Abstract

   Urinary metabolomics is a useful non-invasive tool for large-scale
   screening of disease-related metabolites. However, no comprehensive
   urinary metabolomic analysis of vitiligo is presently available. To
   investigate the urine metabolic pattern of vitiligo patients, we
   conducted a combined cross-sectional and prospective self-control
   cohort study and an untargeted urinary metabolomic analysis. In the
   cross-sectional study, 295 vitiligo patients and 192 age‐ and
   sex‐matched controls were enrolled, and 71 differential metabolites
   between two groups were identified. Pathway enrichment analysis
   revealed that drug metabolism-cytochrome P450, biopterin metabolism,
   vitamin B9 (folate) metabolism, selenoamino acid metabolism, and
   methionine and cysteine metabolism showed significant enrichment in
   vitiligo patients compared with the status in healthy controls. In the
   self-control cohort, 46 active vitiligo patients were recruited to
   analyse the urinary metabolic signatures after treatment. All of these
   patients were asked to undertake follow-up visits every 2 months three
   times after first consulting and the disease stage was evaluated
   compared with that at the last visit. Folate metabolism, linoleate
   metabolism, leukotriene metabolism, alkaloid biosynthesis, and tyrosine
   metabolism were predicted to be involved in vitiligo activity. Our
   study is the first attempt to reveal urinary metabolic signatures of
   vitiligo patients and provides new insights into the metabolic
   mechanisms of vitiligo.

   Subject terms: Immunology, Biomarkers, Diseases

Introduction

   Vitiligo is a common acquired pigmentary disorder characterised by
   depigmentation of skin resulting from the destruction of epidermal
   melanocytes. The incidence of vitiligo has been estimated to be 1% of
   the global population^[38]1. It has major impacts on patients’ social
   activity and mental health, causing severe distress. Unlike other
   cutaneous disorders, no erythema or scaling is present in vitiligo
   lesions. In some atypical or early-stage cases of vitiligo, it is quite
   difficult to diagnose and differentiate from other hypopigmented
   diseases^[39]2,[40]3. This may cause a delay in treatment at an early
   stage and some patients with other diseases can even be misdiagnosed
   with vitiligo^[41]4. Moreover, the course of vitiligo is unpredictable
   and it is difficult to assess the treatment response at an early
   stage^[42]5–[43]7, which means that treatment evaluation is often
   postponed. Therefore, a biomarker helping physicians to objectively
   recognise the atypical lesions, follow patients over time, or
   accurately determine the treatment response at an early stage would be
   of great value^[44]6,[45]8,[46]9.

   Many groups have attempted to find vitiligo biomarkers. Clinical signs
   such as Koebner phenomenon, blurred border, and confetti-like
   depigmentation have been described as clinical markers of active
   vitiligo, but these signs only present in a subset of vitiligo patients
   and are not sufficiently objective^[47]10,[48]11. Skin tissue
   biomarkers have been reported, such as basal cell vacuolisation^[49]12,
   CD8^+ lymphocyte infiltration^[50]13, and increased expression of heat
   shock protein-70^[51]14, CXCL9^[52]15, and sCD25^[53]16. However, skin
   biopsy is a traumatic examination and it is difficult to apply to
   patients with more than one active period. Moreover, an overlap in
   histological findings has been found between active and stable
   vitiligo^[54]9.

   Blood biomarkers of vitiligo, including soluble CDs (sCD25,
   sCD27)^[55]6, chemokines (CXCL9, CXCL10)^[56]17, S100B^[57]11,
   cytokines (IL-1β, IL-10, IL-17)^[58]18,[59]19, and homocysteine^[60]20,
   have been reported to be related to the occurrence and activity of
   vitiligo. However, these markers still need further clinical
   verification and different studies on them have even shown contrasting
   results^[61]9. Thus, there is a need for more study to find
   non-invasive biomarkers that could help to diagnose and monitor the
   stage of vitiligo.

   Metabolomics is a widely used biological approach for identifying and
   measuring the changes in biological samples^[62]21,[63]22. Urine
   examinations are considered as non-invasive, rapid diagnostic methods
   that have been used in scientific research and clinical applications.
   Recent studies have shown that urine metabolomics has become a useful
   method to identify biomarkers for some skin diseases, such as
   psoriasis^[64]23, dermatomyositis^[65]24, melanoma^[66]25,
   syphilis^[67]26, and atopic dermatitis^[68]27. However, to date, few
   studies focusing on changes in urine metabolites in vitiligo patients
   have been performed. Previous metabolic studies mostly focused on a few
   metabolites in vitiligo, such as urinary catecholamines and
   vanillylmandelic acid on a small scale^[69]28,[70]29. However, no
   comprehensive urinary metabolomic analysis of vitiligo is currently
   available.

   In this study we conducted a urine metabolomic analysis in a
   cross-sectional and prospective self-control cohort, and attempted to
   identify the metabolic biomarkers for both the diagnosis and the
   treatment response of vitiligo patients. We first analysed the urine
   samples from 295 Chinese vitiligo sufferers who volunteered to
   participate and 192 healthy controls, along with investigating the
   metabolic features and biomarkers of the vitiligo patient group. Then,
   in the self-control cohort, we targeted the active patients and
   investigated the variation in urine metabolites after treatment. The
   analysis contributes to our ability to diagnose this disease, revealing
   metabolic changes involved in disease activity, and helping to identify
   the urinary metabolic patterns in different effective stages and
   potential biomarkers for treatment responses (Fig. [71]1). Our study is
   the first attempt to reveal urinary metabolic signatures of vitiligo
   patients. These results might be helpful to explore the metabolic
   changes involved in the pathogenesis of vitiligo, and to diagnose this
   disease and monitor its treatment response in a clinical setting.

Figure 1.

   [72]Figure 1
   [73]Open in a new tab

   The work flow of this study.

Results

Clinical characteristics

Discovery cohort and validation cohort

   Urine samples (midstream) were randomly divided into a discovery cohort
   of 211 vitiligo patients and 113 age- and sex-matched healthy human
   adults, and a validation cohort of 84 vitiligo patients and 79 healthy
   human adults. Among the 295 vitiligo patients, 19 segmental vitiligo
   patients were enrolled. Their detailed demographics and disease
   subtypes are shown in Tables [74]1 and [75]S1.

Table 1.

   Demographics of healthy subjects and vitiligo patients enrolled in this
   study.
   DHC
   (n = 113) DVC
   (n = 211) p* VHC
   (n = 79) VVC
   (n = 84) p^# SAV (n = 46)
   Average age (years) 25.18 ± 15.61 23.03 ± 13.53 0.26 24.38 ± 14.16
   23.21 ± 13.28 0.97 32.7 ± 12.45
   Sex 0.48 0.64
   Female 55 94 48 48 19
   Male 58 117 31 36 27
   [76]Open in a new tab

   DHC: Discovery cohort healthy control. DVC: Discovery cohort vitiligo
   cases. VHC: Validation cohort healthy control. VVC: Validation cohort
   vitiligo cases. SAV: Self-control active vitiligo cohort.

   *p-value of Chi-square test comparing DHC with DVC.

   ^#p-value of Chi-square test comparing VHC with VVC.

Self-control cohort

   Forty-six active vitiligo patients were recruited at Peking Union
   Medical College Hospital. Among these patients, 44 patients undertook
   the first follow-up visit (2 months after the first consultation), 44
   patients undertook the second follow-up visit (4 months after the first
   consultation), and 43 patients undertook the third follow-up visit
   (6 months after the first consultation; for more details, see Tables
   [77]1 and [78]S1). By comparing the digital follow-up photographs, wood
   lamp images, and clinical examination results obtained at the last
   visit, 104 effective/improved visit points were recorded. Then, we
   performed metabolic profiling between the different follow-up visits
   and the baseline among the patients in whom treatment was effective to
   further investigate whether metabolic profiles could reflect
   improvement of vitiligo and detect the urine metabolites with a
   tendency to change in association with this.

Urine metabolomic pattern of vitiligo patients compared with healthy controls

Metabolites differentially expressed between vitiligo patients and healthy
controls were identified

   To eliminate the potential confounders between the vitiligo and healthy
   groups, subjects were gender- and age-matched. LC–MS-based urine
   samples from vitiligo patients and healthy controls yielded 2500
   features after quality control (QC) filtering. Tight clustering of the
   QC samples indicated good repeatability of analysis (Fig. [79]S1). The
   principal component analysis (PCA) score plot did not show an obvious
   trend of separation between the vitiligo patients and the healthy
   controls (Fig. [80]2a). However, the orthogonal partial least squares
   analysis (OPLS-DA) model achieved better separation (Fig. [81]2b).
   Permutation tests were carried out to confirm the stability and
   robustness of the supervised models presented in this study (Fig.
   [82]S2a). Differential metabolites were assigned based on VIP
   values > 1 and adjusted p-values < 0.05. In total, 71 differentially
   expressed metabolites were identified, with 61 metabolites upregulated
   and 10 downregulated in the vitiligo group compared with the levels in
   the healthy controls (Table [83]S2). Further metabolic comparison
   between the 19 segmental vitiligo samples and 19 age- and
   gender-matched nonsegmental vitiligo samples showed that the two
   disease types had no significant differences of urine metabolomics
   (Fig. [84]S2c,d).

Figure 2.

   [85]Figure 2
   [86]Open in a new tab

   Analysis of urine metabolome between vitiligo patients and healthy
   controls. (a) PCA analysis of urine metabolome. (SIMCA 14.0 software,
   Umetrics, Sweden) (b) Score plot of OPLS-DA model between vitiligo
   patients and healthy controls showed a better separation. (SIMCA 14.0
   software, Umetrics, Sweden) (c) Pathway enrichment analysis showed
   significant enrichment (p < 0.05) of several pathways between the two
   groups. (MetaboAnalyst 3.0, [87]https://www.metaboanalyst.ca) (d) A
   metabolites panel consisting of 7alpha-hydroxy-3-oxochol-4-en-24-oic
   acid, deoxyuridine, 3,4-octadienoylglycine, threoninyl-Proline showed
   the predictive ability with an AUC of 0.807 in the validation cohort.
   (MetaboAnalyst 3.0, [88]https://www.metaboanalyst.ca).

   Pathway enrichment analysis showed significant enrichment
   (p-values < 0.05) of several pathways in the vitiligo group compared
   with the case in healthy controls, including drug metabolism-cytochrome
   P450, biopterin metabolism, vitamin B9 (folate) metabolism, selenoamino
   acid metabolism, and methionine and cysteine metabolism (Fig. [89]2c,
   Table [90]2).

Table 2.

   Differential metabolic pathways that may involve in vitiligo
   pathogenesis.
   Pathways                           Possible involved function References