further confirmed this parallel, showing massive lymphocytic infiltration and severe destruction of acinar architecture, mirroring the pathology seen in our patient cohort (Figure 5C).

Given the CD4⁺ T cell dominance, we hypothesized that IL-17A plays a key role in this process. Indeed, anti-PD-1 treatment led to a robust upregulation of Il17a gene expression in the SMGs (Figure 5D) and a concomitant, significant reduction in salivary flow (Figure 5E). To test the functional relevance of this pathway, we administered a neutralizing anti-IL-17A antibody. This intervention not only attenuated the upregulation of IL-17A in the gland tissue but also led to a significant recovery of saliva production compared to the anti-PD-1 group (Figures 5D, E). Taken together, these data establish a causal link between PD-1 blockade, IL-17-mediated immunopathology, and salivary gland dysfunction, highlighting the IL-17 axis as a promising therapeutic target to ameliorate this immune-related adverse event.

In conclusion, these findings demonstrate a selective expansion of Th17 cells accompanied by enhanced IL-17A production, implicating this pathway as a key mediator of ICI-induced salivary gland inflammation.

We prospectively enrolled patients 18 years or older with a confirmed diagnosis of HNSCC who were treated with pembrolizumab and/or nivolumab at our institution between 2023 and 2025. Exclusion criteria included (1): concomitant radiotherapy (2); a history of specific chronic sialadenitis, such as SjS or IgG4-related disease; and (3) concurrent use of medications known to significantly affect salivary or lacrimal function. The institutional ethics committee approved the study protocol and all amendments of the Hospital of Stomatology of Sun Yat-sen University, China (Grant Number: ERC- [2015]-29) in accordance with the Declaration of Helsinki. Written informed consent was obtained from the patients for the publication of any potentially identifiable images or data included in this article.

Salivary gland tissue samples were obtained intraoperatively from patients who underwent cervical lymph node dissection as part of their surgical treatment. Immediately after excision, each fresh specimen was divided. One portion was placed in ice-cold RPMI 1640 medium and transported on ice for subsequent experiments. Another portion was fixed in 4% paraformaldehyde for 24–48 h at 4°C, followed by standard processing for dehydration and paraffin embedding. The resulting paraffin-embedded blocks were sectioned into 4 μm-thick slices, which were then mounted on glass slides for subsequent analysis. A third portion was snap-frozen in liquid nitrogen and stored at -80°C for future molecular analyses.

All functional assessments were performed at the beginning of the study (before ICI treatment) and at designated follow-up time points. To minimize diurnal variations, saliva samples were collected between 9:00 a.m. and 11:00 a.m. Patients were instructed to fast for at least one hour prior to collection. The flow rate of unstimulated whole saliva (UWS) was measured using the passive drool method over a 15-min period in a pre-weighed sterile tube. Subsequently, the flow rate of the stimulated whole saliva (SWS) was assessed for 15 min, and salivary secretion was stimulated by applying 3% citric acid to the tongue tip every 3 min. Both UWS and SWS flow rates were calculated and expressed in mL/min. Lacrimal secretion was quantified using the standard Schirmer’s test. A sterile filter paper strip was placed in the lower conjunctival fornix of each eye for 5 min. The length of the moistened portion of the strip was measured in millimeters (mm).

Before staining, paraffin-embedded sections (4 μm) were deparaffinized and rehydrated. Briefly, slides were baked at 65°C for 15 min, immersed in xylene, and then passed through a gradient of ethanol. (100%, 95%, 80%, 70%) to distilled water. H&E staining was performed using a commercial kit (Solarbio, Cat# G1120), following the manufacturer’s instructions. Masson’s trichrome staining was conducted similarly (Solarbio, Cat# G1340). Following staining, all sections were dehydrated through a graded ethanol series, cleared in xylene, and cover slipped with a neutral mounting medium. H&E-stained sections were used for histopathological assessment. The severity of lymphocytic infiltration in salivary glands was graded according to the Chisholm and Mason scoring system as follows: Grade 0: no inflammatory infiltrate; Grade 1: mild, diffuse lymphocytic/plasmocytic infiltrate; Grade 2: moderate infiltrate or less than one focus; Grade 3: one focus; Grade 4: more than one focus. A “focus” was defined as an aggregate of at least 50 mononuclear cells per 4 mm² of tissue. For morphometric analysis, whole-slide images (WSIs) were acquired using a digital slide scanner. Quantitative analysis was then performed on non-overlapping, randomly selected fields from each WSI using ImageJ software (NIH, USA).

Before staining, paraffin-embedded sections were deparaffinized in xylene and rehydrated through a graded ethanol series. Heat-induced epitope retrieval (HIER) was performed by incubating sections in sodium citrate buffer (10 mM, pH 6.0) in a pressure cooker for 5 min, followed by natural cooling to room temperature. Following antigen retrieval, endogenous peroxidase activity was quenched by incubating sections in 3% H₂O₂ for 15 min. After blocking non-specific binding, sections were incubated overnight at 4°C with the following primary antibodies: anti-AQP5(abcam ab315855), anti-CK7(abcam ab181598), anti-CD19(HUABIO ET1702-93), anti-CD3(Proteintech 17617), anti-CD4(ZSGB-BIO ZM0418), anti-CD8(ZSGB-BIO ZA0508), anti-CD68(CST 76437), anti-CD16(CST 24326),anti-PD-1(CST 86163),anti-PD-L1(CST 13684), and anti-IL-17A(abcam ab79056). The next day, sections were incubated with a horseradish peroxidase (HRP)-conjugated secondary antibody. The signal was visualized using a DAB (3,3’-diaminobenzidine) substrate kit. Finally, sections were counterstained with hematoxylin, dehydrated through a graded ethanol series, cleared in xylene, and cover slipped with a neutral mounting medium. For immunofluorescence staining, sections were blocked with 5% bovine serum albumin (BSA) for 60 min at room temperature. Sections were then incubated overnight at 4°C with the desired primary antibodies. After washing with PBST (PBS containing 0.1% Tween-20), sections were incubated with an Alexa Fluor 594-conjugated secondary antibody (1:1000) for 1 h in a dark, humidified chamber. Nuclei were counterstained with Hoechst. Slides were then mounted using an anti-fade mounting medium. Multiplex immunofluorescence staining was performed on 4-um FFPE sections using a commercial kit (Boster, Wuhan, China, Cat# PTSA-45) according to the manufacturer’s instructions. IHC-stained slides were digitized using a whole-slide scanner (Leica, Germany). Immunofluorescence images were captured using an Olympus motorized fluorescence microscope. Quantitative analysis was performed using ImageJ software (NIH, USA). For each marker, the extent of positive staining was quantified by calculating the percentage of the positively stained area relative to the total tissue area within each analyzed field.

Total protein was extracted from SG tissue using RIPA lysis buffer supplemented with a phosphatase and protease inhibitor cocktail. The lysates were cleared by centrifugation at 12,000 ×g for 15 min at 4°C, and the resulting supernatants were collected. The protein concentration was determined using a BCA Protein Assay Kit (Thermo Fisher Scientific).

For electrophoresis, equal amounts of protein (20 µg per lane) were resolved by SDS-PAGE and subsequently transferred to polyvinylidene difluoride membranes (PVDF) (Millipore, Billerica, MA, USA). Following transfer, the membranes were blocked with 5% non-fat dry milk in Tris-buffered saline containing 0.1% Tween-20 (TBST) for 1 h at room temperature. Membranes were then cut according to the expected molecular weights of the target proteins. The cropped membranes were incubated overnight at 4°C with specific primary antibodies: anti-AQP5, anti-CK7, anti-β-actin(Proteintech 66009). After washing with TBST, the membranes were incubated with appropriate HRP-conjugated secondary antibodies for 1 h at room temperature. Protein bands were visualized using an enhanced chemiluminescence (ECL) substrate (SuperSignal West Pico, Thermo Fisher Scientific) and imaged using a LI-COR Odyssey Clx Imaging System.

Total RNA was extracted from SG tissue samples using the FastPure^® Cell/Tissue Total RNA Isolation Kit V2 (Vazyme, Nanjing, China) according to the manufacturer’s protocol. The concentration and purity of the extracted RNA were determined using a spectrophotometer. First-strand cDNA was synthesized from total RNA using the HiScript^® III Reverse Transcriptase kit (Vazyme). RT-qPCR was performed using ChamQ Universal SYBR qPCR Master Mix¹ (Vazyme) on a QuantStudio™ 7 Flex Real-Time PCR System (Thermo Fisher Scientific, USA). The thermal cycling conditions were as follows: initial denaturation at 95°C for 5 min, followed by 40 cycles of 95°C for 15 s and 60°C for 60 s. A melt curve analysis was performed at the end of each run to confirm the specificity of the amplified products. The primers used for this study were: for human samples IL-17 5′-CTCTGTGATCTGGGAGGCAAA-3′ 5′- CTCTTGCTGGATGGGGACA-3′, IL-4 5′-CCAACTGCTTCCCCCTCTG-3′ 5′-TCTGTTACGGTCAACTCGGTG-3′, IL-10 5′-TCTCCGAGATGCCTTCAGCAGA-3′ 5′- TCAGACAAGGCTTGGCAACCCA-3′, IFN-Y 5’-TCGGTAACTGACTTGAATGTCCA-3’ 5’- TCGCTTCCCTGTTTTAGCTGC-3’, TGF-b 5’-GGCCAGATCCTGTCCAAGC-3’ 5’- GTGGGTTTCCACCATTAGCAC-3’, GAPDH 5′- GGAGCGAGATCCCTCCAAAAT-3′, and 5′- GGCTGTTGTCATACTTCTCATGG-3′. For mouse samples IL-17 5′- AAGCTGGACCACCACATGAA-3′ 5′- CCCTGAAAGTGAAGGGGCAG-3′, GAPDH 5′- AACTTTGGCATTGTGGAAGGG-3′, and 5′- GACACATTGGGGGTAGGAACA-3′.The relative expression levels of target genes were normalized to the endogenous control, GAPDH. The fold change in gene expression was calculated using the comparative Ct (2^-ΔΔCt) method.

The method for preparing proteins from salivary gland samples was the same as that described above. IL-17A concentration was quantified using a commercially available Human IL-17A ELISA kit (feiyue, FY-EH4665) following the manufacturer’s protocol. Briefly, absorbance was measured at 450 nm using a microplate reader (Biotek, USA). The concentration of IL-17A in each sample was calculated from a standard curve generated with recombinant IL-17A standards. The final results were normalized to the total protein content and are expressed as picograms of IL-17A per milligram of total protein (pg/mg).

SCCVII cells were maintained in DMEM/F12 medium with 10% fetal bovine serum and 0.4% penicillin-streptomycin solution, which we obtained from the Peking University School and Hospital of Stomatology as a gift. To obtain SCCVII, a murine squamous cell carcinoma derived from immunocompetent mice C3H/HeN (C3H), we purchased female C3H mice (4–6 weeks of age) from Beijing Vital River Laboratory Animal Technology Co., Ltd and fed them under specific pathogen-free conditions at the Laboratory Animal Center of Sun Yat-sen University. SCCVII cells were used for tumor inoculation when they reached the exponential growth phase. For tumor models, 100 μL of FBS-free DMEM/F12 containing 5×10⁵ cells were injected into the shaved right flanks of C3H mice subcutaneously. When the tumor volume reached 50 mm³, tumor models were treated with PBS (untreated group) or antimouse-PD-1 (treated group)(Selleck,Houston,USA, clone:RMP1-14). These treatments were administered on days 7, 9, and 11 with the same dose, 200 μg/mouse i.p. All animal experiments were conducted in strict accordance with protocols approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University (SYSU-IACUC-2025-002999).

On day 12 post-tumor inoculation, mice bearing established tumors were randomized into treatment groups. The treatment group received an antimouse-IL-17A neutralizing antibody (Selleck,Houston,USA, clone:17F3, dose: 100 μg/mouse), while the control group received a corresponding isotype control antibody (Selleck,Houston,USA, clone:MOPC-21, dose: 100 μg/mouse). The treatment regimen consisted of three intraperitoneal (i.p.) injections administered at 3-day intervals.

At the study endpoint on day 33, saliva flow rate was measured. Mice were first anesthetized with an i.p. injection of sodium pentobarbital (50 mg/kg) and placed in a prone position. Salivation was then stimulated with pilocarpine hydrochloride (2 mg/kg). Following a 10-minute latency period, whole stimulated saliva was collected from the oral cavity for 10 minutes using pre-weighed cotton balls. Saliva volume was determined by the change in weight. At the designated experimental endpoints, mice were humanely euthanized. Specifically, euthanasia was induced by an intraperitoneal (i.p.) injection of an overdose of sodium pentobarbital (150 mg/kg), followed by cervical dislocation as a secondary physical method to ensure irreversible death. The submandibular glands were promptly harvested, with one lobe fixed in 4% paraformaldehyde for histological analysis and the other lobe snap-frozen in liquid nitrogen for molecular analysis. All animal carcasses were disposed of according to institutional biohazardous waste guidelines.

For clinical samples, single-cell suspensions were prepared from fresh SG tissue. Briefly, tissues were mechanically minced and then enzymatically digested in a solution containing collagenase II (2 mg/mL; Worthington) and DNase I (4 mg/mL; Worthington) for 60 min at 37°C with continuous agitation. The resulting cell suspension was filtered through a 70-μm cell strainer to remove undigested tissue. Following centrifugation (400 ×g, 5 min), red blood cells were lysed by incubating the cell pellet with RBC Lysis Buffer for 1–2 min. The reaction was stopped with excess FACS buffer (PBS containing 2% FBS), and the cells were washed twice before being counted and resuspended for staining. For staining, cells were first incubated with Zombie Violet™ Fixable Viability Kit (BioLegend) for 15 min to enable discrimination of live and dead cells. After washing, non-specific antibody binding was blocked by incubating cells with a human Fc receptor blocking reagent (e.g., Human TruStain FcX™, BioLegend) for 20 min. Subsequently, cells were stained with a cocktail of the following fluorochrome-conjugated surface antibodies¹ for 30 min at 4°C in the dark: FITC anti-human CD45, Brilliant Violet 510™ anti-human CD3, Brilliant Violet 605™ anti-human CD19,Brilliant Violet 650™ anti-human CXCR3,PE anti-human CD25, PE-Cy7 anti-human CCR6, APC-Cy7 anti-human CD4, and Red Fluor 710 anti-human CD8. For intracellular staining of the transcription factor Foxp3, cells were fixed and permeabilized using a Foxp3/Transcription Factor Staining Buffer Set (eBioscience) according to the manufacturer’s protocol. Cells were then stained with PE-Cy7 anti-human Foxp3 (eBioscience).

For tumor-model mice, the bilateral submandibular glands of each mouse were separated, cut into small pieces, and digested in a solution containing collagenase II (2 mg/mL; Worthington) and DNase I (4 mg/mL; Worthington) for 60 min at 37°C with continuous agitation. A single-cell suspension was obtained by grinding tissue blocks, from which we selected live cells after incubation with Zombie dye at 4°C for 20 min. After centrifugation (400 ×g, 5 min), the cells were washed with buffer and blocked with antimouse CD16/32 at 4°C for 30 min. Subsequently, the cells were incubated with surface-staining fluorescent antibodies (PerCP anti-mouse CD45, PE-CY7 anti-mouse CD3, AF700 anti-mouse CD4, PE anti-mouse CD8a, APC anti-mouse CD19) at 4°C for 20 min.

Data were acquired on a BD LSRFortessa™ flow cytometer (BD Biosciences). Compensation and data analysis were performed using FlowJo™ software (v10, BD Life Sciences).

All quantitative data are presented as the mean ± standard error of the mean (SEM). Statistical analyses were performed with GraphPad Prism software (v.9.0, GraphPad Software, La Jolla, CA, USA). Comparisons between two independent groups were conducted using an unpaired two-tailed Student’s t-test. For comparisons among three or more groups, one-way analysis of variance (ANOVA) followed by an appropriate post hoc test (e.g., Tukey’s) was employed. To assess the relationship between two continuous variables between CD4⁺ cells,CD8⁺ cells and AQP5⁺ Cells, correlation analysis was performed. Pearson correlation coefficient (r) was calculated for normally distributed data, while Spearman rank correlation coefficient (ρ) was used for non-normally distributed data. A P-value less than 0.05 was considered statistically significant.