Integrated omics analysis of sweat reveals an aberrant amino acid metabolism pathway in Vogt–Koyanagi–Harada disease

Summary

Vogt–Koyanagi–Harada (VKH) disease is an autoimmune disease leading to visual impairment. Its pathogenic mechanisms remain poorly understood. Our purpose was to investigate the distinctive protein and metabolic profiles of sweat in patients with VKH disease. In the present study, proteomics and metabolomics analysis was performed on 60 sweat samples (30 VKH patients and 30 normal controls) using liquid chromatography tandem mass spectrometry. Parallel reaction monitoring (PRM) analysis was used to validate the results of our omics analysis. In total, we were able to detect 716 proteins and 175 metabolites. Among them, 116 proteins (99 decreased and 17 increased) were observed to be significantly different in VKH patients when compared to controls. Twenty‐one differentially expressed metabolites were identified in VKH patients, of which 18 included choline, L‐tryptophan, betaine and L‐serine were reduced, while the rest were increased. Our multi‐omics strategy reveals an important role for the amino acid metabolic pathway in the pathogenesis of VKH disease. Significant differences in proteins and metabolites were identified in the sweat of VKH patients and, to some extent, an aberrant amino acid metabolism pathway may be a pathogenic factor in the pathogenesis of VKH disease.

Sweat, as an emerging biofluid with the potential to discover immune biomarkers, was used to understand the pathogenic mechanisms of Vogt‐Koyanagi‐Harada (VKH) disease. In our study, we utilized proteomic and metabolomics analysis to investigate the distinctive composition of sweat in VKH patients. Significant differences in proteins and metabolites were identified in the sweat of VKH patients, and an aberrant amino acid metabolism pathway may be a pathogenic factor in the pathogenesis of VKH disease.

Introduction

VKH disease is a systemic autoimmune disease directed at melanocyte‐associated antigens and is characterized by a recurrent bilateral granulomatous panuveitis, often coupled with systemic complications such as vitiligo, poliosis, alopecia and auditory abnormalities [1, 2]. Although VKH disease is a serious sight‐threatening disease, we still have a limited understanding of its pathogenesis. Previous studies have demonstrated that a viral infection, as the triggering factor, may drive this disease by inducing a humoral autoimmune response [3]. Moreover, many genes, including human leukocyte antigen (HLA)‐DR4, HLA‐DRB1, interleukin (IL)‐23R and adenosine (ADO) were shown to be associated with VKH disease [4]. Similarly, several studies have indicated that various cytokines secreted during T cell‐induced immune responses were likely to be involved in the pathogenesis of this disease [5, 6]. Most of these studies use blood, cerebrospinal fluid or intraocular fluid samples to study disease mechanisms and as yet no research has been reported with other body fluids. Recent studies have shown that an increasing number of anti‐bacterial components and immune factors were found in sweat, indicating that sweat may reflect immune system activity [7, 8]. However, little is known about the changes of sweat composition in autoimmune disease and was therefore the purpose of the study reported here.

Sweat is a colorless, transparent and hypotonic liquid secreted by the sweat glands, which is crucial for the maintenance of skin homeostasis. As a medium connected to the skin barrier, sweat is irreplaceable in immune function, skin moisture, temperature regulation and biological defense [9, 10]. Not only antibody isotypes such as immunoglobulin (lg)A, lgE and lgG, but also cytokines including tumor necrosis factor (TNF)‐α, transforming growth factor (TGF)‐β and other interleukins have been found in sweat [11]. As a widely accepted biomarker, sweat chloride measurement is of great value in diagnosing and treating cystic fibrosis (CF) [12, 13]. Recently, studies on non‐invasive sweat glucose monitoring, which could replace traditional and traumatic blood glucose monitoring, has been introduced to monitor the blood glucose status of diabetics [14]. Although sweat is a promising biological fluid, its role in biomarker analysis has been limited due to the relatively low abundance of analytes and technical limitations of detection.

In the current study, we used proteomic and metabolomic analysis to investigate the diversity of proteins and metabolites in patients with VKH disease and found significant differences in their profile when compared to healthy controls.

Materials and methods

Ethical statement

Informed consent was obtained from each volunteer participating in our experiments and approval of this project was received from the Ethics Committee of Chongqing Medical University.

Participants

Between June 2018 and January 2019, we enrolled 30 untreated VKH patients and 30 normal controls in our hospital for this study. A definite VKH diagnosis was made in accordance with the requirements of the international nomenclature committee [15]. Participants were excluded if they suffered from hypertension, diabetes, heart disease or infectious disease. In addition, the healthy controls that were recruited for this project did not have any inflammatory disease. The baseline characteristics of participants, including age, gender, body mass index (BMI), cigarette, alcohol, coffee and tea consumption, were collected at the same time.

Sample collection and preparation

Sweat of the participants was obtained, according to an earlier publication [7], at a constant temperature of 70℃. Volunteers were asked not to use any cosmetics or other skin‐contact chemicals for 24 h. Clean water was used to bathe the volunteers, and they needed to stay in the sauna for 20 min. After a break of 10 min, the volunteers were instructed to take another shower and stay in the sauna for another 20 min. At this time, sweat from the armpit was collected and transferred to a centrifuge tube using a pipette. The sweat was transported on ice to the laboratory as soon as possible. The sweat was subsequently centrifuged to remove impurities such as cells and keratin and the supernatant was collected for further analysis. Approximately 600 µl of sweat was collected per individual.

Proteomics

Approximately 20 μg sweat proteins were boiled for 5 min after adding ×5 loading buffer and proteins were separated on a 12·5% sodium dodecyl sulphate–polymerase chain reaction (SDS‐PAGE), with a parameter setting for constant current at 14 mA for 90 min. Protein bands were stained by Coomassie blue staining. Afterwards, the eluted proteins were digested into peptides in accordance with the filter‐aided sample preparation (FASP digestion) method. Thirty μl SDS with DTT (SDT) buffer, containing 4% SDS, 100 mM dithiothreitol (DTT) and 150 mM Tris‐HCl at pH 8.0, was added to 200 μg of the sweat sample. Repeated ultrafiltration (microcon units, 10 kD) was performed after adding UA buffer, a mixture of 8 M urea and 150 mM Tris‐HCl at pH 8.0; 100 μl iodoacetamide [100 mM iodoacetamide (IAA) in UA buffer] was added to the samples and incubated in the dark for 30 min. The filter was then washed five times, three times with 100 μl of UA buffer and twice with 100 μl of 25 mM NH₄HCO₃ buffer. The protein suspensions were digested with 4 μg trypsin (Promega, Shanghai, China) at a constant temperature of 37°C overnight. A C18 cartridge [Empore^TM SPE cartridge C18 (standard density), bed ID 7 mm, volume 3 ml; Sigma, St Louis, MO, USA] was used to desalinate the acquired peptides. Peptides were lyophilized and then taken up in 0·1% formic acid solution. Peptide content was analyzed by high‐performance liquid chromatography (HPLC) using 0·1% formic acid (buffer A) and buffer B comprising 84% acetonitrile and 0·1% formic acid. A reverse‐phase trap column (Acclaim PepMap100, 100 μm × 2 cm, nanoViper C18; Thermo Scientific, Waltham, MA, USA) with an automatic sampler in buffer A was used and separation of the peptide mixture was completed after the sample had passed through a C18 reverse‐phase analytical column (Easy column, 10 cm long, 75 μm inner diameter, 3 μm resin; Thermo Scientific). IntelliFlow technology was used and the flow rate was maintained at 300 nl/min during the entire process. The linear gradient setting during the next 2 h was as follows: elution was performed from 0 to 55% buffer B for 110 min, followed by 55–100% buffer B and 100% buffer B for 5 min.

A 2‐h liquid chromatography tandem mass spectrometry (LC‐MS/MS) analysis was performed using a Q Exactive mass spectrometer (Thermo Scientific) combined with Easy nLC (Thermo Fisher Scientific). It ran constantly in positive ion mode. The data‐dependent top 10 method was considered as the best method to obtain mass spectrometry (MS) data. For better higher‐energy collisional dissociation (HCD) fragmentation, this method allowed for a dynamic selection of the largest number of precursor ions (survey scan 300–1800 m/z). The automatic gain control (AGC) target was subsequently adjusted to 3e6. The appropriate settings for maximum injection time and dynamic exclusion duration were fixed at 10 ms and 40·0 s, respectively. At m/z 200, survey scans were first acquired at a resolution of 70 000, while the HCD spectra were obtained at a resolution of 17 500. The following settings were also optimized: isolation width (2 m/z), normalized collision energy (30 eV) and the underfill ratio (0·1%). The mass spectrometer parameter information is shown in Supporting information, Table S4. Analysis of the MS data was performed using MaxQuant software version 1.3.0.5 (Max Planck Institute of Biochemistry, Martinsried, Germany) and label‐free quantitation (LFQ) was applied for further quantitative analysis. Proteins identified in greater than or equal to two of the three biological replicates were included for subsequent expression analysis.

PRM analysis

Skyline software (MacCoss Laboratory, University of Washington, Seattle, WA, USA) was adopted to optimize the analysis of target peptides on target proteins. Three product ions with high abundance and good continuity were chosen in the selected peptide for quantitative analysis. Data for the peak area for each peptide were imported after Skyline analysis to obtain the relative expression information for each peptide. For each significantly altered protein, the signal intensities for a single peptide sequence was accurately quantified through further calculation.

Metabolomics

Ultra‐HPLC (UHPLC) analysis of each sample was performed using 1290 Infinity LC equipment (Agilent Technologies, Santa Clara, CA, USA). The hydrophilic interaction liquid chromatography (HILIC) separation procedure was performed on a 2·1 mm × 100 mm ACQUIY UPLC BEH 1·7 µm column (Waters, Dublin, Ireland). The column was kept at 25°C, with a flow rate of 0·3 ml/min. The mobile phase A consisted of ammonium acetate (25 mM) and ammonium hydroxide (25 mM) in water, while the mobile phase B contained only acetonitrile. The gradient was maintained at 95% B in the first minute and linearly decreased to 65% B in the next 13 min. From 14 to 16 min, it gradually dropped to 40% B and remained at this level for the following 2 min. During 18–18·1 min, the gradient was immediately increased from 40% B to 95% B and remained as such to min 23. The sweat samples were placed in a 4℃ autosampler during the entire analysis.

Analysis of sweat samples was then performed on a quadrupole time‐of‐flight mass spectrometer (TOF‐MS) (AB Sciex TripleTOF 5600, Waltham, MA, USA), employing electrospray ionization‐positive and ‐negative ion modes, respectively. Parameters of the Ion Source Gas1 and Gas2 were optimized at 60 psi. The curtain gas (CUR) was set at 30 psi, source temperature was kept constant at 600℃ and IonSpray Voltage Floating (ISVF) was fixed at ± 5500 V. When the TOF‐MS scan was operated, the facility parameter was specified to the m/z range 60–1000 Da with the accumulation time at 0·20 s/spectra. However, when comparing product ion scans, the setting was changed to an m/z range of 25–1000 Da and its accumulation time was adjusted to 0·05 s/spectra. In a high‐sensitivity mode, information‐dependent acquisition (IDA) was chosen to obtain the MS/MS data. The declustering potential (DP) has positive and negative modes: ± 60 V at 35 ± 15 eV applied the collision energy (CE). The optimal parameters of IDA were as follows: exclude isotopes within 4 Da, six candidate ions to monitor per cycle. Next, the raw data were imported with XCMS software after being converted to an mzXML file by ProteoWizard. The subsequent processes included peak picking, peak grouping and ion features extraction. Metabolites were identified by comparing the accuracy m/z value [< 25 parts per million (p.p.m.)] and MS/MS spectra.

Statistical and bioinformatics analysis

Student’s t‐test and one‐way analysis of variance (ANOVA) were used to analyze the proteomic data. Proteins with a fold‐change (FC) > 2 or < 0·5 and having a P‐value < 0·05 were deemed to be significantly different. Cluster version 3.0 and Java Treeview software were applied for hierarchical clustering analysis. A protein–protein interaction (PPI) network map was then constructed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://string-db.org/). Based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (https://www.kegg.jp/), a GO functional enrichment analysis and pathway enrichment analysis were performed with OmicsBean (http://www.omicsbean.cn/), a website for comprehensive analysis of omics data.

After preprocessing the metabolic data by Pareto scaling, the next step included statistical single‐ and multi‐dimensional analyses. Student’s t‐test and fold change analysis were part of the single‐dimensional statistical analysis, while partial least‐squares discrimination analysis (PLS‐DA) was used for multi‐dimensional statistical analysis. A metabolite was considered to be statistically different when the variable importance in the projection (VIP) value was > 1 and the P‐value < 0·05. All the identified metabolites were displayed as volcano plots drawn in r language (https://www.r-project.org/) to show significant differences between patient and control groups. The metabolites categories were obtained by the HMDB database (http://www.hmdb.ca/).

Results

Sixty participants were recruited for this project, 30 with VKH disease and 30 normal controls. The baseline characteristics show that there were no statistically significant differences concerning age, gender, BMI or lifestyle (cigarette, alcohol, coffee and tea consumption) between the two groups (Table 1).

Table 1.

Characteristics of recruited VKH patients and normal controls

Characteristic	VKH	Controls	Statistical value	P‐value
Patients (n)	30	30
Male (n, %)	19 (63·3%)	17 (56·7%)	χ2 = 0·278	0·598
Age (years)
Mean ± SD	41·83 ± 10·68	36·43 ± 12·78	t = 1·776	0·081
BMI
Mean ± SD	23·95 ± 3·32	23·29 ± 2·85	t = 0·819	0·416
Cigarette (n,%)	8 (26·7%)	8 (26·7%)	χ2 = 0·000	1
Alcohol (n, %)	4 (13·3%)	4 (13·3%)	χ2 = 0·480	0·488
Coffee (n, %)	1 (3·3%)	1 (3·3%)	χ2 = 0·000	1
Tea (n, %)	6 (20%)	8 (26·7%)	χ2 = 0·373	0·542

To show the features of gender, cigarette, alcohol, coffee and tea, χ² tests were applied.

The characteristics of age and body mass index (BMI) were analyzed using Student’s t‐test.

VKH = Vogt–Koyanagi–Harada.

An overall description of proteomics data

Using label‐free quantitative proteomics techniques, we identified 716 proteins in six biological replicates of sweat samples tested. Strict statistical criteria (FC ≥ 2·0 or ≤ 0·5, P‐value ≤ 0·05) were applied to detect differentially abundant proteins between the groups. A total of 116 proteins were differentially expressed, 17 of which were increased in VKH patients, while 99 proteins were decreased (Supporting information, Tables S1 and S2). Of the differentially expressed proteins, earlier reports have mentioned an involvement of macrophage migration inhibitory factor (MIF), complement factor B (CFB) and complement factor H (CFH) to be involved in VKH disease [16, 17, 18].

Functional analysis of differentially expressed sweat proteins

The hierarchical clustering method was used to cluster the differential proteins and was displayed as a heat‐map, whereby it is clear that the VKH group can be readily distinguished from the controls (Fig. 1). To further characterize the nature of these 116 proteins, we adopted PPI maps using STRING (Fig. 2) and analyzed their interaction. The results showed that several proteins located in the center of the network, such as glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH), 60 kDa heat shock protein, mitochondrial (HSPD1) and fibrinogen gamma chain (FGG), were related to the immune system [19, 20, 21].

Fig. 1.

Heat‐map was performed using 44 proteins with a large difference to reveal distinct expressive pattern. There were three pooled biological replicates in the Vogt–Koyanagi–Harada (VKH) group and the control group, respectively. High‐ and low‐intensity proteins are represented by red and purple.

Fig. 2.

A protein–protein interaction (PPI) network was constructed to visually show the degree of connectivity between differentially expressed proteins. Nodes in red are the up‐regulated proteins in Vogt–Koyanagi–Harada (VKH) patients, whereas gray nodes are down‐regulated proteins.

GO functional enrichment analysis was used to further analyze the functional pathways of the differential proteins (Supporting information, Fig. S1). The results identified 2391 biological process (BP) terms, 267 cellular component (CC) terms and 233 molecular function (MF) terms (P‐value < 0·05). The BP analysis showed that many were related to the following immune system pathways: immune effector process, leukocyte‐mediated immunity, positive regulation of immune response and innate immune response. CC analysis revealed that these proteins largely participated in the so‐called extracellular region, organelle and vesicle formation. Terms were identified in the MF analysis, including primarily associated binding, protein binding, identical binding and receptor binding.

Pathway enrichment analysis was performed to further analyze the biological functions of the differentially expressed proteins. As shown in Fig. 3, protein processing in endoplasmic reticulum, carbon metabolism, Hippo signaling pathway, pyruvate metabolism and pathogenic Escherichia coli infection were the pathways with the largest number of enriched proteins.

Fig. 3.

Pathway enrichment analysis based on differentially expressed proteins. The main categories and subcategories of each pathway are shown in the same color.

Validation of differential proteins

To confirm the reliability of our data, further PRM assays were performed. Four differentially abundant proteins with high connectivity in the PPI network were validated, including GAPDH, malate dehydrogenase, mitochondrial (MDH2), hemopexin (HPX) and lactotransferrin (LTF). As shown in Table 2, comparing label‐free results and PRM results, it was obvious that their expression trends were consistent.

Table 2.

Comparison of quantification results between VKH and normal controls

Accession	Name	Label‐free results	PRM results
		VKH/controls	VKH/controls
P04406	GAPDH	0·32 (down)	0·31 (down)
P40926	MDH2	0·22 (down)	0·26 (down)
P02790	HPX	0·14 (down)	0·37 (down)
P02788	LTF	0·34 (down)	0·09 (down)

VKH = Vogt–Koyanagi–Harada; GAPDH = glyceraldehyde‐3‐phosphate dehydrogenase; MDH2 = malate dehydrogenase, mitochondrial; HPX = hemopexin; LTF = lactotransferrin.

Metabolomic analysis of sweat samples in VKH‐disease

Untargeted metabolomics strategy was used for metabolite profiling of sweat samples and resulted in the detection of 2388 so‐called features in the positive ion mode and 1796 features in the negative ion mode. Using available standards, metabolomics data obtained were matched using an in‐house database. Using this method, we were able to identify 175 metabolites. On the basis of a database from HMDB (http://www.hmdb.ca/), we were able to classify these 175 metabolites into 27 categories, 71% of which were carboxylic acids and derivatives, fatty acyls, organo‐oxygen compounds and benzene and substituted derivatives.

When the PLS‐DA score was analyzed, we observed marked differences between the VKH and control group (Supporting information, Fig. S2). Using a VIP > 1 and P‐value < 0·05, we identified a differential expression of 21 metabolites between these two comparison groups (17 in positive ion mode; seven in negative ion mode; L‐pyroglutamic acid, L‐citrulline and urocanic acid were recognized in both ion modes). Among these 21 differentially expressed metabolites, three (4‐methoxycinnamic acid, Val‐His and palmitic acid) were increased and the rest were decreased (Supporting information, Table S3). The Volcano Plot (Supporting information, Fig. S3), a univariate analysis of integrated fold change analysis and t‐test, was used to assess the significance of the differentially expressed metabolites. Correlation analysis showed that 15 differentially expressed metabolites displayed a positive correlation in the positive ion mode (Fig. 4).

Fig. 4.

Correlation analysis of 17 significantly differentially expressed metabolites [variable importance in the projection (VIP) > 1, P‐value < 0·05) were performed in positive ion mode. The deeper the color, the closer the correlation coefficient is to 1, indicating a positive correlation between the two metabolites.

Integrated analysis of dysregulated proteins and metabolites

We integrated the entire range of differential proteins and metabolites using the KEGG database to obtain further insight into the functional pathways that were possibly associated with VKH. Nine pathways were found to be involved in both omics groups (Fig. 5). Most of the identified pathways belonged to amino acid metabolism, including glycine, serine and threonine metabolism, cysteine and methionine metabolism, phenylalanine metabolism and tyrosine metabolism. Decreasing the glycine, serine and threonine metabolism pathways showed the strongest difference when comparing VKH with controls. Several molecules, such as D‐3‐phosphoglycerate dehydrogenase (PHGDH), choline, L‐tryptophan, betaine and L‐serine, were significantly altered. Three proteins and one metabolite, including MDH2, L‐lactate dehydrogenase B chain (LDHB), malate dehydrogenase, cytoplasmic (MDH1) and L‐serine, were relevant to the cysteine and methionine metabolism pathway. The imbalance of MIF and L‐phenylalanine was recognized in the phenylalanine metabolism pathway. In addition, both MIF and tyramine were altered in the tyrosine metabolism pathway.

Fig. 5.

A total of nine biological pathways were found to have shared proteins (orange) and metabolites (blue) according to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. A metabolic pathway was shown by a column, arranged from left to right according to the number of proteins or metabolites involved. A tall column demonstrated that a large number of molecules are annotated in the biological pathway.

Discussion

In the present study, we show that VKH patients have a different sweat protein and metabolite profile. A total of 175 metabolites and 716 proteins could be identified in sweat, using state‐of‐the‐art metabolomics and proteomics technology. Comparison of sweat samples obtained from VKH patients and controls revealed that 116 proteins and 21 metabolites were differentially expressed, and the reliability of these omics data was further validated by PRM. Taking advantage of bioinformatics analysis, our results indicated that amino acid metabolism pathways, shared by the differentially expressed proteins and metabolites, were significantly altered in the VKH group. Although measurement of the sweat chloride concentration (sweat test) is the gold standard for the diagnosis of an immune disorder such as cystic fibrosis (CF) [22] and has also been proposed to monitor glucose in diabetic subjects [23], no other reports have been published so far, to our knowledge, concerning sweat composition in autoimmune disease. Earlier studies have reported the presence of proinflammatory molecules such as cytokines in sweat and showed a good correlation with plasma levels, suggesting that the analysis of sweat might form an alternative method to assess these factors in ambulatory settings where blood collection is difficult [24, 25, 26]. Factors detected in sweat included calcitonin gene‐related peptide, IL‐6, IL‐8, IL‐1α, IL‐1β, neuropeptide Y, substance P, TNF‐α and vasoactive intestinal peptide. Although sweat may reflect the systemic cytokine profile of an individual following leakage from the blood compartment, it should be noted that local cytokine production may also contribute to the levels observed [27].

Instead of analyzing individual cytokines in sweat we decided to use a multi‐omics approach, which addresses more basic molecular pathways. T cells from patients with autoimmune disease have been shown to have a distinct metabolic microenvironment associated with proliferation and proinflammatory activities, emphasizing the role of such basic pathways in the control of the immune response [28]. Several groups have now reported on the use of omics strategy to investigate the mechanisms operative in a number of autoimmune diseases, such as idiopathic inflammatory myopathies [29], rheumatoid arthritis [28] and systemic lupus erythematosus [30, 31].

In our proteomics data, a close connection was found between many proteins and the metabolism pathway through pathway enrichment analysis (Fig. 3). Among 17 different metabolites in positive ion mode, 11 with positively correlated expression patterns belonged to the amino acid metabolic pathway (Fig. 4). Our further multi‐omics strategy also revealed an important role for this pathway in the pathogenesis of VKH disease (Fig. 5). Our findings are in agreement with recent studies, showing that very basic molecular pathways such as those involved in amino acid metabolism may play an important role in the control of the (auto)immune response [32, 33]. In our study, we observed that an altered PHGDH, choline, L‐tryptophan, betaine and L‐serine participate in glycine, serine and threonine metabolism and that this pathway was associated with VKH disease. PHGDH is a key enzyme that catalyzes 3‐phosphoglycerate to synthesize L‐serine [34]. The low content of L‐serine results in activation of the stress response and cell death. Both choline and betaine, as the precursor substances of trimethylamine N‐oxide (TMAO), are two essential and nutritional biomolecules. Several reports demonstrated that the deficiency of choline and betaine could accelerate the production of reactive oxygen species (ROS) and are thus considered as a key factor in initiating and maintaining inflammatory diseases caused by oxidative damage [35, 36]. Recent studies indicated that the absence of the essential amino acid L‐tryptophan could affect ROS scavenging and immune response control in vitiligo, which is generally regarded as an autoimmune disease and which is also a prominent feature in VKH disease [5, 37]. In fact, numerous studies have suggested that tissue damage during autoimmune disease may be exacerbated by the excess production of ROS [38, 39]. Consistent with these findings, our data revealed that the level of PHGDH, choline, L‐tryptophan, betaine and L‐serine in sweat from the VKH group was lower when compared with healthy controls, indicating that the damage to glycine, serine and threonine metabolism may be linked to the pathogenesis of VKH disease.

The study has several limitations. First, we only included 30 VKH patients and 30 normal controls. Therefore, we were not able to assess the role of gender or particular clinical subtypes. Additional multi‐omics investigations of patient groups with various disease manifestations are required to further elucidate the role of amino acid metabolism pathways in its pathogenesis. Longitudinal studies are also needed to investigate whether disease activity or treatment affects the sweat metabolite or protein profile. Secondly, the use of sauna to collect sweat is a known method to collect sweat [7], and was acceptable to the patients. However, with a more targeted approach it might be easier to obtain sweat samples following pilocarpine stimulation. Futher study is needed to compare the effect of these different sweat collection methods on its composition. Thirdly, our study is a preliminary observation in an emerging field, and further studies are also needed to investigate if sweat analysis could be an alternative to blood sampling in the search for disease biomarkers. A further limitation is that we only analyzed sweat from the armpit. Sweat from different areas should be collected and compared, to improve the physiological interpretation of our data.

In summary, this study showed that the molecular profile of sweat in patients with VKH disease differs from controls and emphasizes a role for certain amino acid metabolism pathways in the pathogenesis of VKH disease.

Disclosures

The authors declare no conflicts of interest.

Author contributions

X. C., S. Y. and P. Y. conceived and designed the experiment; X. C., Q. C. and C. Z. collected samples; X. C. performed the experiments, X. C. and L. Z. wrote the initial manuscript; G. S., A. K. and P. Y. reviewed the data and edited the manuscript.

Supporting information

Fig. S1. Go enrichment analysis of differential expressed proteins associated with biological process (BP), cell component (CC) and molecular function (MF) was performed.

Fig. S2. Partial Least Squares Discrimination Analysis (PLS‐DA) of 60 sweat samples acquired from VKH patients and normal controls.

Fig. S3. The metabolomics data of positive ion mode (a) and negative ion mode (b) were analyzed by volcano plot. The red dots represents the metabolites with FC > 1·5 and P‐value < 0·05.

Table S1. Proteins with significantly higher or lower aboundance in VKH disease

Table S2. Proteins with consistent presence/absence LFQ expression patterns in VKH disease

Table S3. Differential metabolites between the VKH disease group and the normal control group

Table S4. Parameter details obtained from the mass spectrometer