T Cell–Derived IFN-γ Activates Ferroptosis in Lacrimal Luminal Ductal Cells via JAK-STAT Signaling in Sjögren's Disease-Related Dry Eye

Abstract

Purpose

Sjögren's disease (SjD) leads to severe dry eye disease (DED) via lacrimal gland inflammation and dysfunction, yet the specific pathogenic mechanisms remain unclear. In this study, we aimed to deeply identify the vulnerable lacrimal epithelial subclusters and elucidate the immune-epithelial cross-talk.

Methods

We used MRL/lpr mice as a murine SjD model and integrated bioinformatic analysis of single-cell RNA-sequencing (scRNA-seq) and spatial transcriptome datasets. These findings were validated using immunofluorescence (IF), multiplex immunohistochemistry analysis (mIHC), Western blot, transmission electron microscopy (TEM), and flow cytometry. Finally, an in vivo interferon-γ (IFN-γ) neutralization strategy was implemented, with therapeutic effects evaluated by the phenol red thread test and lissamine green staining.

Results

ScRNA-seq and spatial analyses converged to identify luminal ductal epithelia (LDE), marked by CYP2F2, as the primary epithelial population damaged in SjD. This was further correlated with the specific upregulation of both Janus kinase/signal transducer and activator of transcription (JAK-STAT) signaling and ferroptosis pathway. We then confirmed that this periductal damage was driven by the infiltration of IFN-γ–secreting T helper 1 (Th1) cells. TEM analysis validated classical ferroptotic mitochondrial damage specifically in LDE. Systemic IFN-γ neutralization suppressed local Th1 infiltration, which in turn significantly downregulated the JAK2/STAT1 pathway and reversed ferroptosis within LDE. This intervention significantly alleviated DED signs, as evidenced by increased tear secretion and reduced corneal lissamine green staining.

Conclusions

Our findings define a spatially-specific pathogenic axis where infiltrating Th1 cells release IFN-γ to trigger a JAK-STAT and ferroptosis cascade in LDE, highlighting IFN-γ neutralization as a promising therapeutic strategy.

Sjögren's disease (SjD) is a systemic autoimmune disease characterized by pro-inflammatory lymphocytic infiltration of exocrine glands, primarily the lacrimal and salivary glands, leading to their dysfunction.^1–3 The term “Sjögren's disease” is currently adopted as the preferred nomenclature of previous “Sjögren's syndrome”, with respect to the improvement of clarity, precision, and clinical relevance.³ The lacrimal gland (LG) is responsible for secreting the aqueous component of the tear film, a fluid essential for ocular surface health and optimal light refraction.⁴ In SjD, this autoimmune assault on the LG parenchyma leads to severe aqueous-deficient dry eye disease (DED), a condition that has aroused increasing attention due to its complex pathophysiology and high prevalence.⁵ SjD-DED often presents with more severe signs and a higher risk of vision-threatening complications compared to non-SjD DED, as the systemic autoimmune characteristics exert specific damaging effects.⁶ Although the LG is recognized as the primary target in SjD, the precise mechanisms leading to the dysfunction of its secretory parenchyma and the subsequent functional failure are not fully understood, necessitating a more detailed atlas to describe its comprehensive pathology.

Exocrine gland dysfunction in SjD-DED is driven by the progressive loss of parenchymal cells, including both the acinar and ductal epithelia.^7,8 Previous research on SjD-associated exocrine gland damage focus predominantly on the acinar cells as the primary site of damage and tear production failure.^9,10 However, the lacrimal gland is a complex organ comprising both acini and a sophisticated ductal network (including luminal and basal ductal epithelia) responsible for tear modification and transport.¹¹ The role of this ductal system in the initiation and progression of SjD has been underestimated. Growing evidence from single-cell RNA sequencing (scRNA-seq) suggests that luminal ductal epithelial cells (LDE) are not passive victims but active participants in the immune crosstalk,¹² although the specific contribution remains a critical gap in our knowledge.

Concurrently, the mechanism driving this epithelial destruction is being re-evaluated. Apoptosis was long considered the principal mode of cell death in SjD.^13,14 However, recent research has uncovered ferroptosis, a distinct form of regulated cell death initiated by iron-dependent lipid peroxidation, as a novel pathogenic factor in a series of autoimmune and inflammatory diseases.^9,15 Ferroptosis is mechanistically driven by the accumulation of lipid peroxides, a process that is critically prevented by the antioxidant enzyme glutathione peroxidase 4 (GPX4).¹⁶ Although its role is being explored elsewhere, the involvement of ferroptosis in the progressive destruction of the lacrimal gland in the context of SjD remains to be elucidated.

Immune cell infiltration is so far considered a major aspect of lacrimal gland involvement in SjD.^17,18 The pathogenic roles of T cells, especially T helper 1 (Th1) and T helper 17 (Th17) subsets, have been well-documented in SjD, and the inflammatory cytokines they produce are known to mediate tissue damage.^19,20 The Janus kinase/signal transducer and activator of transcription (JAK-STAT) pathway is a critical signaling cascade that acts downstream of key cytokines like interferon-gamma (IFN-γ) and is implicated in driving autoimmune pathology.²¹ Nevertheless, previous research has not yet clearly connected this well-known axis of immune infiltration and cytokine signaling to a specific terminal mechanism of epithelial cell death within the lacrimal gland.

Here, we integrate bioinformatic analysis of both scRNA-seq and spatial transcriptome datasets with in vivo validation in a murine SjD model to explore SjD-related lacrimal gland pathology. Our findings indicate LDE damage depends on IFN-γ derived from infiltrating Th1 cells. We further show this process is driven by IFN-γ-mediated activation of the JAK-STAT pathway, thereby leading to suppression of the key ferroptosis inhibitor GPX4 and subsequent cell death by ferroptosis. Finally, we used systemic IFN-γ neutralization to confirm this pathway's pathogenic role and assess its therapeutic potential target for rescuing this lacrimal gland failure and the SjD-associated DED phenotype.

Method

Animal Model

Female MRL/lpr mice aged 16 weeks were used as the spontaneous SjD model; age-matched female BALB/c mice served as healthy controls. All animals were maintained under specific-pathogen-free conditions (22°C ± 2°C, 12-hour light/dark cycle, free access to food and water). All animal experiments followed the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Ethics Committee of Zhongshan Ophthalmic Center (Guangzhou, China; approval ID: Z2024011).

Tear Secretion Test

To quantify tear production, phenol red threads (Jingming, Tianjin, China) were used. With minimal manipulation, a thread was slid into the inferior conjunctival sac of every mouse and left in situ for 15 seconds. After removal, the distance of color change along the thread was measured; this procedure was performed in triplicate per eye, and the mean value was adopted as the tear secretion result.

Lissamine Green Staining

Corneal epithelial disruption was examined with lissamine green dye. A 1 % lissamine green solution (199583; Sigma-Aldrich, St. Louis, MO, USA) was applied to the inferior conjunctival sac of each mouse. After 30 seconds, excess dye was gently removed. Stained areas were immediately photographed under a microscope equipped with white-light illumination.

Immunofluorescence Staining

After fresh lacrimal glands were fixed in 4% paraformaldehyde (PFA) and made into paraffin sections, samples were deparaffinized, antigen retrieved and permeabilised with 0.1 % Triton X-100 in PBS for 20 minutes and blocked with 3% BSA (Solarbio, Beijing, China) for one hour at room temperature. Primary antibodies—mouse anti-CYP2F2 (Cytochrome P450 family 2 subfamily F member 2) (1:100, sc-374540; Santa Cruz Biotechnology, Dallas, TX, USA), anti-CD4(1:200, ab183685; Abcam, Cambridge, UK), and anti-F4/80(1:200, 70076; Cell Signaling Technology, Danvers, MA, USA) were applied overnight at 4°C. After three washes in PBS, sections were incubated with Alexa Fluor 488- or 594-conjugated secondary antibodies (1:400, 4412, and 8889; Cell Signaling Technology) for one hour at room temperature, then counter-stained with 4′,6-diamidino-2-phenylindole (DAPI, 1:30000, D9542; Sigma-Aldrich) for five minutes. After additional rinses, images were captured on a confocal laser scanning microscope (LSM 980; Zeiss, Oberkochen, Germany).

Terminal Deoxynucleotidyl TUNEL Assay

To assess parenchymal cell death, TUNEL assay was performed using a commercial kit (A111; Vazyme Biotechnology, Nanjing, China) according to the manufacturer's protocol. Briefly, paraffin-embedded lacrimal gland sections were deparaffinized, rehydrated, and underwent antigen retrieval. Sections were incubated with an equilibration buffer for 20 minutes at room temperature. After a PBS wash, the TUNEL reaction mixture was then applied, and sections were incubated in a humidified, dark chamber for one hour at 37°C. After TUNEL staining, anti-CYP2F2 antibody was uploaded overnight, following by secondary antibody incubation and DAPI staining. Finally, slides were imaged using a fluorescent microscope (LSM 980; Zeiss).

Multiplex Immunohistochemical Analysis (mIHC)

MIHC was conducted with a multiplex immunohistochemical kit (AFIHCO24; AFbio, Changsha, China) following the instructions. Briefly, deparaffinized sections were in turn incubated with peroxidase blocking buffer, rinsed with PBS, incubated with 10% goat serum (C-0005; Bioss, Beijing, China) incubated with primary antibody at 4°C overnight, rinsed with PBS, incubated with secondary antibody at room temperature for one hour, rinsed with PBS, incubated with tyramide signal amplification buffer and dye, rinsed with PBS and finally applied for antigen retrieval to complete a cycle. After all tyramide signal amplification dyes were stained, DAPI (D9542, 1:30000, Sigma-Aldrich) was applied and sections were mounted for observation under a Zeiss LSM 980 With Airyscan 2 Microscope. Primary antibodies used included anti-CYP2F2 (1:100, sc-374540), anti-JAK2 (Janus kinase 2)(1:200, 3230; Cell Signaling Technology), anti-pJAK2 (phospho-Janus kinase 2)(1:200, 66245; Cell Signaling Technology), anti-STAT1 (Signal Transducer and Activator of Transcription 1)(1:400, 14994; Cell Signaling Technology), anti-pSTAT1(phospho-Signal Transducer and Activator of Transcription 1) (1:200, 9167; Cell Signaling Technology), anti-FTH1 (Ferritin Heavy Chain 1) (1:2000, ab65080; Abcam), anti-TFRC (Transferrin Receptor) (1:2000, 13208; Cell Signaling Technology), and anti-GPX4 (1:200, ab125066; Abcam). For the measurement of mean intensity, we referred to the previous research,²² and the acinar and ductal epithelia were demarcated and separated, and the threshold was adjusted based on the default detection to cover the fluorescent area. The mean gray value was calculated and adopted for the statistical analysis.

Western Blot Analysis

For lacrimal gland tissue, both globes from one mouse were pooled to form a single replicate (n = 3 mice/group). After rinsing with ice-cold PBS, specimens were lysed on ice for 10 minutes in 200 µL RIPA (Radio-Immunoprecipitation Assay) buffer (P0013B; Beyotime, Shanghai, China) supplemented with 1% protease inhibitor cocktail (P1006; Beyotime). Lysates were cleared by centrifugation at 15,000g, 4°C, five minutes. Protein concentrations were determined with a BCA kit (P0010; Beyotime). Equal amounts of protein were mixed with 4 × SDS loading dye (P0015L; Beyotime) and heated at 100°C for 10 minutes. Samples were loaded on 10% SDS-PAGE gels for one hour at 130 V and then transferred to 0.22 µm polyvinylidene fluoride membranes (Millipore, Sigma-Aldrich) for one hour at 300 mA. Membranes were blocked with QuickBlock (P0252; Beyotime) for 15 minutes at room temperature, followed by overnight incubation at 4°C with the indicated primary antibodies: β-actin (1:3000, 66009-1-Ig; Proteintech, Rosemont, IL, USA), FTH1 (1:1000, ab65080; Abcam), ferritin light chain (FTL) (1:1000, ab69090; Abcam), STAT1 (1:1000, 14994; Cell Signaling Technology), GPX4 (1:1000, ab125066; Abcam), and ACSL4 (1:1000, 22401-1-AP; Proteintech). After three washes in TBST, membranes were incubated for one hour at room temperature with horseradish peroxidase–conjugated goat anti-rabbit (1:5000, SA00001-2; Proteintech) or anti-mouse (1:5000, SA00001-1; Proteintech) secondary antibodies. Immunoreactive bands were detected with enhanced chemiluminescence reagent (Beyotime) and quantified using ImageJ software (National Institutes of Health, Washington, DC, USA).

Flow Cytometry

Extra-orbital lacrimal glands were dissected and minced into 1 mm³ fragments, and enzymatically digested for 30 minutes at 37°C under gentle agitation in 2 mL digestion medium (RPMI-1640 [C11875500BT; Gibco, Thermo Fisher Scientific, Waltham, MA, USA] containing 2 mg mL⁻¹ collagenase IV [17104-019; Gibco, Thermo Fisher Scientific]). Digests were triturated through an 18 G needle, filtered via a 70 µm mesh, and washed twice with ice-cold FACS buffer (PBS, 2% fetal bovine serum [A5256701; Gibco], 2 mM EDTA [E9884; Sigma-Aldrich]). Erythrocytes were lysed with ammonium-chloride-potassium buffer (A1049201; Gibco) for two minutes on ice. For surface markers, V500-anti-CD45 (561487; BD Bioscience, San Jose, CA, USA), PerCP-Cy5.5-anti-CD3 (560527; BD Bioscience), Fluorescein Isothiocyanate-anti-CD4 (100406; Biolegend, San Diego, CA, USA), BV421-anti-CD25 (562606; BD Bioscience) and flexible viability 780 (565388; BD Bioscience) were applied. For intracellular cytokines, cells were stimulated with 50 ng mL⁻¹ PMA + 1 µg mL⁻¹ ionomycin + 5 µg mL⁻¹ brefeldin A (00-4970-93; Invitrogen, Carlsbad, CA, USA) for four hours, fixed/permeabilized, and stained with APC-anti-IFN-γ (505810; Biolegend) and PE-anti-IL-17A (506904; Biolegend). After being rinsed with 1 × PBS with 2% fetal bovine serum, samples were loaded to a BD LSRFortessa x-20 flow cytometer (BD Bioscience). For the gating strategies, FSC-A and SSC-A were applied to draw an initial range to exclude cell debris and FSC-A and FSC-H were used to subset single cells. Then, with the viability dye, all positive stained cells were labeled dead and excluded. Afterward, CD4+T cells were extracted according to a CD45+CD3+CD4+ phenotype. Th1, Th17 and regulatory T cells (Treg) were respectively recognized with IFN-γ, IL-17A, and CD25. Above analysis were performed on Flowjo (v10, BD Bioscience). All data without low quality and other abnormalities including excessive debris and severe deviation were retained for further statistical analysis.

Transmission Electron Microscopy (TEM)

For ultrastructural analysis, lacrimal glands of control and SjD groups were first fixed overnight at 4°C in 2.5% glutaraldehyde dissolved in 150 mM sodium cacodylate (pH 7.4). Post-fixation was carried out with 1 % osmium tetroxide, followed by staining with uranyl acetate. Samples were then dehydrated through an ascending ethanol series and infiltrated with epoxy resin. Ultrathin sections were cut and examined with a Hitachi HT7700 transmission electron microscope (Hitachi, Tokyo, Japan).

ScRNA-Seq Data Analysis

We adopted a public scRNA-seq dataset (GSE132420) of lacrimal gland¹² of preclinical SjD murine model. After quality control, downstream analysis was performed on R platform (version 4.4) through Seurat (version 5.0.1, https://github.com/satijalab/seurat). Differentially expressed genes (DEGs) were calculated with default parameters with function “FindMarkers” in a subcluster, and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis was performed with the DEGs through ClusterProfiler (version 4.10.1, https://github.com/YuLab-SMU/clusterProfiler) with a threshold of adjusted P value <0.05. For the Venn picture, the acinar, luminal ductal and basal ductal epithelia were respectively subset, and DEGs of these three subclusters of SjD group compared to the control group with the function “FindMarkers” provided by Seurat with the default parameters were calculated. Then, three tables of upregulated DEGs were obtained and a Venn diagram to show the overlapping upregulated DEGs in these parenchymal clusters to investigate the common injured traits was drawn with ggVennDiagram (version 1.5.4, https://github.com/gaospecial/ggVennDiagram). For the enriched pathways of LDE in the SjD group, we extracted the LDE cluster, and applied “FindMarkers” to calculate the DEGs with the parameters group.by = “orig.ident”, ident.1 =“SjD”, ident.2 = “CT”, and obtained both upregulated and downregulated ones in this cluster. Then, the upregulated and downregulated DEGs were respectively applied for KEGG analysis through clusterProfiler package, and enriched pathways of upregulated DEGs and downregulated DEGs were exhibited. Pathway scores were calculated through “AddModuleScore” function in the Seurat package, and genesets were from Gene Ontology database with the accession number “GO:0046425’, “GO:0097707” and “GO:0007259”. The scCustomize (version 2.1.2, https://github.com/samuel-marsh/scCustomize) was used for visualization. Cell-cell communication analysis was conducted on CellChat (version 1.6.1, https://github.com/jinworks/CellChat) to describe the overall interaction strength of each cluster. The “netVisual_heatmap” function of Cellchat was applied, with the given parameter (measure = “weight”,color.heatmap = “Reds”). Depth of redness represented strong probability of interactions with other subclusters for a specific group.

Spatial Transcriptome Analysis

We used a published spatial transcriptome dataset (GSE210380) of lacrimal gland from control and a SjD model.²³ Based on the provided integrated data from the research, we performed re-clustering, evaluation of pathway score with given datasets (Supplementary Table S1), and spatial projection with Seurat (version 5.0.1, https://github.com/satijalab/seurat) on R platform. For the pathway score measurement, the genesets of acute inflammatory response (GO:0002526), chronic inflammatory response (GO:0002544), response to typeII interferon (GO:0034341), positive regulation of receptor signaling pathway via JAK-STAT (GO:0046427) were used, and the detailed genes applied were listed in the newly added Supplementary Table S1.

Statistics

Statistical analyses were performed using GraphPad Prism 8.0 software (GraphPad Software Inc., San Diego, CA, USA). Student's t-test was used to make comparisons between two individual groups. Multiple comparisons were conducted with a one- or two-way ANOVA. Summary data were presented as mean ± SD. A P-value <0.05 was defined as statistical significance.

Result

ScRNA-Seq Identified That LDE Exhibits Distinct Ferroptosis via JAK-STAT Signaling in SjD

To dissect the cellular heterogeneity and specific targets in the SjD lacrimal gland, we adopted a well-established Sjögren disease model and a corresponding published lacrimal gland scRNA-seq dataset. Consistent with previous reports, we obtained a similar annotation of each cluster in the lacrimal gland (Fig. 1A), and noticed decreased proportion of lacrimal functional parenchymal cells in the SjD group (Fig. 1B). Among them, the LDE had the most obvious changes in cell-cell interaction strength (Fig. 1C) and thus served as a highly-involved cluster in SjD related injuries. This loss of LDE was corroborated by immunofluorescence staining for the specific marker CYP2F2 (Supplementary Fig. S1A), which showed significant epithelial depletion and broken morphological continuity in the lacrimal gland of SjD group (Fig. 1D).

Figure 1.

ScRNA-seq identified injury of LDE. (A) UMAP of lacrimal gland of CT and SjD groups in the scRNA-seq dataset. (B) Proportion of each cluster in both groups indicated LDE loss and T cell infiltration. (C) Cell-cell communications showed increased strength in LDE measured by weight. (D) IF confirmed LDE loss in the lacrimal gland with statistical analysis (Scale bar: 50 µm, n = 3, *P < 0.05). (E) Venn diagram of DEGs of acinar, luminal ductal, and basal ductal cells, respectively, between CT and SjD groups. (F) KEGG pathways of crossed upregulated DEGs of acinar, luminal ductal and basal ductal cells between CT and SjD groups. (G) Dotplot of marker genes across all subclusters. (H) Representative up-regulated KEGG pathways of LDE. (I) Representative down-regulated KEGG pathways of LDE. (J) Pathway score showed high level of ferroptosis in LDE and JAK-STAT in T cells.

We next investigated the common changes in several major parenchymal clusters, including luminal ductal cells, basal ductal cells, and acinar cells (Fig. 1E), and recognized 200 overlapping genes. Such genes were enriched in pathways related to inflammatory response and antigen presentation (Fig. 1F). To specify the dominating pathogenic factor, we found in the LDE, markers related to ferroptosis (Acsl4) and JAK-STAT signaling (Stat1) had distinct enrichment in the parenchymal clusters (Fig. 1G). As a result, we further analyzed possible changes in the LDE. KEGG analysis of upregulated pathways included ferroptosis and JAK-STAT signaling, while downregulated pathways included protein export, which is the main function of LDE (Figs. 1H, 1I). Module scoring further indicated that the LDE subset possessed the highest ferroptosis pathway score among all parenchymal cells (Fig. 1J).

Spatial Transcriptomics Maps IFN-γ–Driven Immune Microenvironment and JAK-STAT Signaling in LDE

To elucidate the spatial crosstalk between infiltrating immune cells and their parenchymal targets in the context of SjD, we re-analyzed a published spatial transcriptome database. Based on the given markers, we obtained similar sub-clusters, with further division of ductal cells into luminal ductal cells and basal ductal cells (Fig. 2A). With Cyp2f2 and Krt5, we could find basal ductal cells tended to be located in the base and the luminal ductal cells were proximal to the lumen (Fig. 2B; Supplementary Fig. S1B). Top markers of each cluster in both control and SjD group were shown (Fig. 2C), and like the results of scRNA-seq, obvious immune infiltration involving T, B and macrophages was observed in the SjD group, forming several niches in the spatial view (Fig. 2D).

Figure 2.

Spatial transcriptome certified upregulation of inflammatory pathways. (A) UMAP of lacrimal gland of CT and SjD groups in the spatial transcriptome dataset. (B) Transcriptional levels of Cyp2f2 and Krt5 projected in a spatial view. (C) Heatmap of top DEGs of each subcluster. (D) Proportion of each cluster in both groups indicated immune cells infiltration. (E) Vlnplot of GO pathway scores of “Acute inflammatory response” in each subclusters of CT and SjD groups. (F) Vlnplot of GO pathway scores of “Chronic inflammatory response” in each subclusters of CT and SjD groups. (G) Vlnplot of GO pathway scores of “Positive regulation of receptor signaling pathway via JAK-STAT” in each subclusters of CT and SjD groups. (H) Vlnplot of GO pathway scores of “Response to type II interferon” in each subclusters of CT and SjD groups.

Then, following the assumption of activated JAK-STAT signaling in the lacrimal gland through scRNA-seq, we found strong, mainly acute but not chronic, inflammatory response in each parenchymal cluster (Figs. 2E, 2F). We were also aware of higher score of positive regulation of receptor signaling pathway via JAK-STAT pathway (Fig. 2G), especially in the luminal ductal and basal ductal cells, indicating the participation of JAK-STAT signaling. More importantly, also in the luminal ductal and basal ductal cells, the response to type II interferon, that is IFN-γ, was enhanced (Fig. 2H). Representative involved genes in these pathways included CCL family chemokines, CCR family chemokine receptors, TLRs, IL-1/6, and TNF. These genes mediate immune cell chemotaxis (CCL/CCR), innate immune recognition (TLR), and inflammatory cascade amplification (IL-1/6, TNF). Their upregulation indicates robust immune activation and inflammation in the local microenvironment. Together with Th1-derived IFN-γ, these factors may cooperatively promote ferroptosis in lacrimal LDE.

Such changes suggested the ductal clusters might be under the stress of secreted IFN-γ, which could be released by Th1. These combined transcriptomic data implied that LDE ferroptosis might be a target for lacrimal gland injuries in SjD.

Upregulation of JAK-STAT Signaling and Ferroptosis in the SjD Lacrimal Gland

To validate changes of enriched pathways, we selected representative markers of ferroptosis and JAK-STAT signaling and performed western blot assay (Fig. 3A). GPX4 plays an important role in protection of oxidative stress²⁴ and decreased GPX4 expression represented a classical feature of ferroptosis⁹ (Fig. 3B). FTH1 and FTL content stand for ferric storage level,²⁵ and their responsive increase has been reported in DED related injuries.^26,27 In our results, increased FTH1 and FTL might indicate reactive compensation of ferroptosis (Figs. 3C, D). ACSL4 initiated lipid peroxidation and led to organelle membrane damage,^28,29 and its upregulation suggested activation of ferroptosis (Fig. 3E). Besides, STAT1 is a downstream effector of JAK-STAT signaling,^30,31 and its content was also elevated under the background of SjD³² (Fig. 3F). To confirm the spatial specificity of these changes, we performed mIHC. Co-localization analysis with CYP2F2 demonstrated that the LDE exhibited a ferroptotic phenotype, featured by the upregulation of ACSL4, a typical initiator of lipid peroxidation as a detrimental factor to the mitochondrial membrane, and TFRC, a key transporter of ferrous iron as a well-established marker of ferroptosis (Fig. 3G). As a result, we proved upregulation of ferroptosis and JAK-STAT signaling in lacrimal gland.

Figure 3.

Upregulation of ferroptosis in the lacrimal gland. (A) Western blot of proteins related to ferroptosis and JAK-STAT signaling. (B–F) Statistical analysis of Western blot protein level (n = 4, *P < 0.05, **P < 0.01). (G) MIHC staining of CYP2F2, ACSL4 and TFRC of lacrimal gland of WT and MRLlpr groups (Scale bar: 50 µm). (H) TEM showed shrinked mitochondria in acinar, ductal and periductal zone of lacrimal gland in SjD group (2k×, scale bar: 5 µm, 6k×, scale bar: 1 µm, 700×, scale bar: 10 µm).

To further obtain direct evidence of ferroptosis, we adopted TEM to observe microstructures in the lacrimal gland. Both the acinar and LDE showed more intense staining in SjD, but mitochondria in LDE presented classical characteristics of ferroptosis, including shrinkage and vague cristae (Fig. 3H). In the periductal zone, obvious infiltration of immune cells was also observed, which suggested possible immune-parenchymal interactions and the major role of LDE during SjD related injuries.

Th1-Derived IFN-γ Drives Local Periductal T-Cell Recruitment

After we found clues of periductal immune infiltration, we first probed which kind of immune cell, myeloid cells or lymphocytes mainly participated in the damage. In the periductal zone, infiltration of CD4 labeled T cell was found in SjD group, while F4/80 labeled macrophages were not found, which primarily led the direction to T cell infiltration (Fig. 4A). Back to the scRNA-seq data, we performed sub-clustering of T cell (Fig. 4B). Both CD4+T and CD8+T clusters had noticeable enlargement, among which specifically increased CD4+Th1 suggested possible IFN-γ dominated damage (Figs. 4C, 4D). As a result, we performed flow cytometry in the lacrimal gland. Nearly no T cells were found in the control group, but a certain proportion of IFN-γ+CD4+T cells were enriched in the SjD group (Figs. 4E, 4F). Thus, we considered IFN-γ secreting Th1 might exert pathogenic effects in the lacrimal gland.

Figure 4.

IFN-γ secreting Th1 was infiltrated in the lacrimal gland. (A) Immunofluorescence of CD4, F4/80 and CYP2F2 of lacrimal gland of SjD group (Scale bar: 50 µm). (B) UMAP of immune cells in the lacrimal gland of CT and SjD groups in the scRNA-seq dataset. (C) Proportion of each immune cell subcluster in both groups. (D) Featureplot of “Cd4,” “Cd8a,” “Ifng,” and “Gzmk” of immune cells in the lacrimal gland of CT and SjD groups in the scRNA-seq dataset. (E) Flow cytometry of CD4 and IFN-γ of immune cells collected from the lacrimal gland of CT and SjD groups. (F) Statistical analysis of proportion of T cells in the live collected immune cells (n = 3, *P < 0.05).

To target the identified Th1-dominant pathology, we administered a systemic IFN-γ neutralizing antibody via intraperitoneal injection to SjD model mice and their controls (Fig. 5A). Flow cytometric analysis of cervical draining lymph nodes revealed a significant reduction in the proportion of IFN-γ-secreting Th1 cells following treatment (Figs. 5B, 5C). Concurrently, we observed reduced Th17 infiltration (Fig. 5D, 5E). Although both Th1 and Th17 subsets are implicated in SjD pathogenesis, we identified Th1 cells as the primary effectors, given their predominance in our initial profiling and the lack of significant changes observed in Treg proportions. Crucially, mIHC demonstrated that systemic IFN-γ neutralization effectively attenuated the local infiltration of CD4+ T cells in the periductal region surrounding CYP2F2 positive LDE (Fig. 5F). Collectively, these data demonstrate that the administration of an IFN-γ neutralizing antibody successfully suppresses both systemic and local pathogenic immune infiltration.

Figure 5.

IFN-γ neutralization alleviated Th1 infiltration. (A) Flowchart of anti-IFN-γ antibody intervention. (B) Flow cytometry showed decreased Th1 proportion in cervical draining lymph nodes. (C) Statistical analysis of Th1 proportion (n = 3, *P < 0.05). (D) Flow cytometry showed decreased Th17 proportion in cervical draining lymph nodes. (E) Statistical analysis of Th17 proportion (n = 3, **P < 0.01). (F) mIHC of CD4, F4/80 and CYP2F2 of lacrimal gland of SjD and SjD+anti-IFN-γ antibody groups (Scale bar: 50 µm)

Blockade of the IFN-γ Axis Reverses Downstream JAK-STAT Mediated Ferroptosis and Rescues Lacrimal Gland Function

After we validated decreased IFN-γ secreting Th1 infiltration near the LDE, we further evaluated whether such intervention prevented LDE damage, and investigated the concrete regulation of IFN-γ on downstream JAK-STAT signaling and ferroptosis. We specifically paid attention to JAK-STAT expression in the LDE, and in the CYP2F2 labeled cells, expression of p-JAK2 and p-STAT1, both as active effectors of JAK-STAT signaling, had obvious downregulation in the IFN-γ neutralizing group indicated through mIHC compared to the SjD group (Figs. 6A–D). As a result, JAK-STAT signaling was considered to be moderated after suppression of T-cell recruitment. We next evaluated whether IFN-γ and subsequent JAK-STAT signaling moderation could rescue the LDE from ferroptosis. After IFN-γ neutralization, we applied mIHC and also observed the expression level of ferroptosis preventive marker GPX4 was restored. ACSL4, as the marker of lipid peroxidation aspect of ferroptosis, was downregulated in the IFN-γ–interfering group (Figs. 6E, 6F), whereas FTH1 and TFRC levels remained relatively stable. This indicates that the downregulation of ferroptosis is a positive outcome of IFN-γ blockade. In general, our results supported our assumption that an IFN-γ-JAK-STAT-ferroptosis axis might play an important detrimental role in the lacrimal gland injuries under the background of SjD.

Figure 6.

IFN-γ neutralization prevented activation of JAK-STAT signaling and ferroptosis in the LDE. (A) mIHC staining of CYP2F2, JAK2, p-JAK2 of lacrimal gland of SjD and SjD+ anti-IFN-γ antibody groups (Scale bar: 50 µm). (B) Statistical analysis of p-JAK2 mean intensity of lacrimal gland of SjD and SjD+anti-IFN-γ antibody groups (n = 3, *P < 0.05). (C) MIHC staining of CYP2F2, STAT1, p-STAT1 of lacrimal gland of SjD and SjD + anti-IFN-γ antibody groups (Scale bar: 50 µm). (D) Statistical analysis of p-STAT1 mean intensity of lacrimal gland of SjD and SjD+ anti-IFN-γ antibody groups (n = 3, *P < 0.05). (E) MIHC staining of CYP2F2, FTH1, and GPX4 of lacrimal gland of SjD and SjD+ anti-IFN-γ antibody groups (Scale bar: 50 µm). (F) MIHC staining of CYP2F2, TFRC, and GPX4 of lacrimal gland of SjD and SjD+ anti-IFN-γ antibody groups (Scale bar: 50 µm).

Finally, after we have certified the pathogenic mechanism of IFN-γ in the lacrimal gland of SjD-DED, we would like to further assess the therapeutic efficacy of this intervention on DED related signs, as DED presented by abnormal amount and quality of tear is the eventual result. Through the phenol red thread, we detected the tear secretion of each group and found IFN-γ neutralization significantly restored the tear secretion level (Fig. 7A). Besides, as cornea is usually involved due to absolutely insufficient tear secretion and long-term exposure, we measured the corneal injuries through lissamine staining, and a marked reduction in corneal epithelial defects was certified in the IFN-γ neutralization group (Fig. 7B). At the tissue level, TUNEL assay, a common tool to show apoptotic cells, confirmed an apparent decrease in the number of dead CYP2F2-labeled LDE in the IFN-γ neutralization group, indicating that relief of ferroptosis prevented loss of functional LDE (Fig. 7C). Furthermore, TEM analysis as a key ferroptosis indicator revealed the restoration of mitochondrial integrity in the LDE after IFN-γ treatment, characterized by clearer membrane structures and discernible cristae (Fig. 7D). Collectively, these data demonstrate that targeting the Th1-IFN-γ-JAK-STAT-Ferroptosis axis effectively preserves LDE viability and ameliorates SjD-associated DED.

Figure 7.

IFN-γ neutralization relieved dry eye related signs and lacrimal gland injuries. (A) Tear secretion of CT, SjD, and SjD+ anti-IFN-γ antibody groups (n = 5, *P < 0.05). (B) Representative lissamine staining of CT, SjD, and SjD+ anti-IFN-γ antibody groups. (C) TUNEL staining with simultaneous immunofluorescence of CYP2F2 of lacrimal gland of CT, SjD, and SjD+anti-IFN-γ antibody groups (Scale bar: 50 µm). (D) TEM pictures of LDE of CT, SjD, and SjD+ anti-IFN-γ (2k×, Scale bar: 5 µm).

Discussion

SjD is a complex autoimmune disease where the precise mechanisms of lacrimal gland destruction remain largely elusive. Although lacrimal gland pathology in SjD has been linked to CD4+ and CD8+ T cell infiltration,^33,34 bridging these upstream immune signals to a specific terminal executioner pathway in a defined epithelial subset has been lacking.

A notable observation in our study was the spatial distribution of the injury. Previous research on SjD-associated lacrimal gland damage has focused predominantly on the lacrimal gland as a whole or acinar cells,³⁵ given their abundance and primary role in fluid secretion. However, the lacrimal gland comprises a sophisticated ductal network responsible for tear modification and transport, a compartment that has been significantly underappreciated.³⁶ By integrating spatial transcriptomics with high-resolution scRNA-seq, we moved beyond global observations to identify the LDE, marked by CYP2F2, as a major site of immune-parenchymal crosstalk. We found that the LDE exhibited a marked increase in cell-cell communication strength, yet paradoxically suffered from a severe reduction in proportion. Crucially, this parenchymal depletion spatially coincided with characteristic peri-ductal immune infiltration, suggesting that the heightened signaling renders these cells susceptible to immune-mediated damage. This indicates that the ductal network is not merely a passive conduit, but a highly vulnerable immunological interface crucial for the perpetuation of SjD pathogenesis.

Further interrogation of the LDE revealed a concurrent upregulation of ferroptosis and JAK-STAT signaling. Similar alterations have been noted in the salivary glands of SjD patients.^8,15 However, our study is the first to spatially resolve these pathways to the LDE in the lacrimal gland. Both IF and TEM confirmed that ferroptosis was enriched in the LDE rather than acinar cells, highlighting a novel locus of local damage. In the TEM, more intense staining represented by direct deeper darkness in the figure is usually depicted as intense electron density. As for the concrete involved components, we would like to show the mitochondria as a major target, and numerous studies have linked high electron density of mitochondria membrane to ferroptosis.^37,38 So far, the morphological changes under TEM are still the most direct and reliable evidence of ferroptosis, especially distinguished from other programmed cell death, providing reliable support for the ferroptosis. Besides, interestingly under a biomolecular view, although we observed distinct upregulation of the pro-ferroptotic driver ACSL4 and downregulation of the anti-ferroptotic guardian GPX4, the levels of TFRC remained relatively constant. Ferroptosis is a multifaceted process initiated by various distinct drivers, including the ACSL4/lipid peroxidation axis, TFRC-mediated iron overload, or HO-1 dependent mechanisms.³⁹ Our results suggest that in the SjD lacrimal gland, ferroptosis is primarily driven by the ACSL4-mediated lipid peroxidation pathway rather than the canonical TFRC-iron uptake route. Regarding FTH1, although it functions in ferric sequestration, its expression is often upregulated in ferroptotic contexts as a reactive compensatory mechanism. Therefore the lack of significant change in FTH1 levels in our data likely reflects a dynamic equilibrium between pathogenic iron accumulation and the tissue's compensatory response, rather than a static baseline.

The implication of the JAK-STAT pathway extends beyond its canonical role in mediating autoimmune inflammation. Our previous research also illuminated the upregulation of JAK-STAT signaling in the cornea under the DED background,^2,40 and it is increasingly recognized as a critical regulator of cell death modalities, including ferroptosis. Mechanistically, IFN-γ–activated STAT1 has been reported to transcriptionally repress GPX4 activity and sensitize cells to lipid peroxidation. Our data indicated specific upregulation of p-JAK2 and p-STAT1 in LDE spatially coincided with diminished GPX4 expression and ferroptotic features. Thus we propose that JAK-STAT signaling serves as a pivotal “immuno-metabolic switch,” translating the periductal Th1-derived IFN-γ signal into a terminal ferroptotic cascade specifically within the vulnerable LDE.

Regarding the autoimmune microenvironment, we confirmed characteristic peri-ductal infiltration and sought to identify the specific immune drivers of ferroptosis. Our dataset revealed that enlargement of pathogenic T-cell population was the main feature, and changes in the monocyte-macrophage population were relatively minor. We observed noticeable recruitment of CD4+T cells near the LDE rather than F4/80+ macrophage through IF, and flow cytometry further suggested infiltration of IFN-γ secreting Th1 in the lacrimal gland, which corresponded to former research.⁴¹ To delineate the upstream effects of IFN-γ secreting Th1, we applied neutralizing antibodies systemically and observed successful Th1 removal in the cervical draining lymph nodes and lacrimal gland. Such intervention effectively downregulated JAK-STAT signaling and ferroptosis, and served as a candidate therapeutic target for SjD, with regard to the alleviated DED signs in the ocular surface. Interestingly, although we found decreased proportion of both Th1 and Th17, we still considered Th1 played a major role. On the one hand, Th1 occupied a much larger proportion compared to Th17, in both the scRNA-seq dataset and flow cytometry results. On the other hand, Th1 and Th17 were considered to be cooperative in many diseases and thus inhibition of Th1 might exert additional effects on Th17. Th1 and Th17 cells are not simply mutually exclusive pro-inflammatory CD4+ T-cell subsets; rather, they can be induced in parallel within the same inflammatory milieu and may exert overlapping or even collaborative pathogenic effects. In addition, Th17 cells display considerable plasticity and can acquire a Th1-like, IFN-γ-producing phenotype under inflammatory conditions. As Th17 proportion is nearly only 10% of that of Th1, we assumed that Th1 might play a major role and Th17 exerted synergistic effects.

Conclusions

We illustrated an IFN-γ-JAK2-STAT1-Ferroptosis axis specifically in the LDE under the SjD background, which led to progression of DED. Our research also had several limitations. Though we certified changes of both JAK-STAT signaling and ferroptosis in the LDE under regulation of IFN-γ, further experiments were yet needed to better show the regulating effects of JAK-STAT signaling on ferroptosis. Besides, although the acini did not present apparent ferroptosis and JAK-STAT signaling, more work to illustrate whether cooperative pathogenic pathways existed and why LDE were susceptible to IFN-γ mediating ferroptosis excluding the T cell proximity was still anticipated. Nevertheless, IFN-γ neutralization demonstrates promising therapeutic efficacy and represents a viable candidate for drug development targeting lacrimal gland preservation in SjD.