Advertisement

Cytokine RNA In Situ Hybridization Permits Individualized Molecular Phenotyping in Biopsies of Psoriasis and Atopic Dermatitis

Open AccessPublished:May 06, 2021DOI:https://doi.org/10.1016/j.xjidi.2021.100021
      Detection of individual cytokines in routine biopsies from patients with inflammatory skin diseases has the potential to personalize diagnosis and treatment selection, but this approach has been limited by technical feasibility. We evaluate whether a chromogen-based RNA in situ hybridization approach can be used to detect druggable cytokines in psoriasis and atopic dermatitis. A series of psoriasis (n = 20) and atopic dermatitis (n = 26) biopsies were stained using RNA in situ hybridization for IL4, IL12B (IL-12/23 p40), IL13, IL17A, IL17F, IL22, IL23A (IL-23 p19), IL31, and TNF (TNF-α). NOS2 and IFNG, canonical psoriasis biomarkers, were also included. All 20 of the psoriasis cases were positive for IL17A, which tended to be the predominant cytokine, although some cases had relatively higher levels of IL12B, IL17F, or IL23A. The majority of cytokine expression in psoriasis was epidermal. A total of 22 of 26 atopic dermatitis cases were positive for IL13, also at varying levels; a subset of cases had significant IL4, IL22, or IL31 expression. Patterns were validated in independent bulk RNA-sequencing and single-cell RNA-sequencing datasets. Overall, RNA in situ hybridization for cytokines appears highly specific with virtually no background staining and may allow for individualized evaluation of treatment-relevant cytokine targets in biopsies from patients with inflammatory skin disorders.

      Graphical abstract

      Abbreviations:

      AD (atopic dermatitis), Cat (catalog), DC (dendritic cell), IHC (immunohistochemistry), ILC (innate lymphoid cell), KC (keratinocyte), LC (Langerhans cell), RISH (RNA in situ hybridization), RNA-seq (RNA sequencing), scRNA-seq (single-cell RNA sequencing)

      Introduction

      Inflammatory dermatologic disorders are increasingly treated with mAbs that inhibit the activity of specific cytokines (biologics). Clinical selection of biologic medicines is done empirically, and it remains unclear why some patients respond to a particular biologic treatment and others do not. For example, in psoriasis, biologics targeting TNF-α, IL-12/23 (p40), IL-17A, IL-17 family (receptor), and IL-23 (p19) are approved. Whereas the majority of patients with psoriasis treated with newer biologics (e.g., IL-17 axis or IL-23 inhibitors) will experience clear or almost clear skin, a subset will not (
      • Bai F.
      • Li G.G.
      • Liu Q.
      • Niu X.
      • Li R.
      • Ma H.
      Short-term efficacy and safety of IL-17, IL-12/23, and IL-23 inhibitors brodalumab, secukinumab, ixekizumab, ustekinumab, guselkumab, tildrakizumab, and risankizumab for the treatment of moderate to severe plaque psoriasis: a systematic review and network meta-analysis of randomized controlled trials.
      ). In atopic dermatitis (AD), the biologic dupilumab, which targets IL-4Rα and inhibits both IL-4 and IL-13 activity, is approved. However, only approximately one third of patients with AD treated with dupilumab achieve clear or nearly clear skin (
      • Wang F.P.
      • Tang X.J.
      • Wei C.Q.
      • Xu L.R.
      • Mao H.
      • Luo F.M.
      Dupilumab treatment in moderate-to-severe atopic dermatitis: a systematic review and meta-analysis.
      ). One potential explanation for these observations is that underappreciated immunologic heterogeneity exists across patients, especially across patients with AD, and therefore drugs targeting specific cytokines might be more or less effective for individual patients (
      • Glickman J.W.
      • Han J.
      • Garcet S.
      • Krueger J.G.
      • Pavel A.B.
      • Guttman-Yassky E.
      Improving evaluation of drugs in atopic dermatitis by combining clinical and molecular measures.
      ;
      • Tsoi L.C.
      • Rodriguez E.
      • Degenhardt F.
      • Baurecht H.
      • Wehkamp U.
      • Volks N.
      • et al.
      Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
      ).
      Physicians face the increasing challenge of selecting the optimal biologic for individual patients, particularly when many choices might be available for a disease. Apart from medical comorbidities and patient preferences for drug administration, the choice of biologic is generally empirical, and some patients quickly transition to a different agent if they do not experience the desired clinical response. This practice is suboptimal because biologic therapies can have significant adverse effects and are expensive. Ideally, immunologic parameters, including the presence or absence as well as the relative abundance of treatment-relevant cytokines, could be measured in individual patients to help guide personalized diagnosis and treatment. Currently, there are no widely utilized methodologies to assess an individual patient’s cytokine profile as part of clinical practice.
      Previous attempts to characterize cytokine expression or immune polarization (e.g., type 1 vs. type 2 vs. type 17) in skin biopsies with approaches such as immunohistochemistry (IHC) while innovating in concept have been hampered by high background staining owing to nonspecific binding of antibody, limiting interpretation of true positive staining (
      • Cohen J.N.
      • Bowman S.
      • Laszik Z.G.
      • North J.P.
      Clinicopathologic overlap of psoriasis, eczema, and psoriasiform dermatoses: a retrospective study of T helper type 2 and 17 subsets, interleukin 36, and β-defensin 2 in spongiotic psoriasiform dermatitis, sebopsoriasis, and tumor necrosis factor α inhibitor-associated dermatitis.
      ;
      • Miranda E.
      • Roberts J.
      • Novick S.
      • Lapointe J.M.
      • Bruijnzeel-Koomen C.
      • Thijs J.
      • et al.
      Immunohistochemical characterization of the IL-13:IL-4 receptor α axis in the skin of adult patients with moderate to severe atopic dermatitis and healthy controls.
      ;
      • Moy A.P.
      • Murali M.
      • Kroshinsky D.
      • Duncan L.M.
      • Nazarian R.M.
      Immunologic overlap of helper T-cell Subtypes 17 and 22 in erythrodermic psoriasis and atopic dermatitis.
      ). Immunofluorescence, which can be more specific in some instances, is more technically challenging and typically requires specialized tissue collection and processing. For example, a recent study using carefully selected molecular immunologic makers of type-2 versus those of type-17 inflammation was unable to use the markers to distinguish between cases of psoriasis and AD (
      • Cohen J.N.
      • Bowman S.
      • Laszik Z.G.
      • North J.P.
      Clinicopathologic overlap of psoriasis, eczema, and psoriasiform dermatoses: a retrospective study of T helper type 2 and 17 subsets, interleukin 36, and β-defensin 2 in spongiotic psoriasiform dermatitis, sebopsoriasis, and tumor necrosis factor α inhibitor-associated dermatitis.
      ). Although powerful, RNA sequencing (RNA-seq), a widely used research approach to characterize gene expression profiles in tissue, is time consuming and expensive and requires specialized tissue processing, equipment, and expertise, making its incorporation into routine clinical care impractical.
      RNA in situ hybridization (RISH) uses nucleic acid‒based probes to detect mRNA. A potential advantage of this approach is that the probes are highly specific, and labeling can be visualized in formalin-fixed, paraffin-embedded specimens using chromogenic approaches akin to those used for IHC. Despite its potential utility in this setting, RISH has largely been overlooked in clinical dermatology, including in the area of inflammatory dermatoses.
      In this study, we evaluated the ability of RISH to detect pathogenic cytokines in psoriasis and AD and compared the patterns with those in normal skin from healthy controls. We focused on established treatment targets with approved drugs (IL4, IL13, IL12B, IL17A, IL17F, IL23A, TNF), emerging therapeutic targets in AD (IL22, IL31), and canonical psoriasis biomarkers (NOS2, IFNG) (
      • Garzorz-Stark N.
      • Krause L.
      • Lauffer F.
      • Atenhan A.
      • Thomas J.
      • Stark S.P.
      • et al.
      A novel molecular disease classifier for psoriasis and eczema.
      ;
      • He H.
      • Bissonnette R.
      • Wu J.
      • Diaz A.
      • Saint-Cyr Proulx E.
      • Maari C.
      • et al.
      Tape strips detect distinct immune and barrier profiles in atopic dermatitis and psoriasis.
      ;
      • Quaranta M.
      • Knapp B.
      • Garzorz N.
      • Mattii M.
      • Pullabhatla V.
      • Pennino D.
      • et al.
      Intraindividual genome expression analysis reveals a specific molecular signature of psoriasis and eczema.
      ). We assessed the ability of RISH-based cytokine profiles to distinguish between the cases of psoriasis and those of AD as well as to study the intrapatient cytokine heterogeneity within each diagnosis. RISH staining patterns were compared with previously published bulk RNA-seq and single-cell RNA-seq (scRNA-seq) data from psoriasis and AD (
      • Reynolds G.
      • Vegh P.
      • Fletcher J.
      • Poyner E.F.M.
      • Stephenson E.
      • Goh I.
      • et al.
      Developmental cell programs are co-opted in inflammatory skin disease.
      ;
      • Tsoi L.C.
      • Rodriguez E.
      • Degenhardt F.
      • Baurecht H.
      • Wehkamp U.
      • Volks N.
      • et al.
      Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
      ).

      Results

      Cytokine expression patterns differentiate between psoriasis and AD, but heterogeneity exists

      To broadly understand which cytokines and other markers are most different between psoriasis and AD lesions, we analyzed bulk RNA-seq data from lesional skin in a cohort of patients with psoriasis (n = 28), those with AD (n = 27), and healthy controls (n = 38) (
      • Tsoi L.C.
      • Rodriguez E.
      • Degenhardt F.
      • Baurecht H.
      • Wehkamp U.
      • Volks N.
      • et al.
      Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
      ). We found that NOS2 (encoding inducible nitric oxide synthase) was markedly upregulated in psoriasis and was the most significantly differentially expressed transcript between the two conditions (Figure 1a). NOS2 upregulation in psoriasis has been observed previously (
      • He H.
      • Bissonnette R.
      • Wu J.
      • Diaz A.
      • Saint-Cyr Proulx E.
      • Maari C.
      • et al.
      Tape strips detect distinct immune and barrier profiles in atopic dermatitis and psoriasis.
      ;
      • Quaranta M.
      • Knapp B.
      • Garzorz N.
      • Mattii M.
      • Pullabhatla V.
      • Pennino D.
      • et al.
      Intraindividual genome expression analysis reveals a specific molecular signature of psoriasis and eczema.
      ).
      Figure thumbnail gr1
      Figure 1Cytokine expression patterns differentiate psoriasis and AD, but heterogeneity exists. (a) Volcano plot showing the relative mRNA expression of selected genes in cases of psoriasis (n = 28) and in cases of AD (n = 27) on the basis of analysis of RNA-seq data (
      • Tsoi L.C.
      • Rodriguez E.
      • Degenhardt F.
      • Baurecht H.
      • Wehkamp U.
      • Volks N.
      • et al.
      Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
      ). (b) mRNA expression levels of selected genes among individual patients with each diagnosis, corresponding to data in (a). AD, atopic dermatitis; RNA-seq, RNA sequencing.
      We also observed that genes encoding cytokine targets of approved treatments were among the most differentially expressed transcripts. In particular, IL17A and IL13 were markedly upregulated in psoriasis and AD, respectively. IL12B (encoding IL-12/23 p40), IL23A (encoding IL-23 p19), and IL17F were also upregulated in psoriasis, as expected. Significant IL4 expression was not detected in these samples, although it did tend to be higher in AD, consistent with previous reports (
      • Tsoi L.C.
      • Rodriguez E.
      • Degenhardt F.
      • Baurecht H.
      • Wehkamp U.
      • Volks N.
      • et al.
      Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
      ). Emerging treatment targets, including IL31 and IL22 as well as IL36A/G, were increased in AD and psoriasis, respectively.
      Next, we assessed the heterogeneity in the expression of key cytokines within these samples. We found that there was considerable variability within both psoriasis and AD in terms of the predominant druggable cytokine expressed (Figure 1b). For example, some cases of psoriasis were IL17A predominant, but others expressed very little IL17A and were instead IL17F, IL23A, and/or IL12B predominant. Interestingly, some cases of AD did not show significant expression of either IL13 or IL4, and one case was IL17F predominant. These data suggest that there is significant molecular heterogeneity among cases of psoriasis and AD, which might have important implications for optimal treatment selection.

      Inducible nitric oxide synthase (encoded by NOS2) staining differentiates psoriasis from AD

      To assess RISH staining in this setting, we assembled cohorts of patients with psoriasis (n = 20) and AD (n = 26) in which biopsy tissue was available for study (Table 1). We first evaluated whether RISH staining for NOS2 differed between psoriasis and AD, as would be predicted from the analysis of the RNA-seq data. Remarkably, we found that NOS2 staining was present in all the 20 psoriasis cases and that only 1 of 26 cases of AD had any detectable NOS2 expression, and this expression was minimal (Figure 2a–c). No staining was observed in any of the normal controls. NOS2 staining in psoriasis was positive in keratinocytes (KCs) in the upper stratum spinosum (Figure 2b). The apparent specificity of NOS2 with little background emphasized the potential power of RISH as an approach and also provided molecular immunologic validation of the clinicopathologic classifications for cases included in this series.
      Table 1Patient Characteristics
      CharacteristicsPsoriasisAtopic DermatitisNormal
      Number of patients202610
      Age, y, median (range)48 (18–79)57 (24–83)50 (30–72)
      Sex, % M; % F70 M; 30 F68 M; 32 F23 M; 77 F
      Anatomic site, n/total (%)
       Trunk7/20 (35)7/26 (27)5/10 (50)
       Extremities11/20 (55)16/26 (61)5/10 (50)
       Acral2/20 (10)1/26 (4)0/10 (0)
       Head/neck0/20 (0)2/26 (8)0/10 (0)
      Abbreviations: F, female; M, male.
      Figure thumbnail gr2
      Figure 2NOS2 RISH staining differentiates psoriasis from AD and from normal skin. (a) H&E-stained sections (left panels), NOS2 RISH staining (middle panels), and quantitative analysis (right panels). Representative cases are shown for each psoriasis case, AD case, and normal controls. Bar = 100 μM. (b) NOS2 RISH staining in a representative case of psoriasis. Bar = 50 μM. (c) Quantification of NOS2 RISH in psoriasis, AD, and normal skin, shown as the number of positive cells per millimeter of the epidermis. AD, atopic dermatitis; Norm, normal; RISH, RNA in situ hybridization.

      IL17A and IL13 staining patterns also differentiate psoriasis from AD

      Analysis of the RNA-seq data suggested that IL17A and IL13 might have the best ability to differentially label psoriasis and AD. Furthermore, IL-17A and IL-13 are key targets of treatment in these disorders. We therefore stained cases using RISH probes specific for IL17A and IL13. We found that all the 20 cases of psoriasis showed detectable IL17A expression in the epidermis (Figure 3a and b). Interestingly, there were significantly fewer IL17A-positive cells in the dermis, despite the majority of the inflammation being present in the dermis. As with the RNA-seq data, we found that there was considerable variability in the relative abundance of IL17A among psoriasis cases. In contrast, only 2 of the 26 cases of AD had any detectable IL17A-positive cells in the epidermis, and staining was minimal. This may support previous literature suggesting that IL-17 may play a driver role in a small subset of patients with AD (
      • Ungar B.
      • Pavel A.B.
      • Li R.
      • Kimmel G.
      • Nia J.
      • Hashim P.
      • et al.
      Phase 2 randomized, double-blind study of IL-17 targeting with secukinumab in atopic dermatitis.
      ). IL17A staining was not observed in any of the normal controls.
      Figure thumbnail gr3
      Figure 3IL17A and IL13 RISH staining patterns differentiate psoriasis from AD and from normal skin. (a) IL17A RISH staining (left and middle panels) and quantitative analysis (right panels). Black bar = 100 μM, green bar = 30 μM. Representative images are shown for each psoriasis, AD, and normal skin. (b) Quantification of IL17A RISH staining data in psoriasis, AD, and normal controls. The number of positive cells in the epidermis (left) and dermis (right) was quantified independently. (c) IL13 RISH staining (left and middle panels) and quantitative analysis (right panels). Black bar = 100 μM, green bar = 30 μM. Representative images are shown for each psoriasis, AD, and normal controls. (d) Quantification of IL13 RISH staining data in psoriasis, AD, and norm controls; the number of positive cells in the epidermis (left) and dermis (right) was quantified independently. AD, atopic dermatitis; Norm, normal; RISH, RNA in situ hybridization.
      IL13 staining was observed in 85% of the AD cases (Figure 3c and d). Whereas some cases had epidermal predominant IL13, other cases had dermal predominant or exclusive expression (Figure 3c and d). The abundance of IL13-positive cells was also quite variable in the AD cases. Most cases of psoriasis were negative for IL13; however, focal staining was observed in some cases. No IL13 staining was observed in normal skin. Overall, IL17A and IL13 staining patterns alone were sufficient to distinguish between psoriasis (IL17A predominant), AD (IL13 predominant), and normal (negative staining) in 52 of the 56 cases in the series.

      IL17A and IL13 RISH staining is found predominantly in CD3+ T cells

      In both IL17A and IL13 RISH stains, cells that stained positively tended to have a lymphocyte morphology as would be expected. To explore this and also to further validate the specificity of the approach, we performed double staining for CD3 (IHC) and for either IL17A or IL13 in seven randomly selected cases each of psoriasis and AD. We found that the vast majority of IL17A- and IL13-positive cells also colabeled with CD3, suggesting that they may represent type-17‒ and type-2‒polarized T cells, respectively (Figure 4a and c). We focused on the relative abundance of intraepidermal cytokine-producing T cells as a function of total T cells in the epidermis because particularly in psoriasis, most cytokine-producing cells were found here. This analysis showed that cytokine-producing T cells ranged from 3.5% to 15.1% of all intraepidermal T cells in these samples (Figure 4b and d).
      Figure thumbnail gr4
      Figure 4IL13 and IL17A expressions localize to CD3+ T lymphocytes. (a) Double stain showing IL13 RISH (red) and CD3 IHC (brown) in a representative case of AD. Bar = 30 μM. Representative analysis using QuPath and ImageJ is shown. (b) Quantification of the number of IL13-positive cells as a function of total CD3+ cells in AD epidermis (n = 7). (c) Double stain showing IL17A ISH (red) and CD3 IHC (brown) in a representative case of psoriasis. Bar = 30 μM. Representative analysis using QuPath and ImageJ is shown. (d) Quantification of the number of IL17A-positive cells as a function of the total CD3+ cells in psoriasis epidermis (n = 7). AD, atopic dermatitis; IHC, immunohistochemistry; ISH, in situ hybridization.

      IL17F levels generally correlate with IL17A levels in psoriasis

      IL-17F is highly homologous to, is often coexpressed with, and can act synergistically with IL-17A (
      • McGeachy M.J.
      • Cua D.J.
      • Gaffen S.L.
      The IL-17 family of cytokines in health and disease.
      ). It is conceivable that some cases of psoriasis might be relatively more dependent on IL-17F (or IL-17C) given the observation that some patients respond better to IL-17R blockade than to IL-17A blockade (
      • Kimmel G.
      • Chima M.
      • Kim H.J.
      • Bares J.
      • Yao C.J.
      • Singer G.
      • et al.
      Brodalumab in the treatment of moderate to severe psoriasis in patients when previous anti-interleukin 17A therapies have failed.
      ). We next investigated the expression of IL17F in this cohort and compared this with IL17A expression. We found that IL17F was detectable in 19/20 cases of psoriasis and, similarly to IL17A, was generally present in the epidermis (Figure 5a and b). As expected, IL17A tended to be the predominant cytokine in most cases of psoriasis and was highly correlated with IL17F, whereas it was inversely correlated with IL13 (Figure 5c and d). Interestingly, however, in three cases of psoriasis, IL17F was predominant. There was either no or negligible IL17F expression in most cases of AD, and no expression was observed in healthy controls.
      Figure thumbnail gr5
      Figure 5IL17F staining correlates with IL17A staining in psoriasis. (a) IL17F RISH staining (left panel) and quantitative analysis (right panel) in a representative case of psoriasis. Bar = 100 μM. (b) Quantification of IL17A and IL17F RISH staining. The number of positive cells in the epidermis (left) and dermis (right) were quantified independently. (c) Scatter plot of IL17A versus that of IL17F levels in the epidermis of psoriasis and AD cases. (d) Scatter plot of IL17A versus that of IL13 levels in the epidermis of psoriasis and AD cases. AD, atopic dermatitis; Norm, normal; RISH, RNA in situ hybridization.

      IL4, IL22, IL31, IL12B (IL-12/23 p40), and IL23A (IL-23 p19) staining can also be used to characterize psoriasis and AD

      We next used RISH probes for IL4, IL12B, and IL23A to further characterize the skin biopsies from these patients and assess what additional information they might provide, given that there are approved drugs that target these cytokines. We found that IL4 expression was much lower than that of IL13 and was not detected at significant levels in the epidermis of most cases of AD (Figure 6a and d). The predominance of IL13 over IL4 in AD lesional skin is consistent with previous reports (
      • Bieber T.
      Interleukin-13: targeting an underestimated cytokine in atopic dermatitis.
      ;
      • Tsoi L.C.
      • Rodriguez E.
      • Degenhardt F.
      • Baurecht H.
      • Wehkamp U.
      • Volks N.
      • et al.
      Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
      ). The expression of IL4 in some cases but not in others is interesting and might be important in the setting of IL-13‒specific inhibition, which is being evaluated in AD.
      Figure thumbnail gr6
      Figure 6IL4, IL12B, and IL23A staining can also be used to classify cases of psoriasis and AD. (a) IL4 RISH staining (left and middle panels) and quantitative analysis (right panels). Black bar = 100 μM, green bar = 30 μM. Representative images are shown for each psoriasis, AD, and normal control. (b) IL12B RISH staining (left and middle panels) and quantitative analysis (right panels). Black bar = 100 μM, green bar = 30 μM. Representative images are shown for each psoriasis, AD, and normal control. (c) IL23A RISH staining (left and middle panels) and quantitative analysis (right panels). Black bar = 100 μM, green bar = 30 μM. Representative images are shown for each psoriasis, AD, and normal control. (d) Quantification of IL4, IL12B, and IL23A RISH in psoriasis, AD, and normal skin shown as the number of positive cells per millimeter of the epidermis (left panels) or dermis (right panels). AD, atopic dermatitis; Norm, normal; Pso, psoriasis; RISH, RNA in situ hybridization.
      IL12B (IL-12/23 p40) staining within the epidermis was significantly higher in psoriasis than in AD; however, in the dermis, staining was present in both (Figure 6b and d). The significance of IL12B staining in the dermis of some of the AD cases is unclear and may represent bystander cells and/or cells producing but not secreting IL-12/23. Of note, we also observed IL12B expression in some AD cases in the bulk RNA-seq experiments (Figure 1b). Interestingly, IL-12/23 (p40) expression has been previously reported in AD, especially in chronic cases (
      • Caproni M.
      • Torchia D.
      • Antiga E.
      • Terranova M.
      • Volpi W.
      • del Bianco E.
      • et al.
      The comparative effects of tacrolimus and hydrocortisone in adult atopic dermatitis: an immunohistochemical study.
      ;
      • Yawalkar N.
      • Karlen S.
      • Egli F.
      • Brand C.U.
      • Graber H.U.
      • Pichler W.J.
      • et al.
      Down-regulation of IL-12 by topical corticosteroids in chronic atopic dermatitis.
      ), and rarely, patients with AD can improve with ustekinumab (
      • Khattri S.
      • Brunner P.M.
      • Garcet S.
      • Finney R.
      • Cohen S.R.
      • Oliva M.
      • et al.
      Efficacy and safety of ustekinumab treatment in adults with moderate-to-severe atopic dermatitis.
      ;
      • Shroff A.
      • Guttman-Yassky E.
      Successful use of ustekinumab therapy in refractory severe atopic dermatitis.
      ). No staining for IL12B was observed in the normal controls. IL23A (IL-23p19) staining within the epidermis was present in most of the psoriasis cases and was rarely present in AD cases (Figure 6c and d). Relatively less staining for IL23A was present in the dermis of both conditions. We were struck that most of the detectable IL23A production using this approach in psoriasis was in the epidermis, as opposed to that in the dermis. No staining for IL23A was observed in the normal controls.
      Interestingly, the predominance of staining for both IL12B and IL23A in psoriasis was within the epidermis. Whereas clusters of positive cells predominated in psoriasis, only individual positive cells were present in positive cases of AD (Figure 7a and b). Although IL-12 and IL-23 have been reported to be predominantly produced by dendritic cells (DCs) in the dermis in psoriasis (
      • Hawkes J.E.
      • Yan B.Y.
      • Chan T.C.
      • Krueger J.G.
      Discovery of the IL-23/IL-17 signaling pathway and the treatment of psoriasis.
      ), production by DCs that either enter the epidermis or reside in the epidermis (e.g., Langerhans cells [LCs]) in psoriasis has also been described (
      • Martini E.
      • Wikén M.
      • Cheuk S.
      • Gallais Sérézal I.
      • Baharom F.
      • Ståhle M.
      • et al.
      Dynamic changes in resident and infiltrating epidermal dendritic cells in active and resolved psoriasis.
      ;
      • Nakajima K.
      • Kataoka S.
      • Sato K.
      • Takaishi M.
      • Yamamoto M.
      • Nakajima H.
      • et al.
      Stat3 activation in epidermal keratinocytes induces Langerhans cell activation to form an essential circuit for psoriasis via IL-23 production.
      ).
      Figure thumbnail gr7
      Figure 7IL22, IL31, IFNG, and TNF staining patterns in psoriasis and AD. (a) Representative IL23A and IL12B staining patterns in psoriasis. Bar = 50 μM. (b) Representative IL23A and IL12B staining patterns in AD. Bar = 50 μM. (c) Quantification of IFNG, IL22, IL31, and TNF RISH in psoriasis, AD, and normal skin, shown as the number of positive cells per millimeter of the epidermis (upper panels) or dermis (lower panels). AD, atopic dermatitis; Norm, normal; Pso, psoriasis; RISH, RNA in situ hybridization.
      We also stained the cases using probes for IL22, IL31, IFNG (IFN-γ), and TNF (TNF-α). As expected, IFNG staining was increased in psoriasis relative to that in AD (Figure 7c). IL-31 and IL-22 are emerging treatment targets in AD. Although IL-22 inhibition in AD has been overall less effective than hoped, cases with significant IL-22 expression appear to respond more optimally to IL-22 blockade (
      • Brunner P.M.
      • Pavel A.B.
      • Khattri S.
      • Leonard A.
      • Malik K.
      • Rose S.
      • et al.
      Baseline IL-22 expression in patients with atopic dermatitis stratifies tissue responses to fezakinumab.
      ). IL-22 has also been implicated in psoriasis and may be a cytokine associated with epithelial hyperplasia during chronic inflammation (
      • Zheng Y.
      • Danilenko D.M.
      • Valdez P.
      • Kasman I.
      • Eastham-Anderson J.
      • Wu J.
      • et al.
      Interleukin-22, a T(H)17 cytokine, mediates IL-23-induced dermal inflammation and acanthosis.
      ). We found that IL-22 was slightly higher in AD cases but was not significantly different between psoriasis and AD (Figure 7c). IL31 was significantly upregulated in AD compared with that in psoriasis; most expression was present in the dermis (Figure 7c).
      TNF staining was not able to distinguish between AD and psoriasis (Figure 7c). Analysis of bulk RNA-seq data (Figure 1a) also showed this. This has been described previously and may relate to the observation that TNF mRNA is preproduced intracellularly and held under translational repression (
      • Salerno F.
      • Engels S.
      • van den Biggelaar M.
      • van Alphen F.P.J.
      • Guislain A.
      • Zhao W.
      • et al.
      Translational repression of pre-formed cytokine-encoding mRNA prevents chronic activation of memory T cells.
      ), and so measurement of mRNA levels may not accurately estimate TNF-α activity.

      Analysis of cytokine expression patterns with scRNA-seq

      Overall, with the RISH staining, we were struck by the epidermal predominant expression of type-17 cytokines, including IL17A, IL17F, IL12B, and IL23A, in psoriasis. To further evaluate this finding, we turned to a recently published scRNA-seq study of psoriasis and AD by
      • Reynolds G.
      • Vegh P.
      • Fletcher J.
      • Poyner E.F.M.
      • Stephenson E.
      • Goh I.
      • et al.
      Developmental cell programs are co-opted in inflammatory skin disease.
      . In this study, the authors used scRNA-seq to compare between psoriasis (n = 3), AD (n = 4), and healthy skin (n = 5). Before dissociation for scRNA-seq, the epidermis and dermis were separated from each other and were analyzed in distinct samples for each case. Thus, this provided an ideal dataset with which to validate epidermal versus dermal cytokine mRNA expression patterns as determined by RISH as well as cell-type specificity of expression.
      There were 528,253 cells in the study by
      • Reynolds G.
      • Vegh P.
      • Fletcher J.
      • Poyner E.F.M.
      • Stephenson E.
      • Goh I.
      • et al.
      Developmental cell programs are co-opted in inflammatory skin disease.
      . We focused only on T cells, innate lymphoid cells (ILCs), NK cells, other myeloid cells, and KCs according to the cell-type designations as determined by the original authors. In addition, we only analyzed data from lesional skin samples, resulting in a total of 127,219 cells for analysis. Clustering of these data was visualized using t-distributed Stochastic Neighbor Embedding (Figure 8a, b, and d). Next, we looked at the proportion of T cells in the epidermal versus the proportion in the dermal preparations across the samples. In psoriasis, although most T cells were located in the dermis, the vast majority of IL17A-producing T cells were found in the epidermis (Figure 8c). In contrast, although T cells were more evenly distributed between the epidermis and the dermis in AD, the majority of IL13 production was in the dermis (although there was some in the epidermis too) (Figure 8c). Overall, these patterns are highly consistent with our RISH data, although we did detect relatively more epidermal IL13 in our samples.
      Figure thumbnail gr8
      Figure 8Analysis of scRNA-seq data in psoriasis (n = 3), AD (n = 4), and normal control (n = 5) skin. Reanalysis of data from
      • Reynolds G.
      • Vegh P.
      • Fletcher J.
      • Poyner E.F.M.
      • Stephenson E.
      • Goh I.
      • et al.
      Developmental cell programs are co-opted in inflammatory skin disease.
      . (a) t-SNE plot of KCs, myeloid cells, and T cells colored by sample type. (b) t-SNE in (a) colored by epidermal location versus dermal location. (c) Bar graphs showing the proportion of different T cell populations in the epidermis and dermis by diagnosis. (d) t-SNE plots from a showing expression patterns of individual genes. (e, f) Dot plots showing cytokine expression by various cell types. For c–f, cell identity is based on designations assigned by
      • Reynolds G.
      • Vegh P.
      • Fletcher J.
      • Poyner E.F.M.
      • Stephenson E.
      • Goh I.
      • et al.
      Developmental cell programs are co-opted in inflammatory skin disease.
      . AD, atopic dermatitis; Avg, average; Exprn, expression; ILC, innate lymphoid cells; KC, keratinocytes; Mac, macrophages; Rel, relative; scRNA-seq, single-cell RNA sequencing; t-SNE, t-distributed stochastic neighbor embedding.
      Next, we looked more broadly at cytokine expression among T cells, NK cells, ILCs, macrophages, DCs, and KCs as a function of epidermal versus that of dermal derivation (Figure 8e). Most cytokine expression was from T cells, consistent with our observations, but some were also found in ILCs and NK cells. In psoriasis, the majority of IL12B and IL23A expression was from myeloid cells (Figure 8f). DCs in the epidermis produced the highest levels of IL12B in psoriasis. Whereas the magnitude of IL23A expression was higher in the epidermis, the proportion of cells producing IL23A was slightly higher in the dermis.

      Principal component analysis using RISH patterns distinguishes psoriasis from AD

      Finally, we performed principal component analysis on these cases using the cytokine and NOS2 RISH staining data. This analysis showed that psoriasis cases generally clustered together and that AD cases also generally clustered together (Figure 9a–c). Biplot analysis of the principal component analysis plots showed that IL13, IL4, IL31, and IL22 were the drivers of clustering along the AD principal component PC2, whereas NOS2, IL17A, IL17F, IFNG, IL23A, and IL12B were the drivers of clustering along the psoriasis principal component PC1.
      Figure thumbnail gr9
      Figure 9Psoriasis and AD cases can be classified on the basis of cytokine expression patterns determined by RISH. (a) PCA of psoriasis, AD, and normal controls based on RISH staining pattern for IL17A, IL17F, IL12B, IL23A, NOS2, IL13, and IL4 in the epidermis. (b) Biplot of the PCA data for psoriasis drivers. (c) Biplot of the PCA data for AD drivers. AD, atopic dermatitis; PCA, principal component analysis; RISH, RNA in situ hybridization.

      Discussion

      In the past several years, there has been a revolution in the molecularly directed treatment of inflammatory skin diseases, including psoriasis and AD. However, clinically implementable and actionable molecular diagnostics in this area have lagged, particularly as they might relate to personalized treatment selection. A trial-and-error approach to biologic treatment remains standard of care; however, such an approach (i) is inefficient, (ii) is expensive, (iii) is potentially anxiety inducing to patients because treatment outcomes are unpredictable, and (iv) may be accompanied by potentially avoidable adverse effects. In addition, some patients may remain on a medication to which they have a suboptimal response when a better alternative in a different class may be available (e.g., IL-12/23p40 vs. IL-17A vs. IL-23 inhibitor in psoriasis).
      In this study, we show that RISH for disease-causing cytokines is a specific, feasible approach that can identify pathologic cytokines in psoriasis and AD, with IL17A and IL13 appearing to provide the most information. RISH is also an approach that can be easily and cost-efficiently implemented in dermatopathology laboratories because its workflow is analogous to that of IHC. The results are also rapid, making it conceivable that data obtained through RISH-based analyses could be incorporated into real-time clinical decision making. In contrast, RNA-seq is expensive, requires specialized tissue collection and processing (including data normalization), and can take weeks or more.
      We observed that there is variability within both psoriasis and AD in terms of the predominant druggable cytokine expressed. In psoriasis, most cases were IL17A predominant, but others were instead IL17F, IL23A, and/or IL12B predominant. Psoriasis cases demonstrated relatively less molecular immunologic heterogeneity than AD cases, consistent with previous observations and with the excellent responses observed with IL-17 and IL-23 inhibitors in most patients with psoriasis. How this molecular heterogeneity relates to differences in response to treatment and potentially to the development of paradoxical eruptions when a particular cytokine is inhibited will be an area of significant interest moving forward.
      Interestingly, most IL17 mRNA expression appeared to be within the epidermis in psoriasis, a pattern observed both in the RISH staining and in the scRNA-seq data. These findings were somewhat surprising because they are not necessarily in agreement with previous studies looking at cytokine expression patterns in psoriasis in tissue sections, which have reported mostly a dermal expression of these cytokines using IHC and/or immunofluorescence (
      • Johansen C.
      • Vinter H.
      • Soegaard-Madsen L.
      • Olsen L.R.
      • Steiniche T.
      • Iversen L.
      • et al.
      Preferential inhibition of the mRNA expression of p38 mitogen-activated protein kinase regulated cytokines in psoriatic skin by anti-TNFα therapy.
      ;
      • Miyagaki T.
      • Sugaya M.
      • Suga H.
      • Kamata M.
      • Ohmatsu H.
      • Fujita H.
      • et al.
      IL-22, but not IL-17, dominant environment in cutaneous T-cell lymphoma.
      ;
      • Zhang L.
      • Yang X.Q.
      • Cheng J.
      • Hui R.S.
      • Gao T.W.
      Increased Th17 cells are accompanied by FoxP3(+) Treg cell accumulation and correlated with psoriasis disease severity.
      ). Additional study will be needed to reconcile these observations, which may relate to differences in technique and detection of mRNA versus of protein.
      We also somewhat surprisingly found that most of the IL12B and IL23A production in psoriasis, as detected by RISH, was within the epidermis compared with that in the dermis. In the scRNA-seq dataset, IL12B expression was predominantly epidermal, whereas IL23A was produced in both the epidermis and dermis (
      • Reynolds G.
      • Vegh P.
      • Fletcher J.
      • Poyner E.F.M.
      • Stephenson E.
      • Goh I.
      • et al.
      Developmental cell programs are co-opted in inflammatory skin disease.
      ). Previous studies have also found the expression of these cytokines by DCs that either enter the epidermis or reside in the epidermis (e.g., LCs) (
      • Martini E.
      • Wikén M.
      • Cheuk S.
      • Gallais Sérézal I.
      • Baharom F.
      • Ståhle M.
      • et al.
      Dynamic changes in resident and infiltrating epidermal dendritic cells in active and resolved psoriasis.
      ;
      • Nakajima K.
      • Kataoka S.
      • Sato K.
      • Takaishi M.
      • Yamamoto M.
      • Nakajima H.
      • et al.
      Stat3 activation in epidermal keratinocytes induces Langerhans cell activation to form an essential circuit for psoriasis via IL-23 production.
      ), whereas other studies have found mostly dermal expression. There was a negligible expression by KCs. As with IL-17, additional study will be needed to reconcile the differences in location of cytokine production observed in different studies with different techniques.
      In AD, although most cases were IL13 predominant, some cases of AD did not show significant expression of either IL13 or IL4, and one case was IL17F predominant. Furthermore, there was relatively more or less IL22 and IL31 staining in individual cases. Of note, in occasional cases, IL13 was not detected at all; whether this represents biology or instead a technical or sampling limitation remains to be determined.
      Considerable clinical heterogeneity exists in AD and this may be driven, in part, by underlying immunologic heterogeneity (
      • Czarnowicki T.
      • He H.
      • Krueger J.G.
      • Guttman-Yassky E.
      Atopic dermatitis endotypes and implications for targeted therapeutics.
      ;
      • Tsoi L.C.
      • Rodriguez E.
      • Degenhardt F.
      • Baurecht H.
      • Wehkamp U.
      • Volks N.
      • et al.
      Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
      ). Along these lines, some groups have even divided AD into immunologic endotypes, given the observation that IL-17 may play a more prominent role in children with AD and in patients with AD who are of Asian descent and that IL-22 may play a more prominent role in some African Americans with AD (
      • Czarnowicki T.
      • He H.
      • Krueger J.G.
      • Guttman-Yassky E.
      Atopic dermatitis endotypes and implications for targeted therapeutics.
      ). Overall, RISH may provide a practical means by which to dissect this molecular heterogeneity in a quantifiable and practical fashion. We hypothesize that patients with purer IL13‒IL4 dysregulation may respond best to the blockade of this axis with dupilumab, whereas those with mixed or other immunologic drivers may respond suboptimally and may theoretically respond better to blockade of other cytokines or to more broadly acting cytokine blockers, such as Jak inhibitors. Future studies will be required to test this hypothesis.
      Limitations of this study include a relatively small sample size, a retrospective design, and the inability to correlate RISH cytokine profiles with treatment response to biologics. In the future, it would also be useful to see how well RISH patterns from different lesions in individual patients correlate with each other in space and time. The patients with AD in the RISH cohort were also older, in general, and may not be fully representative of all patients with AD. Prospective studies of patients starting biologic therapies are needed. The cost of the assay is also a consideration. The cost of each RISH stain is roughly comparable with that of an IHC stain, and so it is unlikely to be practical to perform large cytokine panels on individual cases in clinical practice; however, one might envision scenarios where smaller, more targeted panels could be useful.
      This study shows the feasibility of RISH for the detection of cytokines in inflammatory skin diseases. RISH for cytokines could be readily expanded to additional cytokine targets and could be further applied to these and other inflammatory skin diseases. Minimal background staining is observed with the approach, making it very straightforward to quantify and suggesting that reproducibility among different laboratories is likely to be high. Overall, we predict that this approach will be a useful and efficient tool to molecularly phenotype inflammatory skin diseases with implications for both research and clinical care, including personalized treatment selection.

      Materials and Methods

      Retrospective case series

      Cases were selected after searching the Yale Dermatopathology (New Haven, CT) database for archival material. A total of 20 cases of psoriasis, 26 cases of AD, and 10 cases of normal skin from healthy controls were selected. For psoriasis and AD, biopsies were of lesional skin. For psoriasis, cases were only included if they had classic clinical and histologic features, as determined by board-certified and experienced dermatologists and dermatopathologists. Medical chart review, including a review of clinical photographs to confirm the clinical diagnosis, was performed as needed. AD cases showed typical features of subacute-to-chronic spongiotic dermatitis histopathologically, and the clinical presentation was typical of AD. The clinical diagnosis of AD was made by experienced dermatologists. The clinical presentation was further verified using medical record and clinical photograph review. The Hanifin‒Rajka criteria were utilized (
      • Hanifin J.M.
      • Rajka G.
      Diagnostic features of atopic dermatitis.
      ). Only cases that could be confidently classified as AD on the basis of these criteria were included. The vast majority of biopsies were from adults reflecting biopsy acquisition patterns in AD at our institution. Normal skin from excision tips of nonchronically sun-exposed sites was included as controls. This work was reviewed and approved by the Yale Institutional Review Board.

      RISH

      RISH was performed using the RNA scope kit (Bio-Techne, Minneapolis, MN) following the manufacturer’s instructions. Briefly, slides were deparaffinized and then treated with hydrogen peroxide and protease from the RNA scope kit (catalog [Cat] #322330). Antigen retrieval was performed with RNA scope target retrieval reagent (Cat #322000). Probe hybridization and amplification steps were then performed according to the manufacturer’s instructions. Slides were counterstained with hematoxylin and coverslipped. Probes for IL17A (Cat #310931), IL17F (Cat #310941), IL12B (IL-12 p40) (Cat #402071), IL23A (IL-23 p19) (Cat #562851), IL4 (Cat #315191), IL13 (Cat #586241), IL22 (Cat #560811), IL31 (Cat #436751), TNF (TNF-α) (cat # 310421), IFNG (310501), and NOS2 (Cat #424991) were purchased from Bio-Techne. Staining with a positive control probe, for PPIB (Cat# 313901), a housekeeping gene (
      • Nazet U.
      • Schröder A.
      • Grässel S.
      • Muschter D.
      • Proff P.
      • Kirschneck C.
      Housekeeping gene validation for RT-qPCR studies on synovial fibroblasts derived from healthy and osteoarthritic patients with focus on mechanical loading.
      ;
      • Pachot A.
      • Blond J.L.
      • Mougin B.
      • Miossec P.
      Peptidylpropyl isomerase B (PPIB): a suitable reference gene for mRNA quantification in peripheral whole blood.
      ), was performed on five cases each of psoriasis, AD, and normal skin to ensure equal staining of cells in epidermis and dermis (data not shown). Staining with a negative control probe, DapB (Cat #310043), an E. coli gene, was performed in a subset of cases and showed no staining (data not shown).
      The CD3 IHC was performed using an anti-CD3ε antibody (D7A6E) acquired from Cell Signaling Technology (Danvers, MA) (Cat #85061). Double staining was performed using the RNA scope kit (Bio-Techne) as described earlier, followed by IHC using standard methods. Briefly, after in situ hybridization signal detection, antigen retrieval was performed with citrate buffer (pH 6.0) (Cat #00500, Thermo Fisher Scientific, Waltham, MA). Peroxidase activity was quenched using 3% hydrogen peroxide (Cat# JT-2186-01, JT Baker, Phillipsburg, NJ). The primary antibody was incubated and then detected using a species-specific secondary antibody and the ImmPRESS horseradish peroxidase reagent (Vector Laboratories, Burlingame, CA) and diaminobenzidine substrate (Vector Laboratories). Slides were counterstained with hematoxylin and were coverslipped.

      Quantification of RISH and IHC staining

      Slides were scanned using a Hamamatsu Nano Zoomer S210 (Yale Dermatopathology). For certain analyses, QuPath (version 0.2.3) (
      • Bankhead P.
      • Loughrey M.B.
      • Fernandez J.A.
      • Dombrowski Y.
      • McArt D.G.
      • Dunne P.D.
      • et al.
      QuPath: open source software for digital pathology image analysis.
      ) and ImageJ (Fiji version 2.10) were utilized to assist with quantification (
      • Doube M.
      • Kłosowski M.M.
      • Arganda-Carreras I.
      • Cordelières F.P.
      • Dougherty R.P.
      • Jackson J.S.
      • et al.
      BoneJ: free and extensible bone image analysis in ImageJ.
      ). The number of positive cells in each specimen was quantified separately for the epidermal and dermal portions. The number of positive cells was divided by the width of the biopsy in millimeters to correct for differences in biopsy size (cells per mm). For a cell to be counted as positive, staining needed to be clearly apparent with the ×20 objective, and at least two dots per cell had to be present (most positive cells had >2 dots). Faint single dots apparent only with the ×40 objective were not counted as positive. The slides were scored in a blinded fashion by consensus (AW and WD). Although the diagnosis was not considered during the scoring, morphologic features in some cases could not be entirely ignored.
      For singly stained slides, the images were imported into QuPath. A region of interest that covered the area of interest was selected. Stain vectors were defined by selecting a region of interest that was representative of the stain. The positive cell detection feature was utilized to identify the positive cells for each marker. The automated detection was manually verified, and in some cases, adjustments were made manually. To facilitate visualization, cells were colored on the basis of whether or not they were positive for the marker.
      For double-stained slides, the images were imported into QuPath, and the color vectors were defined using the region of interest as described earlier. The he_heavy_augement model, a brightfield image nucleus StarDist (arXiv: 1806.03535) approach, was then used in QuPath to detect cell nuclei. After nucleus detection using this approach, the images were then imported into Fiji for color deconvolution on the basis of the stain vectors defined in QuPath. Color deconvolution allowed the separation of the different channels from each stain. Brightness and contrast for each channel were then adjusted (these were done manually) to maximize signal to background ratio using Fiji. The deconvoluted, optimized images were then reoverlaid using Photoshop (version 21.2.0).

      RNA-seq analysis

      A previously published, bulk RNA-seq dataset was utilized (
      • Tsoi L.C.
      • Rodriguez E.
      • Degenhardt F.
      • Baurecht H.
      • Wehkamp U.
      • Volks N.
      • et al.
      Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
      ). Read counts were downloaded from Gene Expression Omnibus (GSE121212). Data analysis was performed using Partek Flow software (version 9.0, build 9.0.20.0510) (Partek, Inc, St. Louis, MI). AD and psoriasis lesional skin and healthy control samples were extracted from the rest of the data set (which also included nonlesional skin). Read counts for these samples were extracted to generate bar graphs using tidyr (version 1.1.0), ggplot2 (version 3.3.2), and data.table (version 1.12.8). The entire dataset was normalized in Partek Flow. DESeq2 was used for differential gene expression analysis to compare psoriasis samples with AD samples in the dataset. The Volcano plot was generated using dplyr (version 1.0.0), ggplot2 (version 3.3.2), and ggrepel (version 0.8.2). The principal component analysis plot was generated by Factoextra (1.0.7) and FactoMineR (2.3) for the manually quantified results.

      scRNA-seq analysis

      A recently published scRNA-seq dataset comprising AD, psoriasis, and healthy control skin samples (
      • Reynolds G.
      • Vegh P.
      • Fletcher J.
      • Poyner E.F.M.
      • Stephenson E.
      • Goh I.
      • et al.
      Developmental cell programs are co-opted in inflammatory skin disease.
      ) was analyzed to examine the patterns of cytokine expression. In this study, the epidermal and dermal portions of the tissue were separated and then dissociated and analyzed separately. The data from this study are freely available, and we downloaded them from https://zenodo.org/record/4569496#.YE9kGLRKi-V. We focused our analysis of these data on lesional AD, lesional psoriasis, and healthy control libraries, which were reanalyzed in Seurat (version 3.2.0). Nonlesional AD and psoriasis skin libraries were not included in our analyses. T cells, myeloid cells, and KCs were subsetted on the basis of the cell-type assignments from the original publication, as determined by AutoGeneS, and were then reanalyzed. We retained these cell-type identities for the subsetting because certain analyses groups were combined, for example, T cells (Tc, Tc17_T helper type 17, Tc_IL13_IL22, T helper, regulatory T cell), ILCs (ILC1_NK, ILC2, ILC3), macrophages (Inf_Mac, Macro_1, Macro_2, Mono_Mac), DCs (DC1, DC2, MigDC, moDC_1, moDC_2, moDC_3), LC (LC_1, LC_2, LC_3, LC_4), and KCs (differentiated, proliferating KC, undifferentiated KC). In total, we included 127,219 cells in our analysis of the data. The NormalizeData command with a scale factor of 10,000 was used to normalize counts. The data were then further scaled on the basis of the number of transcripts and center gene expression values. Cells were clustered using FindNeighbors and FindClusters commands. The data were visualized by performing t-Distributed Stochastic Neighbor Embedding. The FeaturePlot command was used to generate gene-specific t-Distributed Stochastic Neighbor Embedding plots, and DotPlot command was used to generate dot plots.

      Statistical analyses

      Statistical analyses were performed using GraphPad Prism (version 9.0.0) (GraphPad, San Diego, CA). P-values were determined for pairwise comparisons using Student’s unpaired t-tests. R2 correlation values were calculated in Prism. Bar graphs were created in Prism.

      Data availability statement

      No -omics data were generated as a part of this work. Accession numbers for previously generated -omics data are included. Raw data generated from the study are available on reasonable request.

      ORCIDs

      Author Contributions

      Conceptualization: WD; Data Curation: AW, WD; Formal Analysis: AW, MJM, MKM, WD; Investigation: AW, WD; Methodology: AW, MJM, MKM, CJK, WD; Resources: GP, JMM, MB, BAK, WD; Software: AW, MKM; Supervision: WD; Visualization: AW, WD; Writing - Original Draft Preparation: AW, ALF, MDV, JMC, CJK, WD; Writing - Review and Editing: AW, CJK, WD

      Acknowledgments

      We would like to thank the staff of Yale Dermatopathology laboratory (New Haven, CT), particularly D. Mekael. WD is supported by a Career Development Award from the Dermatology Foundation. BAK is supported by the Ranjini and Ajay Poddar Fund for Dermatologic Disease Research. This study has been approved by the Yale Institutional Review Board.

      Conflict of Interest

      MDV’s spouse is employed by Regeneron. BAK is an investigator for Concert Pharmaceuticals, Eli Lilly and Company, and Pfizer. BAK is a consultant to and/or has served on advisory boards for Aclaris Therapeutics, Arena Pharmaceuticals, Bristol-Meyers Squibb, Concert Pharmaceuticals, Dermavant Sciences, Eli Lilly and Company, Pfizer, and VielaBio; he is on speaker’s bureau for Pfizer, Regeneron, and Sanofi Genzyme. WD has research support from Pfizer and is a consultant for TWi Biotechnology. WD and MB are consultants for Eli Lilly and receive licensing fees from EMD, Sigma, and Millipore in unrelated work. The remaining authors state no conflict of interest.

      References

        • Bai F.
        • Li G.G.
        • Liu Q.
        • Niu X.
        • Li R.
        • Ma H.
        Short-term efficacy and safety of IL-17, IL-12/23, and IL-23 inhibitors brodalumab, secukinumab, ixekizumab, ustekinumab, guselkumab, tildrakizumab, and risankizumab for the treatment of moderate to severe plaque psoriasis: a systematic review and network meta-analysis of randomized controlled trials.
        J Immunol Res. 2019; 2019: 2546161
        • Bankhead P.
        • Loughrey M.B.
        • Fernandez J.A.
        • Dombrowski Y.
        • McArt D.G.
        • Dunne P.D.
        • et al.
        QuPath: open source software for digital pathology image analysis.
        Sci Rep. 2017; 7: 16878
        • Bieber T.
        Interleukin-13: targeting an underestimated cytokine in atopic dermatitis.
        Allergy. 2020; 75: 54-62
        • Brunner P.M.
        • Pavel A.B.
        • Khattri S.
        • Leonard A.
        • Malik K.
        • Rose S.
        • et al.
        Baseline IL-22 expression in patients with atopic dermatitis stratifies tissue responses to fezakinumab.
        J Allergy Clin Immunol. 2019; 143: 142-154
        • Caproni M.
        • Torchia D.
        • Antiga E.
        • Terranova M.
        • Volpi W.
        • del Bianco E.
        • et al.
        The comparative effects of tacrolimus and hydrocortisone in adult atopic dermatitis: an immunohistochemical study.
        Br J Dermatol. 2007; 156: 312-319
        • Cohen J.N.
        • Bowman S.
        • Laszik Z.G.
        • North J.P.
        Clinicopathologic overlap of psoriasis, eczema, and psoriasiform dermatoses: a retrospective study of T helper type 2 and 17 subsets, interleukin 36, and β-defensin 2 in spongiotic psoriasiform dermatitis, sebopsoriasis, and tumor necrosis factor α inhibitor-associated dermatitis.
        J Am Acad Dermatol. 2020; 82: 430-439
        • Czarnowicki T.
        • He H.
        • Krueger J.G.
        • Guttman-Yassky E.
        Atopic dermatitis endotypes and implications for targeted therapeutics.
        J Allergy Clin Immunol. 2019; 143: 1-11
        • Doube M.
        • Kłosowski M.M.
        • Arganda-Carreras I.
        • Cordelières F.P.
        • Dougherty R.P.
        • Jackson J.S.
        • et al.
        BoneJ: free and extensible bone image analysis in ImageJ.
        Bone. 2010; 47: 1076-1079
        • Garzorz-Stark N.
        • Krause L.
        • Lauffer F.
        • Atenhan A.
        • Thomas J.
        • Stark S.P.
        • et al.
        A novel molecular disease classifier for psoriasis and eczema.
        Exp Dermatol. 2016; 25: 767-774
        • Glickman J.W.
        • Han J.
        • Garcet S.
        • Krueger J.G.
        • Pavel A.B.
        • Guttman-Yassky E.
        Improving evaluation of drugs in atopic dermatitis by combining clinical and molecular measures.
        J Allergy Clin Immunol Pract. 2020; 8: 3622-3625.e19
        • Hanifin J.M.
        • Rajka G.
        Diagnostic features of atopic dermatitis.
        Acta Derm Venereol Suppl (Stockh). 1980; 92: 44-47
        • Hawkes J.E.
        • Yan B.Y.
        • Chan T.C.
        • Krueger J.G.
        Discovery of the IL-23/IL-17 signaling pathway and the treatment of psoriasis.
        J Immunol. 2018; 201: 1605-1613
        • He H.
        • Bissonnette R.
        • Wu J.
        • Diaz A.
        • Saint-Cyr Proulx E.
        • Maari C.
        • et al.
        Tape strips detect distinct immune and barrier profiles in atopic dermatitis and psoriasis.
        J Allergy Clin Immunol. 2021; 147: 199-212
        • Johansen C.
        • Vinter H.
        • Soegaard-Madsen L.
        • Olsen L.R.
        • Steiniche T.
        • Iversen L.
        • et al.
        Preferential inhibition of the mRNA expression of p38 mitogen-activated protein kinase regulated cytokines in psoriatic skin by anti-TNFα therapy.
        Br J Dermatol. 2010; 163: 1194-1204
        • Khattri S.
        • Brunner P.M.
        • Garcet S.
        • Finney R.
        • Cohen S.R.
        • Oliva M.
        • et al.
        Efficacy and safety of ustekinumab treatment in adults with moderate-to-severe atopic dermatitis.
        Exp Dermatol. 2017; 26: 28-35
        • Kimmel G.
        • Chima M.
        • Kim H.J.
        • Bares J.
        • Yao C.J.
        • Singer G.
        • et al.
        Brodalumab in the treatment of moderate to severe psoriasis in patients when previous anti-interleukin 17A therapies have failed.
        J Am Acad Dermatol. 2019; 81: 857-859
        • Martini E.
        • Wikén M.
        • Cheuk S.
        • Gallais Sérézal I.
        • Baharom F.
        • Ståhle M.
        • et al.
        Dynamic changes in resident and infiltrating epidermal dendritic cells in active and resolved psoriasis.
        J Invest Dermatol. 2017; 137: 865-873
        • McGeachy M.J.
        • Cua D.J.
        • Gaffen S.L.
        The IL-17 family of cytokines in health and disease.
        Immunity. 2019; 50: 892-906
        • Miranda E.
        • Roberts J.
        • Novick S.
        • Lapointe J.M.
        • Bruijnzeel-Koomen C.
        • Thijs J.
        • et al.
        Immunohistochemical characterization of the IL-13:IL-4 receptor α axis in the skin of adult patients with moderate to severe atopic dermatitis and healthy controls.
        J Invest Dermatol. 2021; 141: 440-443.e4
        • Miyagaki T.
        • Sugaya M.
        • Suga H.
        • Kamata M.
        • Ohmatsu H.
        • Fujita H.
        • et al.
        IL-22, but not IL-17, dominant environment in cutaneous T-cell lymphoma.
        Clin Cancer Res. 2011; 17: 7529-7538
        • Moy A.P.
        • Murali M.
        • Kroshinsky D.
        • Duncan L.M.
        • Nazarian R.M.
        Immunologic overlap of helper T-cell Subtypes 17 and 22 in erythrodermic psoriasis and atopic dermatitis.
        JAMA Dermatol. 2015; 151: 753-760
        • Nakajima K.
        • Kataoka S.
        • Sato K.
        • Takaishi M.
        • Yamamoto M.
        • Nakajima H.
        • et al.
        Stat3 activation in epidermal keratinocytes induces Langerhans cell activation to form an essential circuit for psoriasis via IL-23 production.
        J Dermatol Sci. 2019; 93: 82-91
        • Nazet U.
        • Schröder A.
        • Grässel S.
        • Muschter D.
        • Proff P.
        • Kirschneck C.
        Housekeeping gene validation for RT-qPCR studies on synovial fibroblasts derived from healthy and osteoarthritic patients with focus on mechanical loading.
        PLoS One. 2019; 14e0225790
        • Pachot A.
        • Blond J.L.
        • Mougin B.
        • Miossec P.
        Peptidylpropyl isomerase B (PPIB): a suitable reference gene for mRNA quantification in peripheral whole blood.
        J Biotechnol. 2004; 114: 121-124
        • Quaranta M.
        • Knapp B.
        • Garzorz N.
        • Mattii M.
        • Pullabhatla V.
        • Pennino D.
        • et al.
        Intraindividual genome expression analysis reveals a specific molecular signature of psoriasis and eczema.
        Sci Transl Med. 2014; 6: 244ra90
        • Reynolds G.
        • Vegh P.
        • Fletcher J.
        • Poyner E.F.M.
        • Stephenson E.
        • Goh I.
        • et al.
        Developmental cell programs are co-opted in inflammatory skin disease.
        Science. 2021; 371eaba6500
        • Salerno F.
        • Engels S.
        • van den Biggelaar M.
        • van Alphen F.P.J.
        • Guislain A.
        • Zhao W.
        • et al.
        Translational repression of pre-formed cytokine-encoding mRNA prevents chronic activation of memory T cells.
        Nat Immunol. 2018; 19: 828-837
        • Shroff A.
        • Guttman-Yassky E.
        Successful use of ustekinumab therapy in refractory severe atopic dermatitis.
        JAAD Case Rep. 2014; 1: 25-26
        • Tsoi L.C.
        • Rodriguez E.
        • Degenhardt F.
        • Baurecht H.
        • Wehkamp U.
        • Volks N.
        • et al.
        Atopic dermatitis is an IL-13-dominant disease with greater molecular heterogeneity compared to psoriasis.
        J Invest Dermatol. 2019; 139: 1480-1489
        • Ungar B.
        • Pavel A.B.
        • Li R.
        • Kimmel G.
        • Nia J.
        • Hashim P.
        • et al.
        Phase 2 randomized, double-blind study of IL-17 targeting with secukinumab in atopic dermatitis.
        J Allergy Clin Immunol. 2021; 147: 394-397
        • Wang F.P.
        • Tang X.J.
        • Wei C.Q.
        • Xu L.R.
        • Mao H.
        • Luo F.M.
        Dupilumab treatment in moderate-to-severe atopic dermatitis: a systematic review and meta-analysis.
        J Dermatol Sci. 2018; 90: 190-198
        • Yawalkar N.
        • Karlen S.
        • Egli F.
        • Brand C.U.
        • Graber H.U.
        • Pichler W.J.
        • et al.
        Down-regulation of IL-12 by topical corticosteroids in chronic atopic dermatitis.
        J Allergy Clin Immunol. 2000; 106: 941-947
        • Zhang L.
        • Yang X.Q.
        • Cheng J.
        • Hui R.S.
        • Gao T.W.
        Increased Th17 cells are accompanied by FoxP3(+) Treg cell accumulation and correlated with psoriasis disease severity.
        Clin Immunol. 2010; 135: 108-117
        • Zheng Y.
        • Danilenko D.M.
        • Valdez P.
        • Kasman I.
        • Eastham-Anderson J.
        • Wu J.
        • et al.
        Interleukin-22, a T(H)17 cytokine, mediates IL-23-induced dermal inflammation and acanthosis.
        Nature. 2007; 445: 648-651