Assessment of Treatment-Relevant Immune Biomarkers in Psoriasis and Atopic Dermatitis: Toward Personalized Medicine in Dermatology

Abstract

Immunologically targeted therapies have revolutionized the treatment of inflammatory dermatoses, including atopic dermatitis and psoriasis. Although immunologic biomarkers hold great promise for personalized classification of skin disease and tailored therapy selection, there are no approved or widely used approaches for this in dermatology. This review summarizes the translational immunologic approaches to measuring treatment-relevant biomarkers in inflammatory skin conditions. Tape strip profiling, microneedle-based biomarker patches, molecular profiling from epidermal curettage, RNA in situ hybridization tissue staining, and single-cell RNA sequencing have been described. We discuss the advantages and limitations of each and open questions for the future of personalized medicine in inflammatory skin disease.

INTRODUCTION

Atopic dermatitis (AD) and psoriasis are two common chronic inflammatory dermatoses. Immune dysregulation drives AD and psoriasis pathogenesis (Di Cesare et al., 2009; Guttman-Yassky et al., 2011a, 2011b; Leung, 1999). Early studies in psoriasis showed an expansion of CD8+ T cells (Chang et al., 1994) capable of producing type 1 cytokines such as IFNγ and TNFα (Austin et al., 1999). IL-17 and IL-23 were subsequently found to be increased in psoriatic lesions and are now thought to represent a pathologic hallmark of this disease (Lee et al., 2004; Lowes et al., 2008; Piskin et al., 2006; Teunissen et al., 1998). These cytokines and other downstream proinflammatory signals mediate the characteristic changes of this disease (Chan et al., 2006; Liang et al., 2006; Teunissen et al., 1998).

AD lesional skin shows increased levels of IL-4 and IL-13, which are produced primarily by skin-infiltrating T helper (Th) 2 cells (Akdis et al., 1997; Renz et al., 1992; van der Heijden et al., 1991; van Reijsen et al., 1992). These type 2 cytokines are thought to be central to AD pathogenesis. Th22 cells may also be increased, resulting in excess IL-22 production (Gittler et al., 2012). IL-31, another type 2 cytokine, has also been implicated in AD pathogensis, particularly pruritus (Cheung et al., 2010; Dillon et al., 2004; Takaoka et al., 2006).

Despite the characteristic patterns of inflammation in most cases of AD and psoriasis, the Th1/Th17 versus Th2 paradigm may not apply neatly in all cases. There is evidence for intradisease heterogeneity within psoriasis and AD (Liu et al., 2022a; Tsoi et al., 2019; Wang et al., 2021a). Some studies have also suggested that molecular overlap may exist in some patients with psoriasis and AD (Moy et al., 2015). For example, some cases of AD may show an unexpected Th17/IL-23 signal (Brunner et al., 2018; Esaki et al., 2016; Koga et al., 2008), and this may be more common in pediatric individuals with AD and Asian individuals (Noda et al., 2015). Th1-driven inflammation may increase with chronicity in AD (Gittler et al., 2012). Th22 elevation has been reported in Black and African American patients with AD, with conflicting data on the role of Th1/17 inflammation in this population (Sanyal et al., 2019; Wongvibulsin et al., 2021). Overlapping Th2 and Th1/17 patterns have also been illustrated in palmoplantar pustulosis (McCluskey et al., 2022).

Biologic therapies have revolutionized treatment for psoriasis and AD. There are currently several cytokine-targeted biologic therapies approved for psoriasis, including TNFα inhibitors, IL-23 (p19) inhibitors, an IL-12/IL-23 (p40) inhibitor, IL-17A inhibitors, an IL-17 receptor (IL-17RA) inhibitor, and an IL-36 receptor (IL-36R) inhibitor. An oral TYK2 inhibitor was also recently approved for psoriasis. There are two currently approved biologics for AD in the United States: dupilumab, an IL-4Ra inhibitor (blocking the activity of both IL-13 and IL-4), and tralokinumab, an IL-13–specific inhibitor. In addition, two oral Jak inhibitors, upadacitinib and abrocitinib, which inhibit signaling downstream of many cytokines, were recently approved for AD. There are also numerous clinical trials for novel AD biologics, including drugs targeting IL-22, IL-31, and TSLP, among others.

Biologic therapies blocking the IL-17/IL-23 axis have shown 75% improvement in PASI-75 in approximately 75% of patients (Armstrong et al., 2020). In AD, IL-4Ra blockade with dupilumab results in a 75% improvement in the Eczema Area and Severity Index (EASI) 75 in approximately 50% of patients (Silverberg et al., 2021). Presently, medication choice is largely based on population-level efficacy data, medical comorbidities, and physician preference but, in most cases, is ultimately relies up trial and error (Aggarwal et al., 2022). Given the growing number of AD and psoriasis biologics aimed at different molecular targets, there is a need for approaches to assist with rational biologic treatment selection.

Biomarkers can be used for many purposes in inflammatory skin disease: to aid in diagnosis, to define clinical subtypes, to predict disease progression or severity, to monitor treatment response, or to inform therapy selection. Biomarkers for predicting disease progression or severity and monitoring therapy response in inflammatory skin disease have been reviewed elsewhere (Corbett et al., 2022; Ramessur et al., 2022; Renert-Yuval et al., 2021). In this paper, we focus on immune biomarkers that aim to inform biologic therapy selection and, in some cases, aid in the diagnosis of clinically/histopathologically indeterminate rashes (CIRs) (Figure 1). The notion of distinct clinical endotypes, which has been proposed in AD (Czarnowicki, 2019), correlates demographic factors such as age and ethnicity with underlying immunology and could be used to inform therapy selection. However, in this study, we focus on methods for selection of the best therapeutic target for individuals based on their unique immunologic profiles. In addition, although genetic polymorphisms; microbiome-based, metabolomic, or lipidomic biomarkers; as well as imaging-based biomarkers have been explored in inflammatory skin disease, we have chosen to focus on RNA- and protein-based immunologic biomarkers, which have the richest body of literature and the most direct relevance to therapy selection in inflammatory skin disease.

Figure 1. RNA- and protein-based biomarkers for biologic therapy selection and diagnosis.

Various methods exist for collecting biomarkers from the skin, including shave or punch biopsy or minimally invasive approaches such as tape stripping, superficial epidermal curettage, or microneedle patches. In addition, systemic biomarkers from the blood can be measured. RNA analysis techniques include RT-PCR, microarray, sequencing approaches, or in situ analysis using hybridization probes. Proteins can be analyzed by IHC/IF, immunoassay, MS-based methods, proximity extension, or aptamer-based methods for high-throughput proteomics. Each method of analysis is labeled with compatible collection methods on the basis of references in this review. The RNA or protein levels of different immunologic molecules or other molecular correlates could be used for rational selection of biologic therapy and to aid in the diagnosis of CIRs. CIR, clinically/histopathologically indeterminate rash; IF, immunofluorescence; IHC, immunohistochemistry; RISH, RNA in situ hybridization; scRNA-seq, single-cell RNA sequencing; MS, mass spectrometry.

SKIN BIOMARKERS

Early studies used microarray profiling of skin biopsies to characterize transcripts enriched in psoriasis (Oestreicher et al., 2001; Suárez-Fariñas et al., 2012; Zhou et al., 2003), AD (Gittler et al., 2012; Guttman-Yassky et al., 2009; Nomura et al., 2003; Rodríguez et al., 2014), and shared gene signatures between psoriasis and AD (Choy et al., 2012). Other studies used RNA sequencing to identify differentially expressed genes in AD (Suárez-Fariñas et al., 2015) and psoriasis (Gudjonsson et al., 2010; Li et al., 2014; Swindell et al., 2013; Tsoi et al., 2015) and define subtypes of psoriasis on the basis of gene expression signatures (Ainali et al., 2012; Swindell et al., 2012). Several studies proposed classifiers to aid in the diagnosis of AD and psoriasis (Guttman-Yassky et al., 2009; Inkeles et al., 2015; Quaranta et al., 2014), including validation with prospective samples (Garzorz-Stark et al., 2016) and classification based on gene signatures from uninvolved skin (Tsoi et al., 2019).

Gene expression profiling of skin biopsies has also been employed to monitor response to treatment in AD (Beck et al., 2014; Guttman-Yassky et al., 2019a; Möbus et al., 2021) and psoriasis (Johnston et al., 2014; Krueger et al., 2019; Sofen et al., 2014). In one small-scale study in psoriasis, baseline levels of IL20, IL21, and p40 mRNA correlated with response to IL-12/23 inhibition (Gedebjerg et al., 2013). However, no large-scale studies have shown the ability of bulk transcriptomic profiling of the skin at baseline to predict response to biologic therapy. In AD, baseline Th2 biomarkers in the skin did not correlate with improvement in EASI or pruritis scores after treatment with IL-4Rα blockade (Beck et al., 2014). In a combined analysis of gene expression data of psoriatic lesions from many studies, authors achieved robust prediction of PASI improvements using gene expression values from 2 to 4 weeks of treatment but poor predictive accuracy using baseline measurements (Rosa da et al., 2017).

Proteomic analysis of skin biopsies has identified biomarkers in psoriasis (Carlén et al., 2005) and AD (Noh et al., 2016; Pavel et al., 2020), including immunohistochemistry (IHC)- and immunofluorescence (IF)-based methods, to classify AD and psoriasis skin with reasonable accuracy (D’Erme et al., 2015; Garzorz-Stark et al., 2016). To our knowledge, no studies have evaluated the ability of proteomic profiling of the skin to predict response to biologic therapy.

BLOOD BIOMARKERS

Many studies have uncovered alterations in circulating immune cells and levels of serum proteins in psoriasis and AD. Serum biomarkers have been useful to predict disease severity and progression and to monitor response to treatment, including in clinical trials (Beck et al., 2014; Guttman-Yassky et al., 2019a; Kim et al., 2018). The use of serum biomarkers for these purposes has been reviewed earlier (Mikhaylov et al., 2021a; Pourani et al., 2022; Sobolev et al., 2022).

Circulating biomarkers could aid in the selection of biologic treatment or the diagnosis of CIRs. High-throughput techniques such as proximity extension assays and aptamer-based methods have been employed to measure serum protein levels in AD and psoriasis (Brunner et al., 2017; Wang et al., 2017), in addition to more traditional immunoassays (Kolbinger et al., 2017). In one study, authors found that enhanced NF-κB signaling in immune cells from the blood of patients with psoriasis before therapy initiation was associated with poor response to TNFα inhibition (Andres-Ejarque et al., 2021). In an analysis of serum biomarkers from Chinese patients with AD before and after dupilumab treatment, one group found that baseline levels of three proteins–CD25/sIL-2Rα, IL-31, and IL-36β–had a modest correlation with response to treatment (Wu et al., 2023). Serum biomarkers have been used to classify adult and pediatric patients with AD into distinct clusters on the basis of their biomarker profiles (Bakker et al., 2022; Thijs et al., 2017), but it remains to be determined whether this classification could be used to predict response to biologic therapy. One recent study showed that dupilumab was equally effective in extrinsic-type (elevated IgE) AD and intrinsic-type (normal IgE) AD (Gelato et al., 2023).

MINIMALLY INVASIVE AND OTHER INNOVATIVE APPROACHES TO MEASURE IMMUNOLOGIC BIOMARKERS IN THE SKIN FOR BIOLOGIC THERAPY SELECTION

Tape strip profiling

Tape stripping is a minimally invasive method to profile the superficial epidermis. Adhesive tape strips are repeatedly applied to a patient’s skin, each time sampling a deeper portion of the stratum corneum (Figure 2). Approximately 20 tape strips are thought to be necessary to reach the granular cell layer of the epidermis (Kim et al., 2019), as each tape strip removes approximately one layer of corneocytes. Most studies to date have used less than 20 tape strips (Hughes et al., 2021). The procedure elicits only mild discomfort and/or erythema.

Figure 2. Skin sampling and processing for different investigational techniques for personalized therapy selection in inflammatory skin disease.

(a) For tape stripping, an applicator is applied to the skin for consecutive tape strips. RNA or proteins are extracted from the tape strips and quantified. (b) For the microneedle-based dermal biomarker patch (Mindera Health, San Diego, CA), a microneedle patch which has probes extending into the superficial dermis is applied to the skin. RNA is extracted, reverse transcribed, and sequenced. (c) For molecular profiling from epidermal curettage (Castle Biosciences, Friendswood, TX), the skin is scraped with a curette, and RNA is isolated from the skin sample and analyzed by RT-PCR. (d) For RISH, a skin biopsy is obtained. A probe complementary to the RNA molecule of interest is added to the tissue slide, and a chromogenic or fluorescent marker is added in the detection step. (e) For single-cell RNA sequencing, a biopsy is obtained, and the skin sample is dissociated into single cells. Single-cell cDNA library preparation is performed, and the cDNA library is sequenced using next-generation sequencing. FFPE, formalin-fixed, paraffin-embedded; RISH, RNA in situ hybridization.

Early studies of tape stripping used various techniques to quantify the mRNA levels of immune-related genes (Benson et al., 2006; Morhenn et al., 1999; Wong et al., 2004). He et al. (2021) performed RNA sequencing on tape strips from patients with AD, patients with psoriasis, and healthy controls. The authors showed that the skin of patients with AD had increased Th2-related transcripts (including IL13, CCL17, and CCL18) and that psoriasis skin had elevated Th17-related (IL17A/F) and innate (IL36A/IL36G) transcripts. They also showed that high levels of nitric oxide synthase 2 (NOS2) transcripts alone could identify psoriasis samples. Transcriptomic profiling of tape strip samples has also been applied to study pediatric AD (Guttman-Yassky et al., 2019b; Pavel et al., 2021) and hand eczema (Sølberg et al., 2022).

Dyjack et al. (2018) used tape stripping to identify heterogeneity within patients with AD, defining a group of type-2-high patients with elevated expression of IL13, IL4R, CCL22, and CCR4 and more severe eczema. Mikhaylov et al. (2021b) studied transcript levels from AD tape strips before and after dupilumab treatment, noting decreases in chemokine genes, including CCL13, CCL17, and CCL18, and increases in barrier-related transcripts after treatment toward the levels of healthy controls. Some gene expression changes correlated with responses to dupilumab treatment for all patients.

Investigators have also studied tape strip samples using proteomic approaches. Inoue et al. (2011) performed ELISA on tape strip samples to show that IL-18 levels are higher in the skin of patients with AD than in the skin of healthy controls. Méhul et al. (2017) used mass spectrometry and multiplex ELISA to define a proteomic signature of psoriatic lesions, and the same group also showed that proteomic profiling of tape strips could differentiate between psoriasis and cutaneous T-cell lymphoma (Méhul et al., 2019). In another study, authors found that the levels of IL-36γ quantified by ELISA from tape strip samples could accurately differentiate between psoriasis and AD (elevated in psoriasis), even in clinically challenging cases (Berekméri et al., 2018). Multiplex immunoassays have been used to quantify a larger number of protein biomarkers from tape strip samples of adult and pediatric patients with AD (Clausen et al., 2020; Hulshof et al., 2019; McAleer et al., 2019), including in response to dupilumab (He et al., 2020a). Mass spectrometry–based analysis of tape strip samples has also been used to compare protein biomarkers in the skin of patients with AD with and without food allergy (Goleva et al., 2020; Leung et al., 2019).

Tape stripping has several advantages for potential applications in personalized molecular profiling. The first is that it is minimally invasive, allowing for a higher number of sites or frequency of sampling and less discomfort for the patient. Second, direct comparisons between tape stripping and biopsy samples have shown that tape stripping may enrich for expression changes within the epidermis (Dyjack et al., 2018; Tsoi et al., 2022).

Limitations of tape stripping include that it predominantly samples the stratum corneum (and possibly the upper granular cell layer). One study found that proteomic profiling of tape strips was less sensitive to detecting cytokine changes in AD skin compared with RT-PCR of biopsy samples (Simonsen et al., 2021). In addition, there are reported difficulties in standardizing tape stripping on the basis of sampling procedure (number of tape strips and amount of pressure applied) and characteristics of the sampled skin, such as stratum corneum thickness and skin hydration (Bashir et al., 2001; Berekméri et al., 2019). Processing samples for RNA sequencing or proteomics requires specialized equipment, specific expertise, and data normalization, making it potentially expensive and limiting clinically relevant turnaround times (Table 1).

Table 1.

Comparison of Investigational Approaches for Personalized Therapy Selection in Inflammatory Skin Disease

	Tape Stripping	Dermal Biomarker Patch (Mindera Health)	Epidermal Curettage (Castle)	RISH Staining	scRNA-seq
Sampling technique	1–20 tape strips applied to the skin	Dermal biomarker patch applied to the skin for 5 minutes	10 scrapes with curette	Skin biopsy	Skin biopsy
Layers of skin samples	Stratum corneum	Epidermis and upper dermis	Not publicly disclosed	Epidermis and dermis	Epidermis and dermis
Cryopreservation necessary	Yes. Must store tape strips at –80 °C	No. Can be stored at 4 °C and processed within 72 hours	Not publicly disclosed	No	Samples typically processed fresh
Likelihood of scar	Low	Not publicly disclosed	Not publicly disclosed	High	High
Integration into clinical workflow1	Can be collected relatively rapidly; requires specific expertise, normalization, and possibly slow turnaround time.	Can be collected relatively rapidly. Turnaround time and cost of commercial service not yet clear.	Can be collected relatively rapidly. Turnaround time and cost of commercial service not yet clear.	Integrates well with IHC workflow in dermatopathology laboratories with rapid turnaround for specific targets.	Resource-intensive sample processing, sequencing, and data analysis.
Throughput/number of markers	Full transcriptome (for RNA-based methods)	Full transcriptome	28 transcripts	One marker per staining; multiple markers can be batched	Full transcriptome
Requires sequencing	Yes	Yes	No	No	Yes
Data analysis	Computational pipelines for microarray or RNA- sequencing data.	Computational pipelines with proprietary algorithms.	Comparison of count threshold values from qPCR.	Manual or automated quantification of positive cells on microscopy slide.	Computational pipelines for scRNA-seq data.
Histologic analysis inherent	No	No	No	Yes	No

Abbreviations: RISH, RNA in situ hybridization; scRNA-seq, single-cell RNA sequencing.

¹Cost will also be an important factor in comparing the feasibility of workflow integration, but there are not enough data at this time to compare costs between methods.

Microneedle-based dermal biomarker patch

Microneedles are micron-sized needles that can be used to sample RNA or protein from the skin (Ibrahim et al., 2022; Kim et al., 2022) or interstitial fluid (Samant et al., 2020; Wang et al., 2021b). One approach that is being commercially pursued (Mindera Health) to collect material for biomarker analysis from the skin uses microneedle technology through a dermal biomarker patch (DBP) (Ibrahim et al., 2022). Their DBP is lined with 100 square pyramidal microneedles with a 200 × 200 μm base and 750 μm depth. Microneedles are treated with DNA oligonucleotides to enhance the capturing of mRNA. The DBP is applied to the skin and left on for 5 minutes. The microneedles sample into the upper dermis to a depth of 350–400 μm. Subsequent RNA sequencing showed gene detection comparable with that of the RNA sequencing of punch biopsy specimens.

This technology was used in combination with a machine-learning algorithm to assess biologic responses in patients with psoriasis (Bagel et al., 2021; Mindera Health). DBPs were applied at baseline, and the change in PASI was assessed after 12 weeks of treatment with either an IL-23 inhibitor, an IL-17 inhibitor, a TNFα inhibitor, or an IL-12/23 inhibitor. The authors trained a classifier to predict response to IL-23 inhibitors in a subset of the patients, and they trained IL-17 and TNFα inhibitor response classifiers on the basis of previously generated gene expression data. Their classifiers achieved high positive predictive value (PPV) for response to therapy, with PPV values of 93.1, 92.3, and 85.7% for IL-23, IL-17, and TNFα inhibitor classifiers, respectively, when applied to high-PASI (PASI > 8) patients.

DBP appears less invasive than a traditional skin biopsy. Compared with tape stripping, microneedles can sample down into the dermis. One group showed a higher concentration of RNA in samples collected from microneedle patches than from tape stripping (Kim et al., 2022).

There are limitations to microneedle-based approaches. Applying microneedles to inflamed skin could lead to irritation and/or discomfort. Also similar to tape stripping, transcriptional profiling from microneedles requires specialized equipment and expertise to collect and analyze the data, which may restrict its use to a centralized or commercial setting. The cost and turnaround time of this process are not fully clear, but both could be potential limitations. Whether insurance will cover such approaches, including the Mindera system, is also not yet clear. Finally, because the epidermis is breached, there is potential for scarring.

Molecular profiling from epidermal curettage samples

Another minimally invasive way to collect skin samples is by curetting superficial epidermal tissue. One approach that is being pursued commercially (Castle Biosciences, Friendswood, TX) uses this sample collection technique for molecular profiling of 28 psoriasis and AD-related genes using RT-PCR (Quick et al., 2022). This approach was able to detect gene expression differences between AD and psoriasis.

Molecular profiling from epidermal curettage has the advantage of being minimally invasive, although it is possible that the depth of sampling might vary on the basis of the end user and anatomic location. If the dermo–epidermal junction is not breached, this technique would be expected to be nonscarring. In this case, the relative ease of sample collection would also allow for more frequent sampling or sampling at more sites.

Limitations include the specialized technology and expertise required for sample processing and data analysis, which also necessitate sample processing and analysis at a centralized location. In addition, no data are reported on the reproducibility of the collection method between operators or sites. The breadth of potential biomarkers is also limited by the current approach. Turnaround time, cost, and payor considerations are not yet fully clear for the Castle Bioscience approach under development.

RNA in situ hybridization staining

Biomarkers can be measured directly in skin biopsies using techniques such as IF and IHC. Although promising in concept, IHC and IF for secreted proteins such as cytokines have proven difficult because of often high background staining (Chen et al., 2023; Cohen et al., 2020; Miranda et al., 2021; Moy et al., 2015). RNA in situ hybridization (RISH) detects specific mRNA molecules by hybridization of complementary 18–25 bp probes to tissue sections, amplification of the signal, and detection using a chromogenic or fluorescent molecule (Wang et al., 2012). RISH can be performed on skin biopsies, including frozen and formalin-fixed, paraffin-embedded diagnostic specimens. The location of the desired mRNA can then be visualized in the histologic section. RNAScope is a commonly used, commercially available system for performing RISH. With RNAScope, signal detection and amplification require tandem binding of distinct probes, and thus the assay is considered to be highly specific.

RISH has been used to detect disease-relevant cytokines in AD and psoriasis biopsies. Our group used chromogen-based RISH (with the RNAScope kit, Advanced Cell Diagnostics, Hayward, CA) in a retrospective study to determine the levels and degrees of heterogeneity of druggable cytokines in AD and psoriasis biopsies (Wang et al., 2021a). The study focused on established treatment targets, including IL4, IL12B (IL-12/23 p40), IL13, IL17A, IL17F, IL23A (IL-23 p19), and TNF (TNFα); emerging AD therapeutic targets including IL22 and IL31; and established psoriasis biomarkers NOS2 and IFNG. We found that staining was able to discriminate between psoriasis (generally NOS2+ IL17A+), AD (generally NOS2− IL13+), and healthy control samples (negligible staining).

IL17A and IL13 were most commonly detected at significant levels in psoriasis and AD, respectively. IL12B, IL23A, and IL17F were also increased in many cases of psoriasis. Expression of IL4 was also present in AD but was generally present at markedly lower levels than that of IL13, consistent with earlier studies (Tsoi et al., 2019). Within each disease, we also found distinct molecular heterogeneity, and the predominantly expressed druggable cytokines varied among patients. For example, some cases of psoriasis were IL17A predominant, whereas others were IL23A predominant, and others were mixed. Similarly, although most cases of AD were IL13 predominant, others were IL4 predominant.

In a follow-up study, RISH was used to measure type 2 (IL4, IL13), type 1 (IFNG), type 3 (IL17A and IL17F), and Th22 (IL22) cytokines in patients with eczematous dermatitis treated with dupilumab (Singh et al., 2023). In this retrospective study, cytokines were measured in diagnostic biopsies at baseline, and patterns were correlated with response to dupilumab. We hypothesized that patients with a relatively pure Th2 signal would respond optimally to dupilumab, whereas those with a mixed or alternate polarization might respond suboptimally. Indeed, we found that the best responders to dupilumab had the highest expression of IL13, whereas poor or nonresponders had either low or no IL13 and tended to express relatively higher levels of other cytokines such as IFNG and IL17A.

RISH has certain advantages in its potential application for biomarker analysis in inflammatory dermatoses. First, the technique has high specificity for the mRNA of interest with minimal background staining, making it easy to interpret. It can also be performed on diagnostic biopsies that may have already been obtained; has a rapid turnaround; fits well into standard dermatopathology workflow; and does not require a centralized service, specialized knowledge and/or equipment, or data normalization. In contrast to other approaches, histopathologic analysis is also possible in the event of an unexpected diagnosis.

Limitations of the technique include cost (probes cost roughly as much as an antibody for IHC) and difficulty in analyzing a large cytokine panel for each patient because each cytokine target requires a separate probe and staining procedure. In addition, standardized and validated approaches for scoring cytokine expression need to be developed. For some analytes, detecting mRNA as opposed to protein may be a limitation. Furthermore, RISH requires a biopsy to be performed, which may not always be part of routine clinical care. The biopsy could also result in extra costs for tissue processing and dermatopathological evaluation of the biopsy specimen as well as scarring for the patient.

Single-cell RNA sequencing

Single-cell RNA sequencing (scRNA-seq) allows for transcriptomic analysis of individual cells in tissues. scRNA-seq has been applied to human skin in a variety of disease contexts (Kim et al., 2020a), including inflammatory skin diseases (Cheng et al., 2018; He et al., 2020b; Hughes et al., 2020; Kim et al., 2020b; Reynolds et al., 2021; Rojahn et al., 2020).

Recently, Liu et al., 2022b tested the hypothesis that scRNA-seq profiling could classify CIRs. They profiled CD45+ immune cells from the skin of 31 patients, including seven patients with AD, eight patients with psoriasis, two patients with lichen planus, one patient with bullous pemphigoid, six patients with CIRs, and seven healthy controls. They identified AD and psoriasis-specific gene sets within the tissue-resident memory T-cell population, which included Th2- and Th17-associated genes. When they placed CIRs onto this Th2/Th17 map, they found that the classification of CIRs as more AD like correlated with positive response to dupilumab. They created a free web-based tool RashX for users to upload scRNA-seq to assist with molecular classification.

One advantage of scRNA-seq profiling is that it could allow for both treatment selection and the discovery of new biomarkers in inflammatory skin conditions. The creation of centralized databases such as RashX allows for standardization of data analysis and enhanced data sharing.

Limitations of scRNAseq profiling include the high cost and difficulty of integrating scRNA-seq into clinical work-flows. scRNA-seq requires specialized equipment, samples may need to ideally be processed fresh, and this approach may only be possible in academic settings. Tissue dissociation methods also vary and a biopsy is required which leads to scarring. Depending on the tissue processing approach, the entire tissue might be depleted for scRNA-seq and thus histopathologic analysis could require obtaining a second specimen.

Other technologies

Suction blistering is another technique for RNA and proteomic profiling of the skin. Blisters are induced at the dermo–epidermal junction over a few hours. Blister fluid can be used directly for proteomic analysis (Müller et al., 2012; Szegedi et al., 2015), or cells can be collected from the fluid and/or blister roof for transcriptomic analysis. Rojahn et al. (2020) performed scRNA-seq of blister samples from patients with AD and healthy controls and found that the expression profiles of cells from the blister roof recapitulated cell type–specific expression patterns from biopsy samples. Suction blistering was also used to study the persistence of inflammatory signatures in the skin of patients with AD on long-term treatment with dupilumab (Bangert et al., 2021). This method typically does not result in significant scarring. A long sample collection time is a limitation.

RNA can also be detected from skin surface lipids collected noninvasively by wiping the face or scalp with oil-blotting film (Inoue et al., 2022; Shima et al., 2022; Yamamoto-Hanada et al., 2023). Using this method, researchers have shown the capturing of RNA derived from sebaceous glands, epidermis, and hair follicles and showed an increase in inflammation-related RNA molecules in skin surface lipids of patients with AD.

Another method to sample biological material from the skin involves treating the skin with chemical reagents in combination with gentle mechanical abrasion or ultrasound energy (Hwang et al., 2013, 2012; Muradova et al., 2021; Paliwal et al., 2010). Researchers have applied this technique to measure skin surface molecules in AD and psoriasis by embedding a capture-antibody microarray onto a skin patch with quantification by spot ELISA (Røpke et al., 2021; Schaap et al., 2021).

Skin microdialysis has been used to sample soluble materials from the skin of patients with AD (Neisius et al., 2002; Papoiu et al., 2011; Steinhoff et al., 2003) and psoriasis (Buerger et al., 2012; Krogstad et al., 1997; Salgo et al., 2011; Sjögren et al., 2012). This technique may be particularly useful for quantification of drug concentrations (Garcia Ortiz et al., 2009; Neisius et al., 2002; Quist et al., 2016a, 2016b; Turpeinen et al., 1988), continuous monitoring of soluble material concentrations (Clough et al., 2007), or simultaneous sampling and delivery of a therapeutic (Rukwied et al., 2000). This technique is limited in its ability to detect relatively higher-molecular-weight soluble factors such as cytokines (Baumann et al., 2019).

Integration of Biomarkers

Studies in psoriasis have integrated transcriptomic and proteomic profiling to identify molecules that are differentially expressed at both the RNA and protein levels (Piruzian et al., 2010; Swindell et al., 2015). One study also compared gene expression signatures from skin and blood with those of other skin diseases, finding that the blood signature was more psoriasis specific (Swindell et al., 2016). Ungar et al. (2017) combined serum protein measurements with gene expression measurements from lesional and nonlesional skin of patients with AD. Surprisingly, they found that nonlesional skin biomarkers correlated more strongly with serum biomarkers than lesional skin biomarkers. In a pilot study, Foulkes et al. (2019) performed an integrated analysis of serum biomarkers (protein, mRNA, and microRNA) with mRNA measurements from lesional and nonlesional skin to examine the response of patients with psoriasis to TNFα inhibition. The authors built random forest models to classify responses to treatment on the basis of baseline blood and skin measurements. Although underpowered, their pilot study suggests that baseline skin and blood biomarkers may be able to predict response to therapy, which will be explored in future larger studies (Griffiths et al., 2015).

CONCLUSION

An ideal method for biomarker-driven diagnosis and treatment selection is accurate at characterizing the molecular changes present in a patient sample, reproducible across laboratories and conditions, practical in terms of the cost of reagents and analysis, and clinically feasible. Although there are currently no widely used or approved biomarker-based methods to inform therapy selection in AD and psoriasis, several methods are being developed and evaluated. These technologies have limitations, and there are still many open questions (Table 2). We also lack large-scale data at this time showing the clinical and/or financial feasibility of these approaches. Conceptually, given the growing number of biologics with diverse targets for psoriasis, AD, and other inflammatory dermatoses, further development of technologies for biomarker-based therapy selection will be valuable for patients, physicians, and the healthcare system.

Box 1. Open Questions and Limitations of Existing Technologies.

Open Questions:

1. How does sampling affect biomarker detection?

   1. Anatomic location
   2. Stage of individual lesion
   3. Duration of underlying disease
   4. Treatment status of patient
   5. Severity of disease at moment in time (flare status)
   6. Technical sampling differences between operators
2. What are practical limitations to biomarker detection in the clinic?

   1. Insurance and reimbursement
   2. Perceived utility by clinicians
   3. Time taken to collect samples and coordinate downstream analysis
   4. Standardization among labs
   5. Resource limitations
3. How do immunologic biomarkers impact diagnosis and disease classification?

   1. Nosologic considerations (cases with overlapping morphology and immunology)
   2. Clinically and histologically indeterminate rashes
4. Limitations of existing technologies

   1. Lack large-scale data showing the clinical and/or financial feasibility of these approaches
   2. Antidrug antibodies are not assessed
   3. Dose- and patient-specific pharmacokinetics may influence response to biologic therapy (Tsakok et al. 2019; Wilkinson et al. 2019)

Abbreviations:

AD

atopic dermatitis

CIR

clinically/histopathologically indeterminate rash

DBP

dermal biomarker patch

EASI

Eczema Area and Severity Index

IF

immunofluorescence

IHC

immunohistochemistry

NOS2

nitric oxide synthase 2

PPV

positive predictive value

RISH

RNA in situ hybridization

scRNA-seq

single-cell RNA sequencing

Th

T helper