Alteration of Meibum Lipidomics Profiling in Patients With Chronic Ocular Graft-Versus-Host Disease

Abstract

Purpose: To investigate the characteristics of the lipid profiling in meibum of patients with chronic ocular graft-versus-host disease (coGVHD) and to detect the potential influence of anti-inflammatory therapy on these differential lipids.

Methods: This cross-sectional study included 25 coGVHD patients and 13 non-coGVHD after allogeneic hematopoietic stem cell transplantation. Among those with coGVHD, 14 had prior topical treatment (coGVHD_(T)), and 11 did not (coGVHD_(WT)). All participants completed ocular surface disease index questionnaire and received slit lamp examination, Schirmer's test without anesthesia, ocular surface interferometer, and meibography. Binocular meibum was collected and pooled for lipidomic analysis by liquid chromatography–mass spectrometry.

Results: One hundred and twenty differential lipid species were found among the three groups (96 of coGVHD_(WT) vs. non-coGVHD, 78 of coGVHD_(WT) vs. coGVHD_(T), and three of non-coGVHD vs. coGVHD_(T)). Compared with non-coGVHD group, coGVHD_(WT) group had a significant abnormality of meibum composition, showing a significant decrease in glycerolipids, and an increase in glycerophospholipids and sphingolipids. Similar changes were also observed when coGVHD_(WT) versus coGVHD_(T). CoGVHD severity was negatively associated with mono-unsaturated triglycerides (TG), (β = −214.7; 95% CI, −363.9 to −65.5; P = 0.006) and poly-unsaturated TG (β = −4019.9; 95% CI, −7758.1 to −281.6; P = 0.036). Intensity of immunosuppression was negatively associated with mono-unsaturated TG (β = −162.4; 95% CI, −268.6 to −56.2; P = 0.004) and positively associated with phosphatidylcholine (β = 332.0; 95% CI, 19.2-644.8; P = 0.038).

Conclusions: Altered meibum in coGVHD is characterized by a decrease of glycerolipids and an increase of glycerophospholipids and may be significantly reversed by topical anti-inflammatory therapy.

Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a curative therapy for various hematological diseases.^1,2 Chronic ocular graft-versus-host disease (coGVHD) is the most common ocular complication of patients after allo-HSCT.^3,4 CoGVHD may appear with a panoply of clinical manifestations such as severe dry eye, extensive inflammation, fibrosis at the ocular surface, symblepharon, corneal perforation, etc.^5,6 Meibomian gland dysfunction (MGD) is also commonly seen in patients with coGVHD, although its relation to coGVHD remains elusive.^7,8 Recently, a growing amount of evidence suggests that allo-HSCT recipients have significant alterations in composition, conformation, and physicochemical properties of meibum lipids compared to the healthy controls based on infrared spectroscopy. Allo-HSCT recipients had more ordered and stiffer meibum with increased phase transition temperature and magnitude of the phase transition.^9,10 Additionally, meibum derived from dry eye patients after allo-HSCT contained significantly more straight chain and fewer iso-chain hydrocarbons.¹¹ The change of structure and physicochemical properties in meibum after allo-HSCT has been documented, but a comprehensive analysis using the highly sensitive and accurate method of liquid chromatography-mass spectrometry (LC/MS) has yet to be conducted.¹² 

Meibum-derived lipids constitute the main component of the tear film lipid layer, which could inhibit tears from evaporating and help maintain ocular surface homeostasis.¹³ Wax esters, (O-acyl)-omega-hydroxy fatty acids, cholesteryl esters, minor lipid classes such as free fatty acids, fatty alcohol, triglycerides (TG), diacylglycerol (DG), monoglyceride, ceramide, were reported as the main lipid species in normal humans meibum.¹⁴ Lipid profiles of patients with MGD often have distinct lipid composition abnormalities including a decrease in amounts of unsaturated TG, differential expression of specific (O-acyl)-omega-hydroxy fatty acids, alteration of molar ratio of cholesteryl ester/wax ester, and aberrations in sphingolipid metabolism, and these alterations have been reported to be closely associated with disruption of tear film stability and ocular discomfort.^15–18 

The world prevalence of MGD is as high as 36.8%; however, its pathologic mechanism remains uncertain.¹⁹ Recently, inflammation has been identified as a major risk factor for MGD. Inflammatory stimuli could result in significant proinflammatory cytokine response and meibum abnormalities in the murine model and human meibomian gland epithelial cells, and these changes were reversed by anti-inflammatory interventions.^20,21 Prominent inflammation and fibrosis of meibomian gland microstructure have been detected and depictured in patients with coGVHD.²² Similarly, we previously reported that there was a dramatic inflammatory cell infiltration within MGs in patients with coGVHD.²³ Therefore, on the background of coGVHD-related chronic inflammation, whether and how meibum lipid metabolism changed remains a puzzle. If so, does anti-inflammatory treatment for coGVHD have any impact on such change? Hence, this study aims to provide an insight into the possible changes in meibomian lipid homeostasis of patients with coGVHD and to investigate the potential influence of topical anti-inflammatory therapy by comparing the lipid composition among patients without coGVHD, with coGVHD, and those who had been treated. 

Material and Methods

Reagents

High-performance liquid-chromatography or MS grade organic solvents (isopropanol, acetonitrile, methanol, chloroform, and water) and ammonium formate (99.98%) were purchased from Thermo Fisher Scientific (Waltham, MA, USA), Sigma-Aldrich (St. Louis, MO, USA), TCI (Tokyo, Japan), and Wokai (Beijing, China). Deionized water was obtained from a MilliQ purification system (Millipore, Bedford, MA, USA). 

Human Subjects and Meibum Sample Collection

This prospective, cross-sectional study was conducted at Zhongshan Ophthalmic Center between May 2021 and December 2021. This study adhered to the tenets of the Declaration of Helsinki and the whole study procedure was approved by the Ethics Committee of the Zhongshan Ophthalmic Center (approval number: 2019KYPJ135). All participants were recruited from the ocular surface and tear film disease outpatient clinic and written informed consent was obtained from all subjects before examinations. Participants aged 16 to 65 years who were assessed after at least three months after transplantation were included. Grouping situations were as follows: (1) coGVHD_(WT) group (coGVHD_{(Without Treatment)}), newly diagnosed coGVHD, had no history of standardized ocular treatment (had no use of medication or only use of preservative-free sodium hyaluronate eye drops); (2) coGVHD_(T) group (coGVHD_(Treatment)), previously diagnosed with coGVHD, had received continuous ocular anti-inflammatory therapy (topical corticosteroids or tacrolimus, twice daily) except for other ocular medicine such as anti-infective eye drops, and had an improvement and remained stable for at least three months; (3) non-coGVHD group, did not meet the coGVHD diagnostic criteria and with no symptoms/signs of ocular surface disease. The age and gender were matched among the three groups above. 

Exclusion criteria were as follows: (1) history or presence of ocular surface infection, allergy, or any other active ocular surface diseases; (2) history of ocular trauma; (3) history of ocular surgery, contact lens wearer, or continued use of local medications within the previous three months; (4) history or present illness of diabetes mellitus, thyroid diseases, sex hormone-related disorders, or any systemic disease that may affect the ocular surface; (5) asymmetry of conjunctival hyperemia/corneal fluorescence staining grade between two eyes ≥ 1 grade based on the International Chronic Ocular Criteria Ocular GVHD (ICCGVHD) criteria²⁴; (6) unable to cooperate with examinations or meibum collection; (7) the number of meibomian gland obstruction without meibum exceeds half of the total per eyelid. The coGVHD was diagnosed and graded according to the ICCGVHD criteria, which include the ocular surface disease index (OSDI) questionnaire, Schirmer I test without anesthesia, bulbar conjunctival injection, and corneal fluorescein staining. 

Each patient underwent binocular detailed examinations under the slit lamp microscope. Lid margin abnormality (vascular engorgement, glandular orifices obstruction, lid margin irregularity, and anterior or posterior malposition of the mucocutaneous junction) and meibum quality score were graded and recorded as previous studies.^25,26 Corneal staining was based on the National Eye Institute/Industry grading scheme (0–15).²⁷ Average lipid layer thickness was detected and documented using the LipiView Ocular Surface Interferometer (TearScience Inc, Morrisville, NC, USA).²⁸ Meiboscore for upper and lower eyelids were determined using noncontact meibography, and the total score (0–6) was documented.²⁹ 

The eyelid rims were first cleaned with cotton swabs. Next, meibum samples were collected from the subjects' eyelids by gently squeezing the margin with a gloved finger and using a glass rod with a round head (as shown in the supplementary materials, Supplementary Fig. S1). The collected binocular meibum from all meibomian glands (approximately 0.5 mg) was then transferred to a sterile 1.5 mL Eppendorf tube (Thermo Fisher Scientific) using the glass rod and stored at −80°C for further processing. 

Lipid Extraction and Quantitative Analysis of Lipids Using LC/MS

Before analysis, samples were prepared and lipids were extracted as following procedures: Transfer 15 µL of each sample into 2 mL centrifuge tubes, add 750 µL of chloroform methanol mixed solution (2:1) (pre-cooled at −20°C), vortex for 30 seconds; Put on the ice for 40 minutes, add 190 µL H₂O, vortex for 30 seconds, and still put on the ice for 10 minutes; Spin in a centrifuge at 12,000 rpm for five minutes at room temperature and transfer 300 µL lower layer fluid into a new centrifuge tube; Add 500 µL of chloroform methanol mixed solution (2:1) (pre-cooled at −20°C), vortex for 30 seconds; Spin in a centrifuge at 12,000 rpm for five minutes at room temperature and transfer 400 µL lower layer fluid into the same centrifuge tube above. Samples were concentrated to dryness in a vacuum; Dissolve samples with 200 µL isopropanol, and the supernatant was filtered through a 0.22 µm membrane to obtain the prepared samples for LC/MS. 

Chromatographic separation was used with an ACQUITY UPLC BEH C18 (100 × 2.1 mm, 1.7µm, Waters) column maintained at 50°C. The temperature of the autosampler was 8°C. Gradient elution of analytes was carried out with acetonitrile: water = 60:40 (0.1% formic acid +10 mM ammonium formate) and isopropanol: acetonitrile = 90:10 (0.1% formic acid +10 mM ammonium formate) at a flow rate of 0.25 mL/min. Injection of 2 µL of each sample was done after equilibration. An increasing linear gradient of solvent C (v/v) was used as follows. 

The ESI-MSn experiments were used with the spray voltage of 3.5 kV and −2.5 kV in positive and negative modes, respectively. Sheath gas and auxiliary gas were set at 30 and 10 arbitrary units, respectively. The capillary temperature was 325°C, respectively. The Orbitrap analyzer scanned over a mass range of m/z 150 to 2000 for the full scan at a mass resolution of 35,000. Data-dependent acquisition MS/MS experiments were performed with an HCD scan. The normalized collision energy was 30 eV. Dynamic exclusion was implemented to remove some unnecessary information in MS/MS spectra. 

Data Processing

Data were analyzed using LipidSearch software version 4.1 (Thermo Fisher Scientific). The classes and species of lipids were based on the peaks in LC/MS data. Lipid species annotation was performed through a combination of the online database HMDB, PubChem and LIPID MAPS. After the assessment of lipid identification, peak extraction, peak alignment, and quantification was performed. Before multivariate data statistics analysis, the normalized data were Pareto-scaled. The principal component analysis (PCA) was processed for class discrimination, and the differential lipid screening was performed according to the criteria: one-way ANOVA P value ≤ 0.05.³⁰ 

The relative contents of meibum lipid metabolites were represented as relative intensity. Statistical analyses were performed using SPSS 24.0 (SPSS, Chicago, IL, USA) or R software (Version 3.1.0). Mean values of both eyes per patient were calculated and included in statistical analysis. Values were presented as mean ± standard deviation or median (interquartile range) for continuous variables, and frequency (percentage) for categorical variables. The normality of continuous variables was checked using the Shapiro-Wilk normality test and histogram. 

Quantitative comparisons were made among 11 newly diagnosed coGVHD patients, 14 coGVHD patients receiving treatment, and 13 without coGVHD for all 41 classes, 1964 lipid species were analyzed using the one-way ANOVA to identify lipid species that were significantly different from the other two groups. The lipid levels in patients among the three groups were compared using one-way ANOVA with post hoc Scheffe or Games-Howell tests (adjusted P < 0.05 was considered to be statistically significant). Pearson correlation was performed to determine the lipid species associated with coGVHD severity and other ocular surface parameters. Associations of coGVHD severity, the intensity of immunosuppression, and coGVHD-related lipid content were examined by performing univariable and multivariable linear regression analyses. Variables with a P value < 0.05 in the univariable regression analysis were included in the multivariable regression model. Two-sided P values of < 0.05 were considered statistically significant. 

Results

Demographic Characteristics and Medical History

This study enrolled 38 participants including 13 who had no coGVHD (non-coGVHD), 11 who were newly diagnosed with coGVHD (coGVHD_(WT)), and 14 who had been treated with topical anti-inflammatory treatments (coGVHD_(T)). The majority of participants were young and middle-aged, and the age and gender were matched among the three groups. The time interval between allo-HSCT and the study was significantly longer in the coGVHD_(WT) group compared to both the coGVHD_(T) and non-coGVHD groups (both P < 0.05). coGVHD_(WT) group and coGVHD_(T) group were accompanied by a higher rate of systemic cGVHD than those in the non-coGVHD group (9/11 [81.8%] vs. 14/14 [100%] vs. 4/13 [30.8%], P < 0.001). There was no significant difference in primary disease, radiotherapy, chemotherapy, source of HSCT, donor type, and immunosuppressant taken among the three groups (all P > 0.05). Patient demographics and medical history are shown in Table 1. 

Table 1.

 

Demographic, Transplant Characteristics, and Ocular Surface Parameters of Patients After Allogeneic Hematopoietic Stem Cell Transplantation

 

Comparison of Clinical Characteristics Among the Non-CoGVHD Patients and the CoGVHD Patients With and Without Treatments

Lipid layer thickness, meiboscore, lid margin abnormality score, and meibum quality score did not differ significantly among the three groups (Table 1). Only noninvasive tear film break-up time was significantly lower in the coGVHD_(WT) group than that of the non-coGVHD group (1.45 [0.00, 3.63] seconds vs. 5.25 [3.33, 9.91] seconds, adjusted P = 0.005). coGVHD_(WT) group and coGVHD_(T) group had a higher ICCGVHD score than the non-coGVHD group, respectively (8.5 ± 1.6 vs. 1.8 ± 1.2, adjusted P < 0.001; 7.3 ± 2.2 vs. 1.8 ± 1.2, adjusted P < 0.001). 

Differential Lipids Screening and Functional Enrichment Analysis

Untargeted lipidomic analysis of meibum identified 41 classes, 1964 species among the three groups. The distribution of major lipid components and numbers of differential lipids species were shown in Figure 1A. Most species of meibomian lipids were glycerolipids. PCA of the biological lipid effects led to the classification into three clusters (Fig. 1B). Differentially expressed lipid metabolites were screened based on the PCA model together with the adjusted P value (one-way ANOVA P value ≤ 0.05), and the results showed that the lipid compositions were separated between GVHD_(WT) group and non-coGVHD group, suggesting alterations in the meibomian metabolic profile. However, non-coGVHD group and coGVHD_(T) group were not separated well. MetaboAnalyst was used for functional enrichment analysis in altered lipids from the comparison of coGVHD_(WT) and non-coGVHD groups. Glycerolipid metabolism, sphingolipid metabolism, and glycosphingolipid metabolism were the three most significant altered lipids (Fig. 1C), and their expression patterns were shown in Figures 2A and 3A. 

Figure 1.

 

Distribution of major lipid components and differential lipids among three groups. (A) All meibum lipid species detected by LC/MS in patients after allo-HSCT. The number denote the identified lipid subclasses and the number of lipid species per lipid subclass in meibomian glands. (B) Principal component analysis of the lipid profiles. (C) Pathway enrichment of differential metabolites inferred from comparison between coGVHD without treatment and non-coGVHD groups using MetaboAnalyst. FA, fatty acyls; GL, glycerolipids; GP, glycerophospholipids; PR, prenol lipids; SL, saccharolipids; SP, sphingolipids; ST, sterol lipids; AEA, arachidonoyl ethanolamide; AcCa, acylcarnitine; WE, wax esters; OAHFA, (O-acyl)-ω-hydroxy fatty acids; MG, monoglyceride; LPC, lysophosphatidylcholine; LPG, lysophosphatidylglycerol; MePC, methylphosphocholine; PA, phosphatidic acid; PEt, phosphatidylethanol; PG, phosphatidylglycerol; PI, phosphatidylinositol; LPE, lysophosphatidylethanolamine; LdMePE, lysodimethylphosphatidylethanolamine; PS, phosphatidylserine; CarE, carnitine ester; Co, coenzyme; CerG3GNAc1, simple glucosylceramide series; DGDG, digalactosyldiacylglycerol; Hex3Cer, trihexosylceramide; MGDG, monogalactosyl diacylglycerol; MGMG, monogalactosyl monoacylglycerol; CerP, ceramide phosphate; SM, sphingomyelin; SPH, sphingosine; GM3, ganglioside; AcHexChE, acyl hexosyl cholesterol ester; AcHexCmE, acyl hexosyl campesterol ester; AcHexSiE, acyl hexosyl sitosterol ester; AcHexStE, acyl hexosyl stigmasterol ester; ChE, cholesteryl esters; CmE, campesterol ester; SiE, sitosterol ester; StE, stigmasterol ester; ZyE, zymosterol ester.

Figure 2.

 

Heatmap and relative content of glycerolipids among coGVHD_(WT) (G1), non-coGVHD (G2) and coGVHD_(T) (G3) groups. (A) Heatmap. (B) Relative content of triglycerides among three groups. TG, triglycerides. *Adjusted P < 0.05; **adjusted P < 0.01.

Figure 3.

 

Heatmap and relative content of glycerophospholipids and sphingolipids among coGVHD_(WT) (G1), non-coGVHD (G2) and coGVHD_(T) (G3) groups. (A) Heatmap. (B) Relative content of glycerophospholipids and sphingolipids. BisMePA, Bis-methyl phosphatidic acid; MePC, methylphosphocholine; SPH, sphingosine. *Adjusted P < 0.05, **adjusted P < 0.01.

Species and Relative Contents of Meibum Lipids Were Differentially Expressed Among the Three Groups

Components and relative contents of meibum lipids were compared among three groups and then, further pairwise post hoc comparisons were tested to compare the difference of lipid relative contents in each group. There were 120 species of lipids that had significant differences in pairwise comparisons: 96 species from coGVHD_(WT) versus non-coGVHD, 78 species from coGVHD_(WT) versus coGVHD_(T), and three species from non-coGVHD versus coGVHD_(T). The majority of differential lipids (52 species) were the same in comparison of “coGVHD_(WT) versus non-coGVHD” and “coGVHD_(WT) versus coGVHD_(T)” at the same time. Table 2 showed the components and relative contents of similarly altered lipids between “coGVHD_(WT) versus non-coGVHD” and “coGVHD_(WT) versus coGVHD_(T).” 

Table 2.

 

The Comparisons of Relative Lipid Contents Among Three Groups

 

Compared to non-coGVHD, coGVHD_(WT) had a significant decrease in relative contents of glycerolipids (TG and DG), and sterol lipids (campesterol and stigmasterol), and a significant increase in relative contents of sphingolipids (ceramide and sphingosine), glycerophospholipids (phosphatidylcholine, phosphatidylethanolamine, methylphosphocholine, and bis-methyl phosphatidic acid) and wax esters (all adjusted P < 0.05). Compared to coGVHD_(T), coGVHD_(WT) had similar changes as those above (all adjusted P < 0.05). Only three differential lipids were found in the comparison between coGVHD_(T) and non-coGVHD, which were involved in triglycerides and stigmasterol (all adjusted P < 0.05). 

Unsaturated TG Was Decreased in CoGVHD Patients Without Treatment

Compared with non-coGVHD and coGVHD_(T) respectively, the coGVHD_(WT) group had significantly decreased relative contents of TG and DG (adjusted P < 0.05). Figure 2B showed the relative contents of representative TG among the three groups. Differential DG, DG (35:0e), DG (37:3e), and DG (39:0e) were significantly decreased, unsaturated TG was significantly decreased (TG [18:3_10:4_11:3] excepted) whereas saturated TG was significantly increased in coGVHD_(WT) group than other two groups (all adjusted P < 0.05). 

Glycerophospholipids and Sphingolipids Were Increased in CoGVHD Patients Without Treatment

Most glycerophospholipids and sphingolipids were elevated in the coGVHD_(WT) group than in the other two groups. All Bis-methyl phosphatidic acid, phosphatidylcholine (PC), and phosphatidylethanolamine (PE) were significantly increased in the coGVHD_(WT) group than in the other two groups (all adjusted P < 0.05). Representative differential lipid species such as Bis-methyl phosphatidic acid (14:0), PC (36:2e) and PE (18:0p_20:4) were shown in Figure 3B. Ceramide (Cer) (t18:0_16:0) and sphingosine (d16:1) were significantly increased in the coGVHD_(WT) group than in the other two groups (all adjusted P < 0.05). Only Cer(d18:1_25:0) was significantly decreased in the coGVHD_(WT) group. 

Correlation Analysis Between Differential Lipid Relative Content and coGVHD Severity

Pearson correlation analysis was performed for identifying the differential lipids associated with coGVHD severity and the results were revealed in Table 3. ICCGVHD score was negatively correlated with mono-unsaturated TG (r = −0.492, P = 0.002) and polyunsaturated TG (r = −0.417, P = 0.009), but was positively correlated with PC (r = 0.345, P = 0.034) and PE (r = 0.364, P = 0.025). Further analysis showed the correlation between ocular symptoms and signs and differential lipids: ocular surface disease index score was correlated with all differential lipids (all P < 0.05); bulbar conjunctival injection was positively related to methylphosphocholine (r = 0.329, P = 0.044) but was negatively related to campesterol ester (r = −0.338, P = 0.038) and stigmasterol ester (r = −0.356, P = 0.028); Schirmer I test was positively linked to campesterol ester (r = 0.323, P = 0.048); corneal fluorescence staining was positively associated with PC (r = 0.366, P = 0.024) and PE (r = 0.391, P = 0.015) but was negatively associated with mono-unsaturated TG (r = −0.377, P = 0.020) and polyunsaturated TG (r = −0.332, P = 0.042). However, noninvasive tear film break-up time was not correlated with any of these differential lipids (all P > 0.05). 

Table 3.

 

Association Analysis of Differential Lipids and coGVHD Severity

 

Correlation Analysis Between Lipid Relative Content and Systemic Intensity of Immunosuppression

After allo-HSCT, transplant recipients are routinely treated with immunosuppression to prevent rejection of the graft. Exposure to immunosuppressive agents and steroids after allo-HSCT is thought to correlate with sebaceous hyperplasia.^31–33 Meibomian gland is the biggest sebaceous gland in the body, and its morphological changes were observed in some patients who were receiving high-dose immunosuppressive drugs after allo-HSCT.³² Considering the systemic immunosuppressants’ potential effects on meibomian glands, we furtherly conducted the univariable and multivariable linear regression analysis to investigate whether the intensity of immunosuppression is associated with coGVHD severity-related meibum lipids. ICCGVHD score and intensity of immunosuppression were included for regression analysis. Univariable linear regression analysis showed that both ICCGVHD score and intensity of immunosuppression were related to mono-unsaturated TG, Polyunsaturated TG, and PC (all P < 0.05) (Supplementary Table S1). Based on the scatter plots in Figure 4, the results of the multivariable linear regression analysis indicate that the intensity of immunosuppression was negatively related to monounsaturated TG (β = −162.4; 95% CI, −268.6 to −56.2; P = 0.004), whereas it was positively related to PC (β = 332.0; 95% CI, 19.2-644.8; P = 0.038). 

Figure 4.

 

The scatter plots figure of the multivariable linear regression analysis of intensity of immunosuppression, ICCGVHD score and mono-unsaturated TG, poly-unsaturated TG, and PC between coGVHD_(WT) and coGVHD_(T) groups.

Discussion

A comprehensive metabolic signature of the meibum for coGVHD remains elusive, although conformational changes in the meibum of patients who had allo-HSCT have been reported in several studies.^9,10,34 Increasing evidences suggested that abnormal lipid metabolism may play a role in cGVHD development and progression. Some specific abnormal plasma metabolites (e.g., 2-aminobutyric acid, 1-monopalmitin, diacylglycerols [DG 38:5, DG 38:6], and fatty acid [20:1]) demonstrated high sensitivity and specificity toward predicting cGVHD. Markey's study revealed that cGVHD onset might be associated with gastrointestinal microbial dysbiosis via controlling the systemic concentrations of microbe-derived short-chain fatty acids.³⁵ As for coGVHD, tear lipid dysregulation in glycerophospholipid metabolism, sphingolipid metabolism, and biosynthesis of unsaturated fatty acids were considered as promising tear biomarkers for coGVHD.³⁶ Considering meibum is the main source of tear film lipids and can reflect the pathology of meibomian glands directly³⁷; therefore we analyzed the lipid profiling of meibum instead of tear in the present study. We first provide evidence that coGVHD was associated with the altered lipidomics profiling of meibum, indicating these meibum disorders and coGVHD may interact with each other, resulting in exacerbation of ocular surface homeostasis. 

TG, one of the most important components of meibum lipids, serves as a major energy reservoir.³⁸ Our study showed that among all altered meibum lipids, TG had the most dramatic changes between patients with coGVHD and those without. Compared with non-coGVHD patients, coGVHD patients had a significant reduction in total TG, and a similar change was also observed when compared with the coGVHD_(T) group. These results indicated that these alterations of TG may be relatively specific expressed in coGVHD patients but could be reversed by ocular anti-inflammatory agents. The microbial dysbiosis of the ocular surface and coGVHD-related inflammation may explain this phenomenon.^39,40 Our previous study found that microbial dysbiosis at the ocular surface is associated with the severity of coGVHD.⁴¹ Bacterial lipolytic enzymes could hydrolyze lipids from the host cell to release free fatty acids which are used as an energy source and to modulate host immune responses.⁴² Microbial disturbance and hyperproliferation of invasive bacteria on the ocular surface could induce the overproduction of bacterial lipase, resulting in TG degradation. The plausible impact of ocular surface dysbiosis on meibum metabolism warrants further study. Besides, inflammatory cells infiltration and prominent fibrosis in meibomian glands in patients with coGVHD was another remarkable characteristic.²³ The coGVHD-related inflammatory cascade induces massive meibomian gland cell apoptosis, which may cause a reduction in lipid synthesis and TG content.⁴³ 

Our findings indicated that certain unsaturated TG were reduced in patients with coGVHD and were negatively related to coGVHD severity. Possible explanations include: first, an abnormal increase in meibum saturation could lead to stiffness and higher melting points of meibum lipids, causing meibomian gland obstruction. Mice fed a high-fat diet had hypertrophic meibomian glands and increased levels of lipid species acylated by saturated fatty acids in meibum as well.⁴⁴ Second, the increased saturation may result in tear film instability and poor spreading at the ocular surface.^44,45 Lipid saturation abnormalities were related to signs or symptoms in numerous diseases such as MGD, blepharitis, diabetes mellitus (DM)–related dry eye, etc.^44,46,47 The decreased levels of unsaturated TG seen in DM-related dry eye disease are similar to our findings,⁴⁷ suggesting that this change in TG levels may be a common characteristic induced by chronic ocular surface inflammation, rather than being specific to either DM-related dry eye or coGVHD. 

Another notable alteration in this study was altered sphingolipids and glycerophospholipids in the meibum of patients with coGVHD. Sphingolipids and glycerophospholipids are essential components of the biological membranes, which are modulated by healthy T cells.⁴⁸ In our study, glycerophospholipids significantly increased in the coGVHD_(WT) group than in those without coGVHD or who received anti-inflammatory therapy. Because GVHD could induce basal cell apoptosis of meibomian glands, it is hypothesized that these abnormally increased glycerophospholipids, especially PC and PE, which were positively related to coGVHD severity, might be caused by coGVHD attack, raising a large number of meibomian gland cells apoptosis, and cell membrane disintegration. However, coGVHD-involved epithelial cells damage might be repaired after anti-inflammatory therapy and the contents of glycerophospholipids returned to normal accordingly. Recently, Ma et al.³⁶ suggested that coGVHD-induced abnormal immune response might be associated with dysregulated glycerophospholipid and sphingolipid networks, which can promote immune cell proliferation and inflammatory cytokine production. Consistently, our results based on the meibum sample revealed that pathogenic glycerophospholipids and sphingolipids of tear film might derive from meibum lipids. 

There are some limitations as follows: first, this is a cross-sectional study with a relatively small sample size, making it difficult to establish causality. The self-controlled study is hard to avoid interference of time, systemic infection, systemic immune alterations, and systemic therapy. Therefore we compared the meibum lipid composition among patients with and without coGVHD after carefully matching for age, sex, allo-HSCT condition, systemic therapy, and MGD severity in this study. Second, the detailed information of chemotherapy regimen, including the doses, route, time of administration, and number of chemotherapy treatments, were unavailable in most of patients because in China the records were basically archived in HSCT center but not with the patients. Thus we cannot determine the effect of conditioning regimen in detail. However, a known confounding factor, Myeloablative conditioning regimen, were matched among the three groups in our study. Third, because of the limited amount of meibum, it is impracticable to conduct quantification of meibum lipid contents. Instead, peak intensity on LC/MS as standardized measures was used to evaluate the contents of lipids. 

Taken together, this study investigated the meibum lipid composition in patients undergoing allo-HSCT by using lipidomics and particular attention was focused on those discrepant lipids associated with the coGVHD. Our findings suggested that altered meibum lipid composition in patients with coGVHD is characterized by a decrease of glycerolipids and an increase of glycerophospholipids, and topical anti-inflammatory therapy may at least in part reverse these changes. Decrease of unsaturated TG and increase of PC and PE were all related to more severe coGVHD and corneal fluorescein staining. In addition, the intensity of immunosuppression might interfere with TG or PC. Further experimental validation and mechanism investigation are needed to determine the mechanism underlying such changes. 

Acknowledgments

The authors thank all individuals involved for their contribution to participant recruitment, sample and data collection, and data management for this study. 

Supported by grants from the National Natural Science Foundation of China (grant numbers 82070922), the Science and Technology Program of Guangzhou (grant numbers 202201020544), and the High-level Hospital Construction Project of China (grant number 303020101). The sponsors or funding organizations had no role in the design or conduct of this research; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. 

Disclosure: W. Zhao, None; J. Yang, None; Y. Liao, None; B. Yang, None; S. Lin, None; R. Liu, None; L. Liang, None