Proteomic Analysis of Meibomian Gland Secretions in Patients With Blepharokeratoconjunctivitis

Abstract

Purpose: To screen and compare the differential proteins in meibomian gland secretions between patients with blepharokeratoconjunctivitis (BKC) and healthy individuals and to identify target proteins that may participate in the occurrence and development of BKC.

Methods: Thirteen patients diagnosed with BKC in Shenzhen Eye Hospital and five healthy volunteers were included in this study. Meibomian gland secretions and clinical traits were collected before and after 1 month of standard BKC treatment. Label-free mass spectrometry was used for proteomic detection of meibomian gland secretions. Weighted protein coexpression network analysis (WPCNA) and several different protein analyses were performed to identify hub proteins associated with BKC and its clinical characteristics.

Results: Patients with BKC had significantly lower cleanliness of the eyelid margin, higher palpebral margin scores, more serious clinical manifestations of secretions, and more damaged meibomian gland morphology compared with the healthy controls. One hundred fifteen differential proteins were associated with the clinical traits, which included diagnosis, sex, age, severity, corneal neovascularization, disease course, eyelid margin cleanliness, palpebral margin score, secretion characteristics, and meibomian gland morphology. Four hub proteins related to inflammation and the immune response (namely, S100A8, S100A9, ANXA3, and LCN2) were increased in BKC and remained increased after 1 month of treatment. The cleanliness, blepharon eyelid score, and secretion characteristics were improved after BKC treatment.

Conclusions: S100A8, S100A9, ANXA3, and LCN2 are BKC-associated proteins probably involved in the chronic inflammation of BKC.

Translational Relevance: Hub proteins probably involved in chronic inflammation of BKC were identified by proteomic methods.

Introduction

Blepharokeratoconjunctivitis (BKC) is an ocular surface inflammatory disease of the palpebral margin with secondary conjunctival and corneal involvement, with Asian women and children being particularly susceptible.¹ The pathogenesis of BKC may be correlated with microbial infection, immune response, abnormal meibomian gland secretions, tear film instability, and genetics, among other factors. BKC is often accompanied by abnormal morphology and dysfunction of the meibomian gland, manifesting as tortuosity, shortening and disappearance of the meibomian gland, and granular and toothpaste-like meibomian gland secretions. Previously, it was thought that the meibomian gland only secreted lipids to maintain homeostasis of the tear film. However, recent studies indicate that various proteins are present in the meibomian gland secretions of healthy people, and they may play an important role in maintaining ocular surface homeostasis, including lipid metabolism, anti-inflammation, and anti-infection.^2–4 However, most of the protein components and their functions have not been thoroughly investigated.⁵ 

Proteomic analysis is an effective method for investigating differential proteins expressed in diseases, screening disease markers, and understanding the function and pathogenesis of disease-associated proteins. Therefore, proteomic analysis of pathological tissues is widely used in several ophthalmic diseases, such as xerophthalmia, glaucoma, allergic conjunctivitis, and fundus disease.^6,7 Label-free mass spectrometry is one of the most commonly used protein quantitative analyses, as it offers the advantages of great economic benefit, high-throughput screening, and time and labor savings compared to classical proteomic quantification.⁸ 

Weighted protein coexpression network analysis (WPCNA) is one of the most widely used techniques to summarize modules containing a group of clinical trait-related, correlated proteins by hierarchical clustering and tree-cutting methods.⁹ The proteins in the modules most correlated to clinical traits are recognized to be associated with disease severity. Gene set enrichment analysis (GSEA) uses pre-compiled or customized gene sets of functionally related genes to identify the affected biological processes.¹⁰ Accompanied with differential expression of proteins, bioinformatic analyses can find overlapped hub proteins associated with clinical traits and pathophysiological processes of diseases. However, the use of a single method to identify hub proteins may not be accurate, as such use may not include the full spectrum of the changes caused by diseases. Therefore, a comprehensive bioinformatic analysis should be adopted in order to identify a potential correct target for effective therapeutics. 

Label-free proteomics was used in this study to analyze differences in protein components of meibomian gland secretions between patients with BKC and healthy individuals for identifying differentially expressed proteins and hub proteins in order to provide a theoretical basis for understanding the pathogenesis of BKC. 

Materials and Methods

Patient Enrollment

A total of 13 patients with BKC and five healthy subjects were enrolled at Shenzhen Eye Hospital, Guangdong, China. Patients with BKC had a history of recurrent conjunctival or corneal lesions based on blepharitis. Nine of the patients revisited after 1 month of standard BKC treatment as follows: TobraDex ophthalmic suspension (Alcon, Belgium) was applied four times a day, and TobraDex ophthalmic ointment (Alcon) was used once nightly. After 2 weeks, the therapy was changed to continuous use of 0.02% fluorometholone eyedrops (Santen Pharmaceutical Co., Osaka, Japan) four times a day for 2 weeks. Levofloxacin eyedrops (Santen Pharmaceutical Co.), four times a day, were also added, combined with dextran and hypromellose eyedrops (Alcon), four times a day for 1 month. The meibomian gland was massaged once a week, and the eyelid margin was cleaned by Oust Demodex cleanser (OCuSOFT, Rosenberg, TX) once a day. Five patients were lost to follow-up. Patients were excluded who were allergic to fluorescent sodium, had systematic diseases such as Sjögren's syndrome or graft-versus-host disease, were receiving treatment that might affect meibomian gland function or morphology, or had a medical history of mental disorders or metabolic diseases such as depression, hypertension, or diabetes. Clinical traits that were collected for further analysis included age; sex; disease course, severity, and clinical manifestations such as corneal neovascularization; cleanliness of the eyelid margin; palpebral margin score; secretion characteristics; and meibomian gland morphology. 

Ocular surface scores were used to evaluate corneal neovascularization, cleanliness, palpebral margin, secretion characteristics, and meibomian gland morphology, defined as the maximum score between the two eyes. The cleanliness of the eyelid margins was divided into three grades according to the cleanliness of the eyelashes. The eyelid margin, meibomian gland secretions, and meibomian gland structure and morphology were quantitatively scored according to the standard established by the expert consensus on diagnosis and treatment of meibomian gland dysfunction (MGD) in China.¹¹ The corneal neovascularization score was graded on three levels according to the depth of corneal invasion (Supplementary Fig. S1). The detailed definitions and classifications are provided in Supplementary Table S1. 

This study followed the tenets of the Declaration of Helsinki and was approved by the Ethics Committee of Shenzhen Eye Hospital (No. 20200628-10). All subjects signed informed consents. A flowchart of patient inclusion and the proteomic analysis of meibomian gland secretions is provided in Figure 1. 

Figure 1.

 

Flowchart of the patient inclusion and proteomic analysis.

Sample Collection and Preparation From Meibomian Gland Secretions

After hot compression on the eyes and topical anesthesia, the patient’s conjunctival sac, upper and lower eyelid margins, and eyelashes were cleaned with sterile saline. Then, the upper and lower eyelids were turned over by a technical clinician, and the conjunctival surface of the meibomian gland was pressurized using a sterile round glass rod. Meibomian gland secretions were later collected into sterile 1.5-mL centrifuge tubes and stored at −130°C. 

Lysis solution (100 mL), used for the radioimmunoprecipitation assay (RIPA), was added to the meibomian gland secretions obtained from each patient. Then, 2 µL of phosphatase inhibitors and 1 µL of phenylmethylsulfonyl fluoride were then added to each 100 µL of RIPA lysis solution. The supernatant was transferred to a new 1.5-mL centrifuge tube, followed by ultrasound oscillation for 15 minutes at 4°C and centrifugation at 16,000g for 20 minutes. The protein concentration was determined by the bicinchoninic acid assay (BCA), which measured protein concentration by detecting the formation of Cu⁺, chelated with bicinchoninic acid from Cu²⁺ by the biuret complex in alkaline solutions of proteins at an absorbance of 562 nm.¹² Aliquots of lysates were mixed with 200 µL of 8-M urea in a Nanosep centrifugal device (Pall Corporation, New York, NY). The samples were centrifuged at 14,000g at 20°C for 20 minutes. All subsequent centrifugation steps were performed under the same conditions. The concentrates were diluted with 200 µL of 8-M urea in 0.1-M Tris-HCl (pH 8.5) and were then was centrifuged. Proteins were reduced with 10-mM dithiothreitol for 2 hours at 56°C. After that, the samples were incubated in 5-mM iodoacetamide for 30 minutes in the dark to block reduced cysteine residues followed by centrifugation. The resulting concentrates were diluted with 200 µL of 8-M urea in 0.1-M Tris-HCl (pH 8.0) and concentrated again. This step was repeated twice, and the concentrates were subjected to proteolytic digestion overnight at 37°C. The digests were collected by 390g centrifugation at 45°C. 

Label-Free Proteomic Analysis Using Liquid Chromatography–Tandem Mass Spectrometry

The lyophilized peptide fractions were resuspended in double-distilled water containing 0.1% formic acid, and 2-µL aliquots were loaded into a nanoViper C18 trap column (3 µm and 100 Å; Thermo Fisher Scientific, Waltham, MA). Online chromatography separation was performed with the EASY-nLC 1200 System (Thermo Fisher Scientific). The trapping and desalting procedures were carried out with a volume of 20 µL 100% solvent A (0.1% formic acid). Then, an elution gradient of 8% to 38% solvent B (80% acetonitrile, 0.1% formic acid) at a flow rate of 300 nL/min (0–40 minutes, 5%–38% solvent B; 40–42 minutes, 38%–100% solvent B; 42–50 minutes, 100% solvent B) over 50 minutes was used with an analytical column (50 µm × 15 cm C18, 3 µm and 100 Å). Data-dependent acquisition (DDA) mass spectrum techniques were used to acquire tandem mass spectrometry (MS/MS) data with a Q Exactive Mass Spectrometer (Thermo Fisher Scientific) fitted with a Nanospray Flex Ion Source (Thermo Fisher Scientific). Data were acquired using an ion spray voltage of 1.9 kV and an interface heater temperature of 275°C. The mass spectrometer was operated with full MS scans. For DDA, survey scans were acquired in 250 ms, and up to 20 product ion scans (50 ms) were collected. Only spectra with a charge state of 2 to 4 were selected for fragmentation by higher energy collision energy. Dynamic exclusion was set at 25 seconds. 

The MS/MS data were analyzed for protein identification and quantification using MaxQuant 1.5.6.0 (Max Planck Institute of Biochemistry, Planegg, Germany). The local false-discovery rate (FDR) was 1.0% after searching against the Homo sapiens genome with a maximum of two missed cleavages and one missed terminal cleavage. The following settings were selected: oxidation (M), acetylation (protein N-term), deamidation (NQ), pyro-glu from E, and pyro-glu from Q as variable modifications and carbamidomethylation (C) as a fixed modification. Precursor and fragment mass tolerances were set to 10 ppm and 0.05 Da, respectively. Before statistical analysis, protein expression data were filtered by considering only proteins identified in at least 50% of the biological samples. The K-nearest neighbor method was used to impute the data matrix and manage missing values. Normalization was achieved by dividing by the intragroup mean and performing log2 function transformation. 

Construction of Coexpressed Gene Modules and Identification of Advanced Plaque-Related Modules

WPCNA was performed using R 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria) to identify important proteins that correlated most with eight clinical traits. Samples were clustered by hierarchical clustering analysis, and outlier samples were detected. An appropriate soft threshold value (β) was chosen to make the coexpression network approximate biologically significant scale-free topology (scale independence > 0.8). The adjacency matrix of correlation similarity between two proteins was calculated and converted into a topological overlap matrix (TOM). The module eigenprotein (ME) value of each module was the first component of the protein expression matrix of the corresponding module. The ME values were related to clinical traits (diagnosis of BKC, sex, age, severity of BKC, corneal neovascularization, course of disease, cleanliness of eyelid margin, palpebral margin score, secretion characteristics, and meibomian gland morphology of samples) using hierarchical clustering analysis and Spearman’s correlation analysis. The modules most associated with the clinical traits were defined as those with the highest correlation coefficients, with at least six clinical traits (P < 0.05). Spearman’s correlation analysis was applied to calculate protein significance based on the relationship between the protein intensity spectrum and clinical features and to calculate module membership based on the connectivity between ME, the first component of the protein expression matrix, and their protein intensity spectrum. Proteins with significance > 0.2 and module membership > 0.8 were defined as proteins associated with clinical traits.¹³ Proteins overlapping in the six most relevant clinical features were extracted for further analysis and displayed in Venn diagrams. 

Functional Enrichment Analysis of the Proteins in Modules Most Relevant to Clinical Features

Protein names in the most relevant modules were imported into the DAVID 6.8 database (https://david.ncifcrf.gov/) for Gene Ontology (GO) enrichment analysis,¹⁴ which describes biological processes, cellular components, and molecular functions enriched in gene sets, and for Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis,¹⁵ which describes pathways enriched in gene sets. 

GSEA of the Proteins in Modules Most Relevant to Clinical Traits

The names of total proteins detected by liquid chromatography–tandem mass spectrometry (LC–MS/MS) were input into GSEA software v4.0.3 (http://www.gsea-msigdb.org/gsea/index.jsp). In the run program settings, the gene set “msigdb.v.7.2.symbols.gmt,” consisting of hallmark gene sets, positional gene sets, curated gene sets, regulatory target gene sets, computational gene sets, ontology gene sets, oncogenic signature gene sets, immunological signature gene sets, and cell type signature gene sets, was chosen as the reference gene set, and the parameter “number of permutations” was set as 1000. Samples were divided into a high expression group and a low expression group, and P < 0.05 was considered significantly enriched. 

Identification of Hub Proteins by Overlapping Results of WPCNA and Differential Protein Analysis

Two principal component analyses were performed using clinical trial and proteomic data to assess whether samples from the patients and the healthy controls could be distinguished. Differentially expressed proteins between the two groups were identified based on log2(fold change [FC]) > 2 and FDR < 0.05 by using the limma package. The results were illustrated by volcano and heatmap plots.¹⁶ The differential proteins were cross-referenced with the overlapping proteins identified in the WPCNA analysis to obtain hub proteins responsible for BKC. A Venn diagram was obtained at a bioinformatics and evolutionary genomics website (http://bioinformatics.psb.ugent.be/webtools/Venn/), and the hub protein specificity of the immune cells was identified based on a human protein mapping website (https://www.proteinatlas.org/). According to log2(FC) > 2 and FDR < 0.05, differentially expressed proteins were identified between patients with BKC before treatment and 1 month after treatment. 

Statistical Analysis

Quantitative variables conforming to a normal distribution were expressed as mean ± SD. The quantitative variables not conforming to a normal distribution were expressed as the median (1st quartile, 3rd quartile). Categorical variables are described as a proportion of the cases. SPSS Statistics 25.0 (IBM, Chicago, IL) and Prism 8.0.1 (GraphPad Software, San Diego, CA) were used for statistical analysis and image processing. The independent-sample t-test was utilized to compare the clinical characteristics of patients with BKC with healthy controls, and the paired-sample nonparametric test was used to compare the clinical characteristics of patients with BKC before treatment and after 1 month of treatment. The χ² test was applied for comparing samples among categories. Mating graphs were used to describe changes in corneal neovascularization, cleanliness of the eyelid margin, palpebral margin score, secretion characteristics, and meibomian gland morphology in nine patients with BKC after 1 month of standard treatment. 

Results

Baseline Characteristics and Ocular Surface Scores of Patients with BKC and Healthy Controls

The average age of the 18 enrolled subjects was 25.0 years (range, 16.8–27.5), and 66.7% were female. There were no differences in age or sex between the patients with BKC and the healthy controls. The average course of 13 patients with BKC was approximately 2.00 years (range, 0.75–3.00) years, characterized by lower cleanliness of the eyelid margin, higher palpebral margin score, more serious clinical manifestations of secretions, and more damaged meibomian gland morphology when compared to those from the healthy controls (P < 0.05). Corneal neovascularization exhibited a trend of greater severity in patients with BKC compared with the healthy controls, but the difference was not statistically significant (Table, Fig. 1). 

Table.

 

Baseline Characteristics and Ocular Surface Scores of Patients With BKC and Healthy Controls

 

WPCNA for the Identification of Important Proteins Closely Correlating With Eight Clinical Traits

We detected no outliers in the sample clustering trees based on 342 included proteins. The detected proteins from patients with BKC and the healthy controls were clearly distinct (Fig. 2A). The proteins were grouped into four modules based on the correlation between protein expression and module eigenprotein: turquoise module (115 proteins), blue module (87 proteins), brown module (73 proteins), and gray module (67 proteins) (Fig. 2B). The relationship of the modules with clinical traits was then explored. The turquoise module was the one most positively associated with the diagnosis of BKC, severity, disease course, corneal neovascularization, cleanliness, palpebral margin score, secretion characteristics, and meibomian gland morphology (Fig. 2C). By using the standard of protein significance (the correlation between protein and clinical traits) > 0.2 and a correlation between ME and the total protein expression profile > 0.8, the closely correlated proteins for the above-mentioned eight clinical traits were identified (Fig. 2D). Fourteen proteins were obtained by overlapping the closely correlated proteins in six traits with the turquoise module (diagnosis of BKC, severity, disease course, cleanliness, secretion characteristics, and meibomian gland morphology) (Fig. 2E). Although the blue module was negatively associated with clinical traits, it was not further used due to the insignificant difference between patients with BKC and the healthy controls. 

Figure 2.

 

WPCNA was used to identify important proteins closely correlated with eight clinical traits. (A) Sample clustering to detect outliers. (B) Cluster dendrogram of the turquoise, blue, brown, and gray modules. (C) The relationship between ME and clinical traits. (D) Table of the closely correlated proteins in the turquoise module of eight clinical traits. (E) Venn plot of proteins correlated highly with six traits significantly associated with the turquoise module.

GO and KEGG Pathway Analyses

As shown in Supplementary Figures S2A and S2B, the proteins in the turquoise module closely participated in immunological biological processes associated with pathogen invasion. The proteins in the turquoise module were mostly located extracellularly. These proteins are, in fact, correlated with calcium-dependent proteins, phospholipids, ion-binding activities, and molecular functions, and they participate in phagosome, leukocyte migration, and glucose metabolic pathways. Therefore, our results support that the proteins in the turquoise module are associated with immunology (Supplementary Fig. S2). 

Identification of Hub Proteins by Overlapping Results of WPCNA and Different Protein Analyses

Although patients with BKC and the healthy controls can be differentiated using clinical traits with good representative ability (Fig. 3A), the two groups were not completely distinguished based on 342 proteins (Fig. 3B). Thus, to identify the differential proteins between patients with BKC and the healthy controls, 14 proteins with FC > 4 and FDR < 0.05 were selected (Fig. 3C), nine of which were upregulated in patients with BKC and four downregulated (Fig. 3D). The proteins in the turquoise module overlapped with different proteins with FC > 4; five overlapping proteins were obtained: S100 calcium-binding protein A8 (S100A8), S100 calcium-binding protein A9 (S100A9), annexin A3 (ANXA3), lipocalin 2 (LCN2), and adipocyte plasma membrane associated protein (APMAP). After identifying the cellular source of these five proteins at the website Protein Atlas (https://www.proteinatlas.org/), four immune cell-specified proteins (S100A8, S100A9, ANXA3, and LCN2) that correlated closely with clinical traits and were significantly upregulated in patients with BKC were chosen as the hub proteins. Such proteins were all specific to neutrophils (Fig. 3E). 

Figure 3.

 

Identification of hub proteins by overlapping proteins between the turquoise module and differential proteins with a FC > 4. (A) Principal component analysis by clinical traits. (B) Principal component analysis by proteomics. (C) Volcano plot of different proteins with FC > 4 and FDR < 0.05. (D) Heatmap of differential proteins with FC > 4 between patients with BKC and the healthy controls. (E) Venn plot and comparison of proteins in the turquoise module and differential proteins with a FC > 4.

GSEA of 342 Proteins

Seven gene sets most enriched in BKC, according to GSEA, corresponded to 342 proteins. Their defense response to bacteria and antimicrobial humoral response were enriched, suggesting that the immunological gene sets are, in fact, related to BKC. S100A8 and S100A9 were common proteins correlated to all seven gene sets (Supplementary Fig. S3). In addition, the acute-phase proteins were similar between the two groups, indicating that the traditional acute inflammatory proteins are unaffected by BKC (Supplementary Table S3). 

Mating Graphs of Clinical Symptoms of Patients With BKC Before and After Treatment

The palpebral margin scores and secretion characteristics were significantly decreased in patients after treatment, although cleanliness tended to decrease in patients after treatment. However, meibomian gland morphology was unchanged after treatment (Fig. 4). 

Figure 4.

 

Mating graphs of cleanliness (A), palpebral margin score (B), secretion characteristics (C), and meibomian gland morphology (D) of patients with BKC before and after treatment.

Discussion

To the best of our knowledge, this is the first study to investigate proteomic differences in meibomian gland secretion between patients with BKC and healthy controls. Four hub proteins (S100A8, S100A9, ANXA3, and LCN2) were identified as being significantly associated with the diagnosis and manifestations of BKC, and they remained elevated after 1 month of treatment, indicating that chronic inflammation continued during the BKC process. This study sheds light on protein changes in meibomian gland secretions in BKC, providing further understanding of potential protein therapeutic targets from meibomian gland secretions in both healthy and disease status. 

The eyelid margin, which is 2 mm in width, is divided into a front portion with the eyelashes and a rear portion with the meibomian duct opening, close to the eyeball. Therefore, blepharitis is divided into anterior blepharitis, with eyelash root inflammation, and posterior blepharitis, with MGD and obstruction. When the disease involves the cornea and conjunctiva with inflammatory changes, it is referred to as BKC, which has various clinical manifestations with chronic and recurrent characteristics. Due to repeated ocular surface inflammatory stimulation, BKC, which tends to occur in teenagers, can lead to the formation of scars in the optical area of the cornea and even more serious ocular surface injury, such as corneal stromal dissolution, posterior elastic layer bulging, and corneal perforation. These conditions can cause irreversible vision loss as well as adverse effects on the physical and mental development of children. Moreover, early-stage BKC is often accompanied by a history of recurrent chalazia. In such cases, the morphology and functional performance of the meibomian gland in patients with BKC are worse compared with healthy individuals. Therefore, the role of the meibomian gland in the pathogenesis of BKC cannot be ignored; thus, the proteins secreted from the meibomian gland that are closely related to the occurrence and development of the disease were the focus of this study. 

In this study, we found that the meibomian secretion proteins in patients with BKC are mainly associated with inflammation. Via GO, KEGG, and GSEA, we have shown that most meibomian proteins associated with clinical traits in patients with BKC are indeed closely associated with the inflammatory immune response. Through WPCNA analysis, we have demonstrated that 14 proteins are closely associated with clinical symptoms: cytochrome c oxidase subunit 2 (COX2); aspartate aminotransferase, mitochondrial (GOT2); annexin A1 (ANXA1); guanine nucleotide-binding protein G subunit alpha-2 (GNAI2); S100A8; S100A9; carcinoembryonic antigen-related cell adhesion molecule 5 (CEACAM5); ubiquitin-60S ribosomal protein L40 (UBA52); ANXA3; cytochrome c oxidase subunit 4 isoform 1 (COX4I1); neutrophil defensin 3 (DEFA3); LCN2; brain acid soluble protein 1 (BASP1); and APMAP. These proteins are mainly involved in the inflammatory response, neovascularization, and lipid metabolism processes. Previous studies have shown that keratin-1 overexpression and MGD might occur in meibomian gland tissues treated with proinflammatory stimuli, interleukin-1β (IL-1β), or Staphylococcus aureus crude extracts, suggesting that inflammation plays an important role in the damage caused to meibomian gland tissues.^17,18 Inflammation is also the core mechanism of corneal neovascularization. COX2, LCN2, S100A8, S100A9, and BASP1 may all promote neovascularization. Increased cholesterol in the meibomian gland is an important pathological mediator of MGD.^19,20 ANXA1, S100A8, S100A9, BASP1, and APMAP may all be involved in lipid metabolism; for example, ANXA1 can increase the expression of ABCA1, which promotes cholesterol efflux and reduces cellular cholesterol accumulation.²¹ BASP1 is a cofactor of WT1, and genes in the cholesterol biosynthesis and lipid transport pathways are direct targets of WT1.²² The hub proteins identified in this study are S100A8, S100A9, LCN2, and ANXA3. 

S100A8 and S100A9 act as dimers; they are important calcium-binding proteins and bind to the receptor for advanced glycation endproducts (RAGE) and various Toll-like receptors to induce cell differentiation, apoptosis, proliferation, inflammation, and other processes. Both proteins have been found to be expressed at elevated levels in MGD and corneal neovascularization and to regulate neutrophil chemotaxis and inflammation in ocular surface diseases.^23–27 Moreover, S100A8 and S100A9 are detected in the tears of patients with dry eye disease and MGD. Its expression level correlates positively with the severity of MGD and the symptoms of red eye and blurred vision in patients with dry eye.²⁸ Furthermore, it can reduce adenosine triphosphate (ATP) binding cassette subfamily G member 1 (ABCG1) expression by regulating IL-22 in macrophage cholesterol excretion and can affect reverse cholesterol transport.²⁹ Nonetheless, its involvement in the process of meibomian gland lipid transport requires further exploration. These two proteins are closely related to clinical symptoms, showing the largest change in expression, which might be the warning proteins of BKC. 

As an important member of the annexin family, ANXA3 is involved in a series of important physiological processes, including cell proliferation and differentiation, apoptosis, and signal transduction. ANXA3 has been shown to be significantly elevated in several inflammatory diseases, including sepsis and cancer.³⁰ However, the role of ANXA3 in ocular surface diseases remains unclear. LCN2 is involved in multiple processes, such as apoptosis, fatty acid transport, regulation of inflammation, and metabolic stabilization.³¹ LCN2 protects cells during bacterial infection, causing acute inflammation; however, it has also been shown to be responsible for chronic inflammation when highly expressed. A high expression level of LCN2 has been detected in the aqueous humor or vitreous fluids of patients with idiopathic acute anterior uveitis, central retinal vein occlusion, or rhegmatogenous retinal detachment. Increased expression of LCN2 can also be detected in the retina of age-related macular degeneration (AMD) donors and in mouse models of AMD, which is an upstream regulator of apoptosis in photoinduced retinal degeneration models and a potential therapeutic target for AMD.^32,33 Moreover, expression of LCN2 is increased in the mouse corneal neovascularization model of alkali burn and in the mouse model of Ikaros family zinc finger 1 transgenic mucosal skin inflammation.^34,35 

Consistent with previous proteomic and biochemical studies, S100A8, S100A9, ANXA1, glutathione S-transferase P, and heat shock protein 70 (HSP1A1),^7,36–38 which are considered to be damage-associated molecular patterns (DAMPs), are upregulated in patients with MGD compared to healthy controls. These DAMPs might be associated with Toll-like receptor 4 (TLR4); however, the mechanism of DAMP-induced TLR4 pathway activation in meibomian gland diseases remains unclear. We have also studied other proteins in the turquoise module and proteins with more than a fourfold change. Increased APMAP and decreased Niemann–Pick type C intracellular cholesterol transporter 1 (NPC1) are associated with decreased lipid deposition in cytoplasm^39,40; however, whether these protein changes are correlated with the proinflammatory profiles must still be investigated. Studying the interplay of inflammation and lipid metabolism may provide a potential direction for understanding the chronic inflammation in the process of BKC. A recent case-controlled study used multiplex bead analysis to show that levels of inflammatory cytokines, including interleukin (IL)-6, IL-8, interferon-γ (IFN-γ), IL-12p70, and intercellular adhesion molecule-1 (ICAM-1), were increased in the meibomian gland secretions of patients with BKC, supporting the suggestion that acute-phase inflammation is involved in the process of BKC.⁴¹ Unfortunately, these acute inflammatory markers could not be detected by the MS in this study, probably because their expression levels did not reach the sensitivity level of the MS assay. 

In this study, changes in acute-phase proteins in the healthy group and BKC group and changes in hub proteins in patients with BKC before and after 1 month of treatment were not significantly different, indicating that BKC is not involved in acute ocular surface inflammation but is possibly involved in a chronic inflammatory reaction with an immune response at the core. Therefore, chronic inflammation of the eyelid margin plays a critical role in BKC. The corneal and conjunctival inflammation can be quickly relieved by the use of local hormones, antibiotics, and ocular surface physical therapies as clinical treatment, even though the chronic inflammation of the eyelid margin may not be fundamentally improved. Overall, this persistent chronic inflammation of the eyelid margin is the basis for recurrent BKC. Thus, clinicians should extend the treatment cycle for BKC. Only when the chronic inflammation of the eyelid margin is effectively controlled can BKC treatment result in eye surface stability. In further proteomic exploration, we have discovered that increases in the meibomian proteins S100A8, S100A9, ANXA3, and LCN2 may be associated with the severity of BKC inflammation (Fig. 5). Therefore, dynamic observation of the changes in these hub proteins from meibomian secretions has critical significance for individualized diagnosis and treatment of patients with BKC. However, further animal model validation of BKC is needed. 

Figure 5.

 

Chronic inflammation participates in the process of BKC. Pathogens stimulate upregulation of the chronic inflammatory proteins S100A8, S100A9, ANXA3, and LCN2 in meibomian secretions. These inflammatory proteins are associated with ocular surface traits in the disease state, including toothpaste-like meibomian secretions, unclean palpebral margins, pachyblepharon, and impaired meibomian glands. Moreover, 1 month of treatment does not affect the levels of these chronic inflammatory proteins.

Due to the limitations of the study, future studies are needed to investigate the expression of these proteins and verify their effect on disease symptoms. Moreover, several proteins associated with disease severity must be further investigated with regard to their association with BKC. 

Conclusions

S100A8, S100A9, ANXA3, and LCN2 are important BKC-associated proteins due to their persistent high levels even after 1 month of BKC treatment. The findings indicate that these proteins are involved in chronic inflammation of BKC. 

Acknowledgments

Supported by a grant from the Science, Technology and Innovation Commission of Shenzhen Municipality (GJHZ20190821113607205). 

Disclosure: J. Su, None; H. Li, None; B. Lin, None; S. Li, None; X. Zhou, None; W. Li, None; P. Guo, None