Abstract Our skin influences our physical and mental health, and its chemical composition can reflect environmental and disease conditions. Therefore, through sampling the skin metabolome, we can provide a promising window into the mechanisms of the body. However, the broad application of skin metabolomics has recently been hampered by a lack of easy and widely applicable sampling methods. Here, we present a novel rapid, simple, and, most importantly, painless and non-invasive sampling technique suitable for clinical studies of fragile or weakened skin. The method is called WET PREP and is simply a lavage of the skin which focuses on capturing the metabolome. We systematically evaluate WET PREPs in comparison with the non-invasive method of choice in skin metabolomics, swab collection, using ultra-performance liquid chromatography coupled to mass spectrometry (UPLC-MS^2) on two complementary chromatographic columns (C18 reversed phase and hydrophilic interaction chromatography). We also integrate targeted analyses of key metabolites of skin relevance. Overall, WET PREP provides a strikingly more stable shared metabolome across sampled individuals, while also being able to capture unique individual metabolites with a high consistency in intra-individual reproducibility. With the exception of (phospho-)lipidomic studies, we recommend WET PREPs as the preferred skin metabolome sampling technique due to the quick preparation time, low cost, and gentleness for the patient. Keywords: skin, sampling method, LC-MS^2, metabolome, non-invasive sampling, skin disease, inter-organ crosstalk 1. Introduction The skin is our largest organ and has a tremendous impact on both our physical and mental health [[34]1]. Through descriptions of its chemical composition, such as the metabolome of the skin, we can begin to understand mechanisms in skin physiology, and eventually reach a point at which we can use important endogenous metabolites to modulate the skin phenotype or use the skin metabolome as a diagnostic tool for systemic diseases. The metabolome defines the entirety of low molecular weight compounds, which includes secreted, exogenous, and endogenous substances. In particular, within the skin, numerous low molecular weight metabolites can be excreted by specialized glands, which promise specific novel biomarkers that are of interest for disease and cosmetic applications. The metabolome of the skin thereby reflects dynamic changes that occur in relation to the macro and microenvironment, intrinsic factors, and also in relation to skin and systemic disease [[35]2,[36]3,[37]4,[38]5]. In this context, aging is one intrinsic factor that influences the skin metabolome. When comparing older versus younger skin, not only did coenzyme Q10, an anti-aging metabolite, levels change, but also 56 other metabolites differed, including dehydroepiandrosterone-(DHEA-) sulfate and several amino acids [[39]3]. The physical macroenvironment can be reflected by the metabolome as well. UV light exposure, for example, upregulates lactate, which is the predominant physiological form of lactic acid, a natural moisturizing factor [[40]2]. Along with the abiotic environmental factors, the microbiome influences the chemical profile of the skin, e.g., through hydrolysis of chemicals secreted by the host or through secretions of secondary metabolites against pathogenic strains such as Staphylococcus aureus [[41]5,[42]6]. These pathogenic strains have a role in the development of skin disease, and consequently a diseased skin can be identified by changes in the metabolome, as reviewed in Yan et al., 2017 [[43]7]. Metabolites can furthermore indicate the progression of wound healing in the skin, where linolenic acid is initially increased followed by an increase in adenosine [[44]8]. In summary, taken the recent literature, the skin metabolome gives insights into the physiology of wound healing, topical diseases and their link to systemic comorbidities, and cosmetic aspects and aging related topics. However, due to the lack of convenient sampling methods, the analysis of skin metabolome has not yet become widespread. A few studies have focused on the metabolome and key metabolites in special applications, including psoriasis, UV-B and nutrition effects, acne, and aging [[45]3,[46]9,[47]10]. They have highlighted the importance of metabolomics to gain deeper knowledge of the pathophysiology of interest. Following the rapid development in this perspective, an increasing demand for studies identifying and cataloging the metabolite composition of the skin will require easy and reproducible sampling technologies and standardization to allow for increased repeatability and better cross-sectional studies. Currently, there are numerous methods for sampling involving varying levels of invasiveness. For long-term skin secretion studies, hydrogel micropatches are preferred [[48]11,[49]12]. They are made of a network of hydrophilic polymers that “swell” in relation to the amount of sweat absorbed, and have a minimum sampling time of 10 min [[50]11]. For studies focused instead on epidermal compositions, biopsies, suction blisters, tape stripping, and pre-wet swabs have been implemented [[51]2,[52]3,[53]5,[54]13]. Among the less invasive methods, tape stripping and swab analysis are the most widely used. However, tape stripping is improbable for use in skin disease studies [[55]13], as it is painful for skin-sensitive groups and unfeasible for sampling lesions because it removes an already fragile barrier. In consequence, and to ensure patients’ convenience, the collection of skin swabs is the most favorable method. Unfortunately, swab preparation is time intensive and significantly complicates clinical assays [[56]12]. Therefore, here, we introduce an easy and fast sampling method of the skin metabolome and skin secretome, which avoids skin irradiation and is applicable on lesions because it is pain free. We evaluate the collected metabolome systematically and with a focus on recently highlighted key metabolites of skin relevance. 2. Results The skin metabolome was obtained from 22 healthy individuals at the left and right antecubital fossa by swab and WET PREP sampling. Metabolites were analyzed using UPLC-MS^2, using two complementary separation modes, hydrophilic interaction chromatography (HILIC) and C18 reversed phase (RP), as these are the two most used separation modes in skin metabolomics. Metabolites were assigned to known molecular formulas listed in KEGG (Kyoto Encyclopedia of Genes and Genomes), HMDB (Human Metabolome Database), and LMSD (LIPID MAPS Structure Database) within the absolute error range of 0.005 Da ([57]Table S2). For both sampling methods, we were interested in the detected skin metabolome in general, as obtained in most individuals (shared skin metabolome), as well as the suitability to obtain inter-individual differences. Furthermore, we analyzed the specific coverage of metabolites that have previously been shown to be important in skin related publications. First, we focused on the general coverage of metabolome. In HILIC mode, both sampling methods delivered a comparable number of detected metabolites, whereas in RP, swabs outperformed WET PREPs by 920 annotated compounds ([58]Figure 1A). A closer look at the annotation of unique compounds in swabs shows that 55% of them do at least have one annotation as lipid, and between the two columns, RP detects overall 11% more lipids than HILIC ([59]Figure 1B). Apart from this higher lipid measurement in RP, we observed of both separation modes a high percentage (66% in HILIC, 54% in RP) of commonalities in the composition of the metabolome (including the lipidome) in swabs and WET PREPs, as shown in the Venn diagram ([60]Figure 1B). This is further supported by unsupervised statistics, such as principal component (PCA), where the separation of both methods is explained by a small amount of variance in the first two components ([61]Figure 1C). Nevertheless, both PCA and cluster analysis illustrated a distinct separation between the sampling methods, indicating unique detection and quantities of metabolites in each method ([62]Figure 1C and [63]Figure 2). Figure 1. [64]Figure 1 [65]Open in a new tab Metabolome overview of skin samples according to sampling method, WET PREP (light blue) and swab (red) for HILIC and RP columns: Total number of annotated metabolites (A); Venn diagrams of the total number of compounds according to sampling method with percentage of lipids-compounds had to be present in at least 2 samples per sampling method to be considered (B); PCA score plot (C); Percentage of shared metabolome across samples with HILIC (lighter colors) and RP (darker colors) (D). Figure 2. [66]Figure 2 [67]Open in a new tab Dendrograms of hierarchical cluster analysis of swab and WET PREP for: RP (A) and HILIC (B). WET PREP samples noted in blue lines and swabs in red lines. Sample annotation is according to sampling method: WET PREP (W) and swab (S), individual (number following “W” or “S”), lateral side right (r) and left (l), and city of residence, location a (green) and location b (black). The shared metabolome, which are metabolites detected stably in most individuals, differed according to method, as observed of the distribution plot of detected metabolites across the samples ([68]Figure 1D). For example, in HILIC mode, 40 of the 44 samples shared 20% of the metabolites found in WET PREPs, while the shared metabolome for swabs was only 10% ([69]Figure 1D). This is substantiated by the PCA ([70]Figure 1C) as a tight clustering of samples illustrates a strong shared metabolome. Additionally, inter-individual variations are reflected in the PCA score plot, and are easily visualized by hierarchical clustering ([71]Figure 2). In consensus with Bouslimani et al., there was a stable metabolome between intra-individual replicates, with only three exceptions for swab samples ([72]Figure 2) [[73]5]. The hierarchical clustering shows a distinct separation according to sampling method for RP and HILIC. Within the swab and WET PREP clusters, two further subclusters are seen that can be explained by including the metadata. To begin with the alignment of the individuals’ metadata, participants resided in one of two urban towns that were 80 km apart, with location B being more densely populated and industrial. Despite the proximity of the locations, the samples clustered based on location for RP swabs, WET PREPs, and HILIC swabs, but in HILIC WET PREPs, location appears as secondary in importance ([74]Figure 2). Apart from a short difference in storage time (less than one week, −80 °C), the sample preparation and analysis were done in a mixed randomized order, and the same person collected the samples in both locations; therefore, extrinsic factors, such as pollution, industry, and population size may be responsible for the clustering. Interestingly, in HILIC WET PREP, the first cluster within the same sampling method was not with respect to location of the sampling, but rather in accordance with peculiarities. For example, participant three did not follow the instructions and showered within 24 h prior, and grouped with other individuals who had showered or had additional peculiarities, such as pregnancy and history of skin disorder ([75]Figure 2). This clustering based on those three peculiarities was also seen in the RP WET PREP samples, but less consistent for the swabs, in both modes RP and HILIC ([76]Figure 2). For both RP and HILIC, regardless of sampling method, individuals who exercised prior, such as riding a bike or hiking, clustered with the other individual who had no peculiarity, suggesting exercise has a low influence on the skin metabolome ([77]Figure 2). In regard to the biological sex of the participants, there was no overarching grouping that differed to what is seen in sweat metabolomics [[78]14]. Despite the outperformance of WET PREPs in the detection of a stable shared metabolome over swabs, individual differences were also successfully represented in the WET PREP data. Overall, for both sampling methods, there is a stable shared metabolome, but there are also metabolic inter-individual differences, and based on metabolome similarity, bi-lateral replicates are true replicates. Before we discuss skin relevant metabolites, we would like to discuss differences of the sampling method in regard to metabolic pathways. For this purpose, we performed a Wilcoxon–Mann–Whitney test between swabs and WET PREPs. All p-values were Benjamini–Hochberg corrected. Significant metabolites (p < 0.05) were then uploaded into MetaboAnalyst [[79]15,[80]16], and all analysis was performed with the focus of impact on the downstream nodes, i.e., out of degree centrality. For the HILIC data, the most impacted and significantly different pathways were aminoacyl-tRNA biosynthesis, glycine serine and threonine metabolism, and valine, leucine and isoleucine biosynthesis. Aminoacyl-tRNA biosynthesis was better represented by WET PREP, while glycine serine and threonine metabolism were better represented in swabs. ([81]Figure 3B). For the RP data, the most impacted and different pathways were aminoacyl-tRNA biosynthesis, valine, leucine and isoleucine biosynthesis, and phenylalanine metabolism, where all three pathways were better represented by WET PREP ([82]Figure 3A). Overall, for both HILIC and RP, we see a general difference in pathways involved with amino acid metabolism, where sampling method mostly affects which amino acids are isolated ([83]Figure 3). This leads us to conclude that sampling method does influence the pathway coverage of the skin metabolome. Figure 3. [84]Figure 3 [85]Open in a new tab Pathway analysis chart of significantly different metabolites between WET PREP and swab: for RP (A) and HILIC (B). Metabolites that are significantly different between WET PREP and swab across each column type and with KEGG ID’s, as determined by MassTRIX annotation, were run against the human KEGG database by MetaboAnalyst. The top 20 pathways are shown according p-value and impact factor for out of degree centrality shown. Finally, through cross-reference to previous literature, we focus on specific metabolites that have potential relevance to skin function and health ([86]Table 1). These metabolites covered amino acids, their derivatives, acid, sugar, nucleo(t/s)ides, and aromatics. 19 and 18 of these metabolites were detected in HILIC and RP, respectively, with the majority of these metabolites focused on amino acids and derivatives. Out of these, 63% were detected with a higher intensity in WET PREP as compared to swabs, with glutamine, glutamic acid, and ornithine surpassing 0.3 log fold change ([87]Table 1). In regard to sugar and aromatics, caffeine and glucose were significantly higher detected in WET PREP samples, and fucose was better detected by swab. In addition, O-cresol was detected and was significantly higher in WET PREP samples surpassing one log fold change. Finally, the acid retinoic acid was best detected in swabs. In sum, this indicates an advantage of WET PREP to swabs for sampling these skin relevant metabolites, with an outperformance of detection in WET PREPs in comparison to swabs for 75% of compounds in RP and 60% in HILIC. Table 1. Metabolites of skin relevance. Compounds of interest were compared for their detection in swabs and WET PREP samples. Exact m/z values (0.005 Da) and retention time were used for identification. All compound identification was verified with analytical standards. Log2 fold change WET PREP/swab of the average intensities across all samples are shown. (n.d.) denotes that the compound is not detected. Metabolites are grouped according to chemical type with their references as seen on the