High-Throughput Tear Proteomics Reveals Multi-Faceted Signatures of Thyroid Eye Disease

Abstract

Purpose

Thyroid eye disease (TED) is the most common extrathyroidal manifestation of Graves’ disease (GD). Despite its clinical significance, the pathogenic mechanisms and reliable diagnostic biomarkers for TED remain incompletely defined. Tear fluid offers a noninvasive window into disease-related molecular changes.

Methods

Tear samples were collected using Schirmer strips from 20 patients with TED, 20 patients with GD without ocular involvement, and 14 healthy controls (HCs). Proteomic profiling was performed using a novel pressure cycling technology-pulse data-independent acquisition mass spectrometry (PCT-PulseDIA-MS) workflow.

Results

A total of 5966 tear proteins were quantified. Differentially expressed proteins (DEPs) were identified through pairwise group comparisons. Patients with TED showed the most extensive tear proteomic alterations among the studied groups. One hundred seventy-four DEPs were associated with ophthalmopathy, 14 with autoimmunity, and 13 with hyperthyroidism. The ophthalmopathy-related DEPs were enriched in immune regulation, lipid metabolism, vascular function, and extracellular matrix remodeling. Several key DEPs showed significant correlations with clinical, laboratory, and imaging variables. A three-protein panel comprising calcium-activated nucleotidase 1 (CANT1), insulin-like growth factor-binding protein 7 (IGFBP7), and caspase 14 (CASP14) achieved excellent diagnostic performance in distinguishing TED from GD, with an area under the curve (AUC) of 0.971.

Conclusions

Tear proteomics reveals distinct molecular signatures shaped by the combined influences of ophthalmopathy, autoimmunity, and hyperthyroidism throughout the pathogenesis of TED, underscoring the potential of tear proteins as early, noninvasive biomarkers for disease diagnosis.

Thyroid eye disease (TED), also known as Graves’ orbitopathy (GO), is the most prevalent orbital disease in adults. It is a disfiguring and potentially sight-threatening organ-specific autoimmune condition, characterized by orbital inflammation and tissue remodeling.¹ TED is most commonly associated with Graves’ disease (GD), which affects approximately 2% of the global population.² Nearly half of the patients with GD develop ocular symptoms, and up to 70% exhibit radiologic evidence of orbital involvement.³ Both conditions share a common autoimmune basis involving thyrotropin receptor antibodies (TRAb) that target thyrotropin receptor (TSHR) expressed on thyroid follicular cells and orbital fibroblasts.⁴ Insulin-like growth factor-1 receptor (IGF-1R) autoantibodies further contribute to TED-specific pathology.⁵

Emerging evidence implicates the lacrimal gland in TED pathogenesis, primarily due to the cross-reactivity of TRAb with TSHR on lacrimal gland acinar cells.⁶ Imaging studies have revealed glandular enlargement,^7,8 whereas histological analyses have demonstrated elevated proinflammatory cytokines, chemokines, and immune cell infiltration within the lacrimal microenvironment.⁹ Such inflammation likely disrupts tear secretion and composition, thereby compromising ocular surface homeostasis.¹⁰ These findings raise the possibility that alterations in tear proteomic profiles may provide a window into the local immune microenvironment of TED.

Tear fluid has gained increasing recognition as a promising source of biomarkers for both ocular and systemic diseases due to its noninvasive and accessible nature.¹¹ Proteomic profiling of tear fluid offers unique advantages for ocular disease studies, owing to their direct contact with the eye and functional proximity to disease phenotypes.¹² To date, over 3370 proteins have been identified in human tear samples.¹³ Prior studies have reported dysregulated protein expression in TED, including upregulated proinflammatory proteins and downregulated protective ones.^14–16 However, conventional proteomic approaches are constrained by limited sample volumes and the wide dynamic range of protein abundance. Recent advances in mass spectrometry (MS) technologies allow for accurate quantification of low-abundance proteins from minute tear samples.¹⁷ Nevertheless, technical hurdles persist in attaining efficient and reproducible sample preparation.

In this study, we used a pressure cycling technology (PCT)-assisted protocol to optimize protein extraction and proteolysis digestion from Schirmer strips.¹⁸ We then applied data-independent acquisition mass-spectrometry with multi-injection pulsed gas-phase fractionation (PulseDIA-MS) for high-throughput proteomic profiling from TED, GD, and healthy controls (HCs).¹⁹ Our aims were to comprehensively characterize tear proteomic landscape, identify protein signatures specifically associated with orbital involvement, systemic autoimmunity, and thyroid dysfunction, and discover candidate biomarkers for TED diagnosis. A schematic overview of the study design is provided in Figure 1.

Figure 1.

Schematic Diagram of the study. A multi-step workflow combining pressure cycling technology-pulse data-independent acquisition (PCT-PulseDIA) mass spectrometry-based tear proteomics with multidimensional clinical correlations was used to identify distinct proteins linked to ophthalmopathy, autoimmunity, and hyperthyroidism through comparative analysis across TED, GD, and HCs.

Collectively, this integrative approach highlights the value of tear proteomics for understanding the molecular interplay between orbital pathology and systemic immune-metabolic dysregulation in TED.

Methods

Study Subjects

This cross-sectional observational study included 54 participants from 2 clinical centers: the Department of Ophthalmology, Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine, and the Department of Endocrinology, JiangYuan Hospital Affiliated to Jiangsu Institute of Nuclear Medicine. The cohort comprised 20 patients with TED, 20 patients with GD, and 14 HCs consecutively recruited between June and September 2023. All participants were adults aged 18 years or older.

All patients with TED had a prior GD diagnosis and met Bartley's diagnostic criteria, with all having normal thyroid function at sampling.²⁰ Disease manifestations were evaluated following the 2021 European Group on Graves’ Orbitopathy (EUGOGO) guidelines.²¹ All patients with GD had active hyperthyroidism and showed no clinical or radiological evidence of ocular involvement.²² HCs had normal thyroid function and no autoimmune history.

Exclusion criteria included: (i) systemic autoimmune diseases (e.g., Sjögren's syndrome); (ii) ocular surface diseases (e.g., dry eye, allergic conjunctivitis, and corneal abnormalities); (iii) contact lens wear; (iv) use of medication affecting tear composition, such as eye drops (tear substitutes without preservatives were accepted), antihistamines, and diuretics; and (v) pregnancy or lactation.

Demographic, clinical, and laboratory data were collected, including age, sex, smoking status, and medical history. Laboratory tests included thyroid hormones and autoantibodies. TRAb levels were measured by thyrotropin-binding inhibiting immunoglobulin (TBII) assays using the Chemiluminescence Immunoassay (CLIA) method. Orbital computed tomography (CT) scans were performed on patients with TED and patients with GD. Lacrimal gland morphology was measured on the slice where the gland appeared the largest with reference to previously published methods.^23,24

All participants provided their written informed consents. The study was approved by the local institutional review board (SH9H-2019-T192-2) and conducted in accordance with the Declaration of Helsinki.

Tear Sample Collection

Tear fluids were collected using Schirmer strips (Surgi Edge, India). A 5 × 35 mm filter paper was folded 5 mm from one end and placed in the lower fornix at the junction of the middle and outer thirds of the lower eyelid for 5 minutes. Care was taken to avoid contact with the cornea or lashes. The moistened length was recorded. The strips were stored at −80°C until proteomic analyses.

PCT-PulseDIA-MS Based Proteomic Workflow

Tear proteins were extracted from fragmented Schirmer strips using PCT to enhance lysis and enzymatic digestion. Proteins were reduced, alkylated, and digested into peptides using Trypsin and Lys-C under 20,000 psi cycles (50 seconds high, 10 seconds ambient pressure per cycle, and 30°C for 120 cycles), followed by acidification and desalting using the SOLAµTM solid-phase extraction (SPE) plate (Thermo Fisher, USA). For spectral library construction, peptides from all samples were pooled (100 µg total) and fractionated into 30 fractions via high-pH reversed-phase chromatography. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis was performed in data-dependent acquisition (DDA) mode using the nanoflow Vanquish Neo ultra-high-performance liquid chromatography (UHPLC) system coupled to an Orbitrap Exploris 480 MS (Thermo Scientific, San Jose, CA, USA) equipped with a field asymmetric ion mobility spectrometry (FAIMS) Pro (Thermo Fisher, USA). For the proteomic analysis, the same LC-MS/MS platform was used in PulseDIA mode, using two compensation voltages (−65 V and −45 V) and segmented acquisition to cover a precursor ion mass range of 390 to 1010 m/z with overlapping isolation windows. For DIA acquisition, peptide concentrations were adjusted to 0.2 µg/µL, with 2 µL injected per sample (400 ng total), ensuring consistent data acquisition.

MS data were processed using data-independent acquisition by neural networks (DIA-NNs) software (version 1.8.1) with a DDA-derived spectral library, comprising 14,084 protein isoforms, 14,084 protein groups, and 127,916 precursors in 115,447 elution groups. Static and variable modifications were applied, and results were filtered at a stringent false discovery rate (FDR) < 0.01.

Quality control (QC) analyses were conducted to ensure reproducibility, data consistency, and instrument stability, with results provided in Supplementary Figure S1.

Identification and Classification of Differentially Expressed Proteins

Differentially expressed proteins (DEPs) among the TED, GD, and HC groups were identified using Welch's t test for pairwise comparisons (TED versus GD, TED versus HCs, and GD versus HCs) on log₂-transformed protein intensities. A threshold of |fold change (FC)| > 1.5 and raw P value < 0.05 was applied for DEP screening and visualization. Volcano plots were generated to illustrate the overall distribution of expression differences. To ensure statistical rigor, Benjamini–Hochberg FDR correction was also performed. This combined strategy provides both intuitive visualization of global expression trends and rigorous control of multiple testing.

DEPs were interpreted based on pairwise comparisons and clinical context. For instance, patients with TED had normal thyroid function with prior GD, whereas patients with GD were hyperthyroid without ocular involvement. TED versus GD reflects differences related to ophthalmopathy and thyroid status, whereas TED versus HCs and GD versus HCs comparisons reflect immune and endocrine components. Based on overlaps, DEPs were categorized into TED-specific ophthalmopathy, GD-related autoimmunity, and thyroid dysfunction. DEPs common to all three comparisons were excluded to ensure specificity.

Bioinformatic and Statistical Analysis

Functional enrichment analyses were performed, including Gene Ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations. Protein–protein interaction (PPI) networks were constructed using the STRING database to explore interconnectivity and functional clustering.

To prioritize biomarkers relevant to TED-specific ophthalmopathy, a composite scoring was calculated for DEPs overlapping between TED versus GD and TED versus HCs using the following formula:

$$\begin{array}{c}\hfill \mathrm{Score}=\frac{\left|{log}_2\mathrm{FC}\left(\mathrm{TE}{\mathrm{D}}_{vs.}\mathrm{GD}\right)\right|+\left|{log}_2\mathrm{FD}\left(\mathrm{TE}{\mathrm{D}}_{vs.}\mathrm{HC}\right)\right|}{\left(\frac{\mathrm{adj}.P_{TEDvs.GD}+\mathrm{adj}.P_{TEDvs.HC}}{2}+\varepsilon \right)}\end{array}$$

ε is a small constant to avoid division by zero. A higher score reflects both a greater magnitude of expression difference and stronger statistical significance across both comparisons, as determined by the average log₂ FC and the average Benjamini–Hochberg adjusted P value. The top 20 proteins were selected for clinical correlation analysis. Spearman's correlation assessed associations between protein expression and clinical, ophthalmic, laboratory, and imaging variables. Receiver operating characteristic (ROC) curve analyses evaluated diagnostic performance of candidate biomarkers in differentiating TED from GD.

All statistical analyses were conducted using GraphPad Prism version 9.0 software. Data were presented as mean ± standard deviation (SD) or median with interquartile range (IQR). Comparisons were made using 2-tailed t-test, 1-way analysis of variance (ANOVA), Mann-Whitney U test, Kruskal–Wallis nonparametric test, and Chi-square test, as appropriate. A P value < 0.05 was considered statistically significant.

Results

Demographic and Clinical Data

To investigate tear proteomic alterations in TED and GD, a PCT-PulseDIA-MS based proteomics study was conducted involving 3 distinct groups: 20 patients with TED with normal thyroid function but prior GD, 20 patients with hyperthyroid GD without ocular involvement, and 14 HCs. This design enabled the delineation of proteomic alterations attributable to ophthalmopathy, thyroid dysfunction, and the underlying autoimmune background.

Baseline demographic and clinical characteristics of the 54 participants are summarized in Table 1. The three groups were comparable in terms of sex distribution, age, smoking status, and GD duration (all P > 0.2). The median TED duration was 9.5 months. As expected, patients with GD exhibited significantly higher levels of fT3, fT4, and lower TSH levels than patients with TED (P < 0.0001). TRAb levels tended to be higher in GD than TED, although marginally nonsignificant (P = 0.0501), whereas thyroid peroxidase antibody (TPOAb) and thyroglobulin antibody (TGAb) levels were significantly elevated in GD (P = 0.0210 and P = 0.0030, respectively). Ophthalmic parameters were collected in patients with TED, and lacrimal gland morphology was measured by CT imaging in both TED and GD groups.

Table 1.

Baseline Demographic and Clinical Characteristics of all Subjects

Characteristic	TED (n = 20)	GD (n = 20)	HC (n = 14)	P Value
Sex, n (%)				0.3219
Male	5 (25.0)	4 (20.0)	6 (42.9)
Female	15 (75.0)	16 (80.0)	8 (57.1)
Age, y, means ± SD	36.6 ± 12.6	35.8 ± 14.0	32.7 ±13.4	0.3941
Smoking status, n (%)				0.5170
Smoker	1 (5.0)	2 (10.0)	1 (7.1)
Non-smoker	16 (80.0)	17 (85.0)	13 (92.9)
Former smoker	3 (15.0)	1 (5.0)	0 (0.0)
Duration of GD, mo, median (IQR)	21.5 (6 to 45)	8 (2 to 69)	–	0.2178
Duration of TED, mo, median (IQR)	9.5 (5 to 33)	–	–	–
Thyroid hormone level
fT3 - pmol/L, median (IQR), reference range, 3.53–7.38	4.57 (4.19 to 5.00)	13.40 (7.89 to 24.77)	–	<0.0001‡
fT4 - pmol/L, median (IQR), reference range, 7.98–15.96	12.63 (10.97 to 14.13)	34.62 (18.53 to 54.12)	–	<0.0001‡
TSH - uIU/mL, median (IQR), reference range, 0.56–5.91	1.031 (0.470 to 2.525)	0.005 (0.005 to 0.006)	–	<0.0001‡
TRAb, IU/L - median (IQR), reference range, 0.00–1.75	5.69 (1.88 to 9.55)	8.61 (4.79 to 20.23)	–	0.0501
TPOAb, IU/L - median (IQR), reference range, 0.00–9.00	57.10 (20.30 to 289.00)	205.20 (63.07 to 367.40)	–	0.0210*
TGAb, IU/L - median (IQR), reference range, 0.00–115.00	16.81 (13.85 to 187.60)	327.00 (76.21 to 848.90)	–	0.0030†
Schirmer's test - mm, means ± SD	20.1 ± 7.5	26.5 ±7.1	26.3 ±8.2	0.0172*
Clinical activity score - points, means ± SD, range, 0–7	2.60 ± 1.43	–	–	–
NOSPECS score - points, means ± SD, range, 0–6	3.25 ± 1.33	–	–	–
Proptosis - mm, means ± SD	18.74 ± 2.10	–	–	–
Palpebral fissure height - mm, means ± SD	9.75 ± 2.47	–	–	–
Bahn-Gorman Diplopia Score - points, 0/1/2/3	16/0/3/1	–	–	–
Eye motility - points, 0/1/2/3	10/9/1/0	–	–	–
Coronal length of LG - mm, means ± SD	17.59 ± 3.28	–	–	–
Coronal width of LG - mm, means ± SD	4.47 ± 1.10	–	–	–
Coronal area of LG - cm2, means ± SD	0.58 ± 0.16	–	–	–

fT3, free triiodothyronine; fT4, free thyroxine; GD, Graves’ disease; HC, healthy control; IQR, interquartile range; LG, lacrimal gland; mo, months; SD, standard deviation; TED, thyroid eye disease; TGAb, thyroglobulin antibody; TPOAb, thyroid peroxidase antibody; TRAb, thyrotropin receptor antibody; TSH, thyrotropin; y, years.

Data are presented as means ± standard deviation (SD), median with interquartile range (IQR), or number (percentages), as appropriate. For comparisons among the three groups, chi-square test was used for categorical variables, one-way ANOVA or Kruskal–Wallis test was used for continuous variables, depending on data distribution. Pairwise comparisons of continuous variables were performed using the Mann-Whitney U test.

^* P value < 0.05.

^† P value < 0.01.

^‡ P value < 0.001.

Tear Proteomic Profiling

Proteomic profiling was performed using the PCT-PulseDIA-MS workflow, with five randomly chosen samples running as technical replicates to ensure data robustness. A total of 5966 proteins were quantified (FDR < 1%), of which, 94.8% were shared across all groups, confirming high consistency, whereas TED had the most unique proteins (n = 69), indicating greater tear proteomic alterations.

Uniform Manifold Approximation and Projection (UMAP) visualization (Fig. 2a) suggested a tendency for TED samples to cluster, whereas GD and HCs showed greater overlap. Pairwise comparisons identified 315 DEPs between TED and GD (212 upregulated and 103 downregulated), 523 DEPs between TED and HCs (424 upregulated and 99 downregulated), and only 48 DEPs between GD and HC (29 upregulated and 19 downregulated), supporting more pronounced tear proteomic changes in TED.

Figure 2.

Global proteomic profiling of tear samples in the TED, GD, and HC groups. (a) UMAP (Uniform Manifold Approximation and Projection) plots based on 5966 protein features reveal distinct clustering patterns among TED (red), GD (blue), and HC (green) groups. (b) UpSet diagram illustrating the overlap of DEPs across pairwise group comparisons. The intersection of 174 DEPs between TED versus HCs and TED versus GD defines a proteomic signature potentially associated with TED-specific eye involvement.

The UpSet diagram (Fig. 2b) revealed both specific and shared patterns of DEPs across group comparisons. Notably, 174 DEPs shared between TED versus GD and TED versus HCs likely define a TED-specific tear proteomic signature associated with local ophthalmopathy. Fourteen DEPs related to autoimmunity and 13 DEPs indicative of thyroid dysfunction were identified based on their logical overlap. Besides, two DEPs common to all three comparisons may reflect the convergence of triple pathophysiological influences.

Functional Characterization of TED-Specific Tear Proteins

The 174 DEPs associated with TED-specific ophthalmopathy effectively distinguished TED from both GD and HCs (Fig. 3a). Their absence in the GD versus HC comparison suggests that these expression changes are primarily driven by localized orbital pathology rather than systemic thyroid dysfunction or general autoimmune effects. Heatmap analysis further confirmed the distinct expression patterns of these DEPs (Fig. 3b).

Figure 3.

Pathway enrichment analysis of 174 tear proteins associated with TED-specific eye involvement. (a) Volcano plots depicting differentially expressed proteins (DEPs) identified in TED versus GD and TED versus HC comparisons, using a threshold of foldchange > 1.5 and P value < 0.05. Red and blue dots, respectively, represent upregulated and downregulated proteins, whereas the grey dots denote those insignificant ones. The Venn diagram identifies 174 overlapping DEPs potentially linked to TED-specific eye involvement. (b) Heatmap of the 174 overlapping DEPs demonstrating distinct expression profiles across TED, GD, and HC groups. The color scale (red to blue) indicates relative abundance. (c) Top 10 KEGG pathways sorted by the number of overlapping DEP annotated in the corresponding pathway, highlighting complement and coagulation cascades, as well as cholesterol metabolism, as the most significantly enriched pathways. The top 10 enriched KEGG pathways were classified into Organismal Systems (n = 4), Human Diseases (n = 5), and Cellular Processes. (d) Network visualization of the top five enriched KEGG pathways, showing functional clustering and core protein components. Different colored lines indicate different pathways. (e) Gene ontology enrichment analysis of the 174 shared DEPs depicting biological processes (BP), cellular components (CCs), and molecular function (MF). Bar plots are sorted by Gene Ontology category from up to down. (f) Protein–protein interaction (PPI) network highlighting key hub proteins, with node size reflecting connectivity degree.

KEGG pathway analysis revealed significant enrichment in the complement and coagulation cascades (21 proteins) and cholesterol metabolism (7 proteins; Fig. 3c). Network visualization demonstrated tight functional clustering among these pathway-enriched proteins (Fig. 3d).

Gene Ontology analysis (Fig. 3e) revealed involvement in three aspects: (i) biological processes (BP), including humoral immune response, complement activation, and fibrinolysis; (ii) cellular components (CCs), primarily located in blood microparticle and lipoprotein particle; and (iii) molecular functions (MFs), related to (endo)peptidase inhibitor/regulator activity and glycosaminoglycan binding. These findings align well with the inflammatory and lipid dysregulation known in TED pathophysiology. The PPI network (Fig. 3f) identified hub proteins, including albumin (ALB), beta-2-glycoprotein 1 (apolipoprotein H [APOH]), alpha-2-HS-glycoprotein (AHSG), serpin family C member 1 (SERPINC1 and antithrombin III), vitronectin (VTN), serpin family D member 1 (SERPIND1 and heparin cofactor II), alpha-2-macroglobulin (A2M), plasminogen (PLG), and apolipoprotein C-III (APOC3), suggesting their central roles in TED pathophysiology.

Together, these proteins represent a complex interplay of immune regulation, lipid metabolism, vascular function, and extracellular matrix remodeling, offering a molecular signature of TED-related ophthalmopathy.

Clinical Relevance and Diagnostic Potential of Ophthalmopathy-Associated Proteins

A composite score integrating FC and statistical significance was calculated for the 174 ophthalmopathy-related DEPs. The top 20 were selected for further clinical evaluation (Table 2).

Table 2.

Summary of the Top 20 Ophthalmopathy-Associated Proteins With Differential Expression Among the TED, GD, and HC Groups

	TED vs. GD				TED vs. HC
Protein Name	Log2FC	P Value	Adj. P Value	Class	Log2FC	P Value	Adj. P Value	Class	Score
Q9NRA1-2_PDGFC	−0.9738	0.0000	0.0098	Down	−1.1546	0.0000	0.0283	Down	55.8168
P02747_C1QC	1.5042	0.0002	0.0444	Up	1.6209	0.0003	0.0721	Up	26.8097
Q8WVQ1-3_CANT1	−1.0197	0.0000	0.0080	Down	−0.7895	0.0002	0.0645	Down	24.9636
P02766_TTR	1.2724	0.0000	0.0098	Up	1.0945	0.0006	0.0900	Up	23.7029
P05452_CLEC3B	0.8153	0.0001	0.0440	Up	0.8922	0.0000	0.0283	Up	23.6054
P98160_HSPG2	−1.0184	0.0001	0.0440	Down	−1.4337	0.0004	0.0771	Down	20.2424
O75556_SCGB2A1	0.6950	0.0024	0.0984	Up	1.2025	0.0000	0.0039	Up	18.5546
Q8WVQ1-2_CANT1	−0.8648	0.0000	0.0094	Down	−0.6684	0.0004	0.0771	Down	17.7114
Q9Y343-2_SNX24	1.0726	0.0001	0.0394	Up	1.0583	0.0005	0.0826	Up	17.4715
P01011-2_SERPINA3	1.0920	0.0005	0.0663	Up	1.2321	0.0002	0.0721	Up	16.7915
Q9P2R7-2_SUCLA2	0.7568	0.0011	0.0766	Up	0.9894	0.0001	0.0398	Up	15.0097
O43505_B4GAT1	−0.6901	0.0004	0.0551	Down	−0.7380	0.0001	0.0435	Down	14.4892
P31944_CASP14	−1.5261	0.0010	0.0766	Down	−1.8919	0.0103	0.1654	Down	14.1260
P02655_APOC2	1.5282	0.0006	0.0690	Up	1.4111	0.0033	0.1416	Up	13.9612
Q16270-2_IGFBP7	−1.1567	0.0000	0.0094	Down	−0.8862	0.0024	0.1412	Down	13.5643
Q8N428-2_GALNT16	−0.7587	0.0000	0.0094	Down	−0.6335	0.0006	0.0973	Down	13.0442
P02790_HPX	1.1489	0.0006	0.0693	Up	1.2500	0.0017	0.1266	Up	12.2403
Q9UKM9−2_RALY	0.6953	0.0014	0.0842	Up	0.8591	0.0001	0.0435	Up	12.1763
P04004_VTN	0.9545	0.0002	0.0466	Up	0.9311	0.0010	0.1139	Up	11.7521
P01011_SERPINA3	0.8886	0.0024	0.0991	Up	1.1354	0.0004	0.0771	Up	11.4851

Adj. P Value, adjusted P value; GD, Graves’ disease; HC, healthy control; Log₂FC, Log₂(Fold Change); TED, thyroid eye disease.

Protein names are presented as UniProt gene symbols; their full names are provided at first mention in the main text.

Log₂fold change (Log₂FC), exact P values, and Benjamini–Hochberg adjusted P values are reported. Class indicates the direction of differential expression in TED. Score represents an ophthalmopathy-associated ranking score calculated by averaging the absolute Log₂FC values from both comparisons and dividing by the average adjusted P value, with ε, a small constant, added to avoid division by zero. A higher score indicates greater expression differences and stronger significance across both comparisons.

Spearman correlation identified associations between selected proteins and TED-specific features (Fig. 4a). Notably, 6 DEPs, sorting nexin 24 (SNX24), vitronectin (VTN), apolipoprotein C-II (APOC2), complement C1q subcomponent subunit C (C1QC), transthyretin (TTR), and hemopexin (HPX) were significantly correlated with palpebral fissure height (Figs. 4b–d), potentially reflecting tear film instability driven by increased surface exposure, altered lipid homeostasis, and epithelial stress responses. C-type lectin domain family 3 member B (CLEC3B) negatively correlated with NOSPECS scores and was also linked to eye motility, indicating its involvement in TED-related inflammation, adipogenesis, and fibrosis (Fig. 4e). Additionally, CANT1 correlated with TED duration (r = 0.5419, P value = 0.0136; Figs. 4f, 4g), and IGFBP7 with TRAb levels (r = 0.3821, P value = 0.0164; Figs. 4h, 4i). Importantly, CASP14, a downregulated protein, inversely correlated with multiple lacrimal gland CT indices, including coronal width, length, and area, suggesting its role in tissue remodeling in TED (Figs. 4j–l).

Figure 4.

Association between ophthalmopathy-specific tear proteins and TED-related clinical features. (a) Heatmap of Spearman correlation coefficients between the top 20 TED-specific eye involvement-associated tear proteins and 11 clinical, laboratory, and imaging parameters. Color scale from blue to white indicates correlation direction and strength. Asterisks denote statistical significance (*P value < 0.05, **P value < 0.01, and ***P value < 0.001). (b–d) Three representative DEPs exhibit significant negative correlations with palpebral fissure height (Q9Y343-2_SNX24: r = −0.7342, P value = 0.0002***, R² = 0.5076; P02655_APOC2: r = −0.6834, P value = 0.0009***, R² = 0.5800; P04004_VTN: r = −0.6548, P value = 0.0017**, R² = 0.4067). (e) Correlation analysis reveals inverse correlations between CLEC3B levels and NOSPECS score (r = −0.4754, P value = 0.0342*, R² = 0.1455). (f) Scatter plot showing significant correlations between Q8WVQ1-3_CANT1 expression and TED duration (r = 0.5419, P value = 0.0136*, R² = 0.2312). (g) Violin plot showing significantly decreased CANT1 expression in TED compared to GD and HCs (1-way ANOVA, P value = 3.14e-07***). (h) Scatter plot showing significant correlations between Q16270-2_IGFBP7 and TRAb (r = 0.3821, P value = 0.0164*, R² = 0.0682). (i) Violin plot showing significantly decreased IGFBP7 expression in TED compared to GD and HC (1-way ANOVA, P value = 9.31e-06***). (j) Scatter plots showing significant correlations between P31944_CASP14 expression and multiple orbital imaging markers, including coronal width (r = −0.5616, P value = 0.0190*, R² = 0.2526), length (r = −0.4853, P value = 0.0483*, R² = 0.1521), and area (r = −0.5080, P value = 0.0374*, R² = 0.3514). Spearman's correlation coefficient, P value and R² are indicated on each plot. Shaded areas represent 95% confidence intervals. (k) Violin plot showing significantly decreased CASP14 expression in TED compared to GD and HCs (1-way ANOVA, P value = 0.0023**). (l) Coronal CT image showing lacrimal gland measurements. The area was determined by manually delineating the boundary of the lacrimal gland on the slice where it appeared largest. The length was measured from the anterior to posterior tips, and the width was measured from the widest points perpendicular to the length. (m) Receiver operating characteristic (ROC) curves evaluating the diagnostic value of various biomarkers in distinguishing TED from GD. TRAb alone showed limited discriminative power (AUC = 0.684). The single tear protein CANT1 showed excellent performance (AUC = 0.918). A three-protein panel composed of CANT1, IGFBP7, and CASP14 further improved the performance, reaching an AUC of 0.971.

ROC analysis demonstrated great diagnostic performance (Fig. 4m). The best-performing protein, CANT1, achieved an AUC of 0.918, outperforming TRAb (AUC = 0.684). A 3-protein panel comprising CANT1, IGFBP7, and CASP14 further improved diagnostic accuracy, yielding an AUC of 0.971, with 95% sensitivity, 90% specificity, and a Youden Index of 0.85. These findings suggest that tear proteins have excellent discriminative capacity in distinguishing TED from GD and may serve as a valid noninvasive diagnostic tool.

Clinical Correlation of Tear Proteins Associated With Autoimmunity and Hyperthyroidism

Beyond ophthalmopathy-associated signatures, protein subsets reflecting systemic mechanisms were also examined. Fourteen DEPs shared by TED versus HCs and GD versus HCs were associated with autoimmunity (Fig. 5a). Among them, slit guidance ligand 3 (SLIT3), furin (FURIN), and proline/arginine-rich end leucine-rich repeat protein (PRELP) were positively associated with TPOAb levels, whereas Sec61 translocon alpha 1 subunit (SEC61A1), SLIT3, and caspase 7 (CASP7) are correlated with TgAb levels (Fig. 5b). In parallel, 13 DEPs shared by TED versus GD and GD versus HCs were linked to hyperthyroidism-related effects (Fig. 5c). STAM-binding protein (STAMBP) was positively correlated with GD duration, whereas heterogeneous nuclear ribonucleoprotein C (HNRNPC) inversely correlated with TRAb. Notably, immunoglobulin heavy constant mu (IGHM) and Lactadherin (MFGE8) were significantly correlated with fT3, fT4, and TSH levels, implicating their involvement in thyroid hormone regulation (Fig. 5d).

Figure 5.

Correlation analysis of tear proteins associated with autoimmunity and hyperthyroidism with clinical parameters. (a) Venn diagram showing 14 overlapping DEPs shared by TED versus HCs and GD versus HC comparisons, suggesting potential association with autoimmune processes. Bar plots illustrate relative protein expression levels across TED (red), GD (blue), and HC (green) groups, with Log₂(FC) values listed below. (b, c) Spearman's correlation analysis between selected autoimmune-related proteins and thyroid autoantibodies. O75094-2_SLIT3, P09958_FURIN, and P51888_PRELP are positively associated with TPOAb levels, whereas P61619_SEC61A1, O75094-2_SLIT3, and P55210-2_CASP7 are correlated with TgAb levels. (d) Venn diagram showing 13 shared DEPs in TED versus GD and GD versus HCs, potentially associated with thyroid hormone imbalance. Bar plots display the relative expression levels and corresponding Log₂(FC) values of these DEPs across groups. (e–h) Scatter plots demonstrate significant correlations of hyperthyroidism-related proteins with clinical indicators. O9630-2_STAMBP positively correlates with GD duration (r = 0.3492, P value = 0.0272*, R² = 0.0776). P07910-2_HNRNPC negatively correlates with TRAb levels (r = −0.4040, P value = 0.0108*, R² = 0.1552). Both P01871-1_IGHM and Q08431-2_MFGE8 show significant correlations with FT3, FT4, and TSH levels, implicating their involvement in thyroid hormone regulation. Spearman's correlation was used for all analyses. P values were two-sided (^*P value < 0.05, ^**P value < 0.01, ^***P value < 0.001).

These findings suggest that tear proteomic profiles not only reflect local orbital pathophysiology in TED but also capture molecular alterations related to autoimmune background and thyroid function, providing multidimensional molecular insights into disease mechanisms.

Discussion

TED is an autoimmune condition frequently associated with GD. Although both conditions share a common autoimmune background, TED exhibits distinct local features.²⁵ Increasing evidence highlights lacrimal gland involvement in TED.^8,9 The inflammatory microenvironment and impaired secretory function may alter tear composition, reflecting both local and systemic pathology.^26,27

Tear fluid, easily collected via Schirmer Strips, offers a noninvasive window for discovering translational biomarkers.²⁸ However, previous studies were constrained by methodologies,¹⁵ with biochemical assays like ELISA and spectrophotometry detecting only abundant tear proteins,^29,30 and MS identifying fewer than 1500 proteins.^16,31–33 In this study, we utilized PulseDIA, a data-independent acquisition strategy with a pulsed multi-injection gas-phase fractionation, to improve proteome depth and reduce missing value rates.¹⁹ Combined with PCT-based sample preparation for better protein extraction and digestion, our workflow achieved quantification of 5966 tear proteins from 54 individuals.¹⁸ This enabled robust confirmation of prior findings and the exploration of new biomarkers.^31,34 Building on this, our study provides a comprehensive tear proteome and investigates the mechanisms underlying proteomic dysregulation in TED.

TED tear samples exhibited greater proteomic dysregulation than GD or HCs, driven by three major mechanisms: ophthalmopathy, autoimmunity, and thyroid dysfunction. The 174 DEPs associated with TED-specific ophthalmopathy highlighted orbital involvement as the primary driver of tear proteomic changes. Functional enrichment revealed four primary processes: immune regulation, lipid metabolism, vascular function, and extracellular matrix (ECM) remodeling.

Immune regulation was the most enriched category, with the “complement and coagulation cascades” pathway as the top pathway, involving proteins like complement component 5 (C5), complement C1q subcomponent (C1QC), and complement component 9 (C9). Complement activation is seen in 40% of patients with TED,³⁵ plays a crucial role in antibody-mediated inflammation, and contributes to TED pathophysiology, as supported by previous proteomic studies.^34,36,37 C5a promotes T helper 17 (Th17) cell differentiation,³⁸ leading to proinflammatory cytokine secretion and orbital fibroblast activation.³⁹ C5a-induced Th1/Th2 imbalance is also implicated in TED pathogenesis.^40,41 Elevated levels of cytokines associated with Th1 cells (e.g., IL-12 and TNF-α), Th2 cells (e.g., IL-13), and Th17 cells (e.g., IL-1β, IL-6, and IL-17A) in patients with TED further support this.^42,43

Lipid metabolism was also enriched, with key proteins including apolipoprotein A-I (APOA1) and apolipoprotein C-III (APOC3) in the “cholesterol metabolism” pathway. Elevated APOA1 in TED tears supports previous findings.³⁴ Lipid alterations contribute to tear film instability through meibomian gland dysfunction.⁴⁴ Combined with aqueous tear deficiency, increased ocular surface exposure, and accelerated tear evaporation, these collectively lead to elevated tear osmolarity,⁴⁵ which can induce ocular surface epithelial apoptosis and exacerbate inflammation,¹⁰ affecting 65% to 95% of patients with TED.⁴⁶

Vascular-related proteins like coagulation factor II (F2), PLG, and APOH may contribute to periorbital edema by promoting vascular permeability during inflammation.⁴⁷ Fibrosis-related proteins, including VTN and CLEC3B, indicate excessive ECM deposition within the orbit. Together, these findings offer mechanistic insights into localized TED pathology.

The three-protein biomarker panel consisting of CANT1, IGFBP7, and CASP14 demonstrated superior diagnostic performance in differentiating TED from GD, achieving an AUC of 0.971, sensitivity of 95%, and specificity of 90%, outperforming the previous lysozyme C (LYZ)–lacritin (LACRT)–zinc-alpha-2-glycoprotein (AZGP1) panel (AUC = 0.93),⁴⁸ and the C-C motif chemokine ligand 2 (CCL2)–CD40 ligand (CD40L) panel (AUC = 0.80).⁴⁹ CANT1 regulates proteoglycan and glycosaminoglycan (GAG) metabolism.^50–52 In TED, its dysregulations may impair GAG turnover, promoting pathological ECM accumulation and orbital tissue expansion.⁵³ IGFBP7, unlike IGFBP1-6, has broader IGF-independent effects.⁵⁴ Whereas IGFBP1-6 regulate IGF-1 bioavailability via the PI3K-AKT-mTOR pathway,⁵⁵ IGFBP7 shows context-specific effects in inflammation and fibrosis.^54,56 Its downregulation in TED may indicate lacrimal gland dysfunction and loss of antifibrotic regulatory feedback. Interestingly, tear IGFBP7 levels were positively correlated with serum TRAb, linking systemic immunity to local responses. CASP14 is an epithelial-specific protease involved in terminal differentiation and barrier formation.^57,58 In TED, downregulated CASP14 correlated with reduced lacrimal gland size on CT, suggesting epithelial dysfunction and compromised barriers under chronic inflammation.⁵⁹

Additionally, we explored DEPs linked to systemic mechanisms. Among 14 autoimmunity-related DEPs, SLIT3 exhibited a stepwise decrease from HCs to GD and TED. As an anti-fibrotic and anti-myogenic factor, its loss may promote orbital fibrosis in TED.^60,61 Among 13 hyperthyroidism-related proteins, IGHM may reflect enhanced B-cell activation in GD and TED, contributing to aberrant TSHR stimulation and increased thyroid hormone production.⁶²

Despite the novel findings, this study has several limitations. First, the small sample size and focus on patients with TED with normal thyroid function limit generalizability across the full clinical spectrum. In particular, patients with euthyroid TED with negative TRAb, as well as those with Hashimoto's thyroiditis, were not included, complicating the separation of local ophthalmopathy from systemic autoimmunity or thyroid dysfunction. Likewise, the absence of other orbital diseases, like IgG4-related orbital disease, limits our ability to determine whether the identified proteins are TED-specific or reflect orbital inflammation more broadly. Second, the lack of external validation and confirmation with immunoassays limit clinical applicability and reproducibility. Future studies using targeted MS, ELISA, or Western blotting in larger cohorts are warranted to verify the diagnostic value and the specificity of the identified proteins. Third, longitudinal analysis is needed to capture dynamic changes with disease progression or treatment response. Finally, a multi-omics approach from multiple biofluids and tissues may provide a more comprehensive molecular landscape of TED.

In conclusion, this study provides a holistic tear protein landscape across TED, GD, and HCs, with 5966 proteins quantified using an optimized PCT-PulseDIA-MS workflow. By leveraging the distinct clinical characteristics of each group, this study dissects tear proteomic alterations into components of ophthalmopathy, autoimmunity, and hyperthyroidism, with ophthalmopathy emerging as the dominant driver. These associated DEPs were mainly involved in immune regulation, lipid metabolism, vascular function, and ECM remodeling, offering mechanistic insights into local pathology. Furthermore, a biomarker panel consisting of CANT1, IGFBP7, and CASP14 demonstrated superior diagnostic potential, supporting tear proteomics as a valuable tool for noninvasive TED diagnosis and mechanistic exploration.