Single‐cell transcriptomics in human skin research: available technologies, technical considerations and disease applications

Abstract Single‐cell technologies have revolutionized research in the last decade, including for skin biology. Single‐cell RNA sequencing has emerged as a powerful tool allowing the dissection of human disease pathophysiology at unprecedented resolution by assessing cell‐to‐cell variation, facilitating identification of rare cell populations and elucidating cellular heterogeneity. In dermatology, this technology has been widely applied to inflammatory skin disorders, fibrotic skin diseases, wound healing complications and cutaneous neoplasms. Here, we discuss the available technologies and technical considerations of single‐cell RNA sequencing and describe its applications to a broad spectrum of dermatological diseases.

of an entire tissue or sample. 2 Since the first single-cell RNA-seq (scRNA-seq) study conducted by Tang et al. the development of single-cell transcriptomics led to the revelation of gene expression variability amongst identical or distinct cell types. 3 Importantly, scRNA-seq allowed researchers to uncover de novo cell populations and lineages that may be implicated in differentiation and development. 4 Moreover, scRNA-seq can be used as a diagnostic tool in precision medicine by deciphering disease-related biomarkers and pathways. 5 This approach can be very promising in drug development, in which cell-specific transcriptomic responses to treatment can be analysed. 6 The aim of this review is to summarize the available technologies and the technical considerations of single-cell transcriptomics, as well as provide a comprehensive compendium of scRNA-seq applications in dermatological research.

| AVAIL AB LE TECHNOLOG IE S
Currently, available technologies of scRNA-seq library generation comprise of cell lysis, reverse transcription into first-strand cDNA, second-strand synthesis and cDNA amplification. 7 These technologies can be broadly classified into two categories: full-length transcriptome sequencing and 3' or 5' end counting-based sequencing. 8 Smart-seq is a full-length transcription sequencing method based on a switching mechanism, in which nucleotides at the end of the RNA template are added, allowing the reverse transcriptase to synthesize the complementary cDNA strand. 9 Smart-seq2, which is the most widely used smart-seq technology, provides higher sensitivity and efficiency in capturing RNA molecules. 10 On the contrary, a popular 3' or 5' end counting-based approach is the droplet microfluidics technology known as Drop-seq, 11 which encapsulates cells into independent microdroplets with unique barcoded beads. Each bead has a cellular barcode which is unique to each droplet, as well as unique molecular identifiers (UMIs) representing each RNA molecule. 12,13 This protocol has been developed further and commercialized with the use of the Chromium instrument (10× Genomics), which is widely used in dermatological research projects. 14 The 10× technology has also developed protocols which conduct immune cell mapping and profiling of specific developing leukocytes by screening the VDJ leukocyte-specific genes of T-cell immune receptors (TCR). 15 In another microfluidic technology, characterized as Seqwell, cells are incorporated into capture beads that are confined in subnanoliter wells and sealed with a semipermeable membrane. In the membrane, the beads are removed, followed by cell lysis and mRNA capture. 16 The first commercially available Seq-Well microfluidic platform was the C1 by Fluidigm. 17 Hughes et al. developed a new version of Seq-well, Seq-Well S, 3 in which a randomly primed second-strand synthesis as a second oligonucleotide handle is established after reverse transcription. 18 This method was applied to certain dermatological diseases and is suggested to be simpler, more compatible with fragile cells and able to manage more samples in parallel. 18 Each of the methods above has its own benefits and drawbacks. Although Smart-seq provides a higher coverage amongst transcripts and alternatively spliced mRNA, Drop-seq enables more cells to be sequenced simultaneously, whereas Seq-well prevents cross-contamination between samples. 19 A novel scRNA-seq technique is Smart-seq3, which incorporates full-length coverage and a 5'UMI tagging strategy. 20 This protocol allows a dramatic increase of sensitivity and can estimate gene expression in a larger number of cells. 20

| TECHNIC AL CONS IDER ATIONS
The scRNA-seq experimental procedure consists of four main steps: sample preparation, cell enrichment, library preparation and data analysis ( Figure 1). 7 In regard to the isolation of single cells from skin samples, punch biopsies or larger specimens are obtained and dissociated via mechanical or enzymatic treatment. Because the various layers of the skin have different cell compositions and properties, single-cell dissociation can be quite challenging. 21 Multiple singlecell dissociation approaches are available depending on whether the dermis, epidermis or both are needed for each experiment. For example, a research study that aimed to resolve the basal keratinocyte transition states isolated the epidermal tissue, 22 while in a study aiming at profiling fibroblast subpopulations, the epidermis was discarded, and the dermis was processed. 23 Typically, whole-skin dissociation kits such as the gentleMACS (Miltenyi Biotec) dissociation system can be used. 23,24 Nevertheless, some dissociation enzymes included in the kit favour the isolation of fibroblasts and might provoke the elimination of a subpopulation of immune cells or epidermal cells, abrogating the whole-cell map of the sample. Alternatively, one can incubate the whole skin with dispase, leading to dermal-epidermal dissociation. 25 Keratinocytes, melanocytes and epidermal immune cells are further obtained by the application of trypsin and fibroblasts by digestion with collagenase. 26 Mechanical means are also used, such as mashing, dicing or slicing, enhancing the whole process. 7,26 Tissue dissociation can vary from 2 h to overnight. 26 During incubation, the activity of each enzyme must be taken into consideration since longer incubations can negatively impact cell viability and induce mechanical stress or trigger immune activation to the cells. 21 An alternative approach for skin cell acquisition is the application of suction blistering, whereby an artificial blister is introduced to the skin, eliciting dermal-epidermal dissociation and the formation of a fluid skin sample combined with interstitial fluid. 27 The liquid nature of the biopsy can enhance single-cell dissociation. 28 A major limitation of this method is the inability to capture deeper dermal regions, and therefore, dermal, endothelial cell and macrophage (Mφ) information cannot be provided. 27,28 Another parameter to be considered is the possibility of samples being frozen prior to processing, allowing sample processing at different time points. However, reduction of cell numbers, and the alleviation of the cell transcriptome, are two major considerations.
Several cryoprotectant agents used in tissue preservation have facilitated post-thaw capture of an adequate number of cells, mainly fibroblasts. 29 Another proposed method of preserving the sample is by freezing the single-cell solution after dissociation, which has shown satisfactory cell viability and integrity. 30 However, working with fresh tissue when needed is preferred, and multiple freeze and thaw cycles should be avoided. 21 An unbiased view of the cellular composition of the sample and the projection of its cellular heterogeneity is revealed when one captures all desired cells. Therefore, quality check and specific cell isolation are performed in the cell enrichment step. Obtaining a pure single-cell solution without cell debris, fragmented, or dead and apoptotic cells can reduce artifacts during downstream analysis. 31 Quality check can be done manually by using an inverted microscope and micropipettes, or by applying negative charge in a patch pipette, a method known as micromanipulation. 7 However, these methods are low-throughput and very laborious. High throughput methods such as fluorescence-activated cell sorting (FACS) or magnetic-activated cell sorting (MACS) are predominantly used. 31 FACS has an improved sensitivity even when a cell expresses low levels of the specific labelled markers with an antibody or with a viability dye such as DAPI. 32 A limitation of this method is the relatively large number of cells needed as a starting material and the added stress exerted on the cells during the sorting process. 33 On the contrary, MACS uses enzymes, antibodies and peptides conjugated to magnetic beads for sorting, requiring less time and equipment, but lacks sensitivity and specificity. 31,34 Sorting methods can also be used to isolate specific cell populations of interest, which are further analysed as a distinct sample. For example, a study outlining the vascular endothelial cell heterogeneity in human skin has used both FACS and MACS to enrich for endothelial cells, 35 whereas another study focused on the dendritic cell signalling in psoriasis by using FACS for dendritic and Mφ cell isolation. 36

| Healthy skin
Different studies have dissected the heterogeneity of dermal fibroblasts with scRNA-seq (Table 1)     Four major fibroblast populations were described and categorized as secretory-reticular, secretory-papillary, pro-inflammatory and mesenchymal according to their anatomical location and predicted functional role, but these identities were blurred in aged cells. Notably, aged fibroblasts also displayed a decreased number of interactions with other cell types. 45

| Inflammatory skin disorders
Atopic dermatitis (AD) has been one of the most extensively studied cutaneous diseases with scRNA-seq. Harirchian and colleagues analysed epidermal cells to demonstrate that IL-17A induced targets of A20, an NFκB inhibitor and contributor to different skin rashes, share a similar overexpression in keratinocytes not only from AD, but also from psoriasis and erythrokeratodermia variabilis. 46 This highlights the role of keratinocytes in inflammatory skin disorders and suggests A20 skin upregulation as potential treatment. 46  Finally, in a scRNA-seq study characterizing Langerhans cells, two steady-state (LC1 and LC2) and two activated subsets were revealed. LC2, which were more likely to be activated, bore similarities to monocytes, expressed immunosuppressive genes and were more abundant in psoriatic lesions. 53 The Lafyatis laboratory used scRNA-seq to gain insights into vasculopathy of systemic sclerosis (SSc) by characterizing cutaneous endothelial cells and revealed genes APLNR and HSPG2, which are mediators of Apelin/Elabela-APLNR and TGFβ signalling and could potentially serve as biomarkers of pathogenesis. 54 The authors further explored the disease by analysing myofibroblast populations, which are the driver cell type of fibrosis, the most prominent manifestation of SSc on the skin. 55  In axillary lesions from hidradenitis suppurativa (HS) patients, a chronic inflammatory follicular occlusion condition, monocytes and Mφ exhibited similar transcriptomic profiles to diabetic foot ulcer cells. 58 They also overexpressed a series of markers associated with Fc signalling, metabolic activity, type I and II interferon stimulation and were more polarized towards the M1 phenotype. 58  The cellular composition and molecular drivers in cutaneous lupus erythematosus lesional and non-lesional skin were reported by Billi et al. 63 Normal appearing skin in lupus patients was revealed as a highly type I interferon enriched environment that universally affects the gene expression of all skin cell types, while in lesional skin, accumulated CD16+ DCs arose as potent disease contributors. 63 In other studies, focused on lupus nephritis, skin scRNA-seq was leveraged to determine whether skin biopsies could be utilized as renal disease biomarkers. 64 IFN-inducible genes, including IFI6, STAT1 and IFITM1, were indeed upregulated in keratinocytes of patients with lupus nephritis indicating a systemic response to IFN. 64 Expanding on their previous report, the authors processed more samples and included paired renal and skin biopsies from the same individuals to confirm augmented expression of type I interferon response pathway genes in lupus patients. 65 They also stratified patients as responders and non-responders to treatment and found that non-responders' tubular epithelial cells and keratinocytes overexpressed fibrosis-associated extracellular matrix genes. 65 To better understand the initiation and progression of vitiligo, proliferating and resistance to apoptosis gene expression signature was described. 78 The first study to harness the technology of scRNA-seq in human skin samples examined metastatic melanoma from 19 tumors with diverse clinical and therapeutic backgrounds. 79 Tirosh et al. Alexandros Onoufriadis https://orcid.org/0000-0001-5026-0431