Atopic dermatitis and inﬂammatory skin disease Single-cell analysis implicates T H 17-to-T H 2 cell plasticity in the pathogenesis of palmoplantar pustulosis

Background: Palmoplantar pustulosis (PPP) is a severe inﬂammatory skin disorder characterized by eruptions of painful, neutrophil-ﬁlled pustules on the palms and soles. Although PPP has a profound effect on quality of life, it remains poorly understood and notoriously difﬁcult to treat. Objective: We sought to investigate the immune pathways that underlie the pathogenesis of PPP. Methods: We applied bulk and single-cell RNA sequencing (RNA-Seq) methods to the analysis of skin biopsy samples and peripheral blood mononuclear cells. We validated our results by ﬂow cytometry and immune ﬂuorescence microscopy Results: Bulk RNA-Seq of patient skin detected an unexpected signature of T-cell activation, with a signiﬁcant overexpression of several T H 2 genes typically upregulated in atopic dermatitis. To further explore these ﬁndings, we carried out single-cell RNA-Seq in peripheral blood mononuclear cells of healthy and affected individuals. Memory CD4 1 T cells of PPP patients were skewed toward a T H 17 phenotype, a phenomenon that was particularly signiﬁcant among cutaneous lymphocyte-associated antigen–positive skin-homing cells. We also identiﬁed a subset of memory CD4 1 T cells that expressed both T H 17 ( KLRB1 / CD161) and T H 2 ( GATA3 ) markers, with pseudotime analysis suggesting that the population was the result of T H 17 to T H 2 plasticity. Interestingly, the GATA3 1 /CD161 1 cells were overrepresented among the peripheral blood mononuclear cells of affected individuals, both in the single-cell RNA-Seq data set and in independent ﬂow cytometry experiments. Dual-positive cells were also detected in patient skin by immune ﬂuorescence microscopy. Conclusions: PPP is associated with complex T-cell activation patterns and may explain why biologic drugs that target individual T helper cell populations have shown limited therapeutic efﬁcacy. (J Allergy Clin Immunol 2022;150:882-93.)

Objective: We sought to investigate the immune pathways that underlie the pathogenesis of PPP. Methods: We applied bulk and single-cell RNA sequencing (RNA-Seq) methods to the analysis of skin biopsy samples and peripheral blood mononuclear cells. We validated our results by flow cytometry and immune fluorescence microscopy Results: Bulk RNA-Seq of patient skin detected an unexpected signature of T-cell activation, with a significant overexpression of several T H 2 genes typically upregulated in atopic dermatitis. To further explore these findings, we carried out single-cell RNA-Seq in peripheral blood mononuclear cells of healthy and affected individuals. Memory CD4 1 T cells of PPP patients were skewed toward a T H 17 phenotype, a phenomenon that was particularly significant among cutaneous lymphocyte-associated antigen-positive skin-homing cells. We also identified a subset of memory CD4 1 T cells that expressed both T H 17 (KLRB1/ CD161) and T H 2 (GATA3) markers, with pseudotime analysis suggesting that the population was the result of T H 17 to T H 2 plasticity. Interestingly, the GATA3 1 /CD161 1 cells were overrepresented among the peripheral blood mononuclear cells of affected individuals, both in the single-cell RNA-Seq data set and in independent flow cytometry experiments. Dual-positive cells were also detected in patient skin by immune fluorescence microscopy. Conclusions: PPP is associated with complex T-cell activation patterns and may explain why biologic drugs that target individual T helper cell populations have shown limited therapeutic efficacy. (J Allergy Clin Immunol 2022;150:882-93.) Key words: Single-cell RNA sequencing, scRNA-Seq, T-cell plasticity, palmoplantar pustulosis, PPP Palmoplantar pustulosis (PPP) is a chronic and debilitating skin disorder that manifests as the eruption of neutrophil-filled pustules on the palms and soles. These painful lesions typically occur on a background of inflamed skin, causing functional and occupational disability. 1 PPP has a profound impact on quality of life, but its causes remain poorly understood. The disease preferentially affects adult female subjects and is associated with cigarette smoking. 2 However, the mechanisms mediating the effects of sex and tobacco exposure are unclear. 3 Although it has been suggested that PPP shares common genetic determinants with other pustular diseases, IL36RN mutations (which are frequently observed in generalized pustular psoriasis) have only been reported in a small number of cases. 2,4 As a result of this limited understanding of disease pathogenesis, evidence-based guidelines for the management of PPP are lacking. 5 The response to conventional systemic therapeutics (oral retinoids, methotrexate, and cyclosporine) is variable, and their prolonged use can have toxic effects. 5 Clinical trials of IL-1 (anakinra) and IL-36 (spesolimab) blockers have been carried out on the assumption that PPP has an autoinflammatory pathogenesis, but the studies undertaken so far could not provide evidence of broad clinical efficacy. 6,7 IL-17 (secukinumab) and IL-23 (guselkumab) inhibitors have also been assessed. Although these biologics reduced disease severity, skin clearance was observed in <30% of patients. 8,9 We therefore sought to identify disease drivers and potential therapeutic targets for PPP via transcription profiling of patient cells. We uncovered a complex immunologic landscape, where T H 2 cell activation dominates in skin while circulating T cells are skewed toward a T H 17 phenotype. We also observed evidence of increased T H 17-to-T H 2 plasticity in the circulating and skinhoming T lymphocytes of affected individuals. These findings point to the activation of diverse T helper cell populations in PPP and warrant the investigation of small-molecule therapeutics that can inhibit multiple signaling pathways.

Study participants
This work was carried out in accordance with the principles of the declaration of Helsinki and after receipt of written informed consent from all participants. PPP was diagnosed on the bases of the results of clinical examination and the consensus criteria of the European Rare and Severe Psoriasis Expert Network, or ERASPEN. 10  Sampling and RNA sequencing of skin biopsy samples Two-millimeter acral skin biopsy samples were obtained from healthy controls (HCs) or APRICOT trial participants. Patients were recruited at their baseline visit after a recommended washout period and before treatment initiation. 6 Lesional biopsy samples were taken from inflamed skin (avoiding pustules), while nonlesional (NL) samples were taken via biopsy from adjacent uninvolved skin. Total RNA was extracted using a miRVana isolation kit (Thermo Fisher Scientific, Waltham, Mass). After poly-A selection and library preparation, samples were run on an Illumina HiSeq instrument (Illumina, San Diego, Calif) to generate 150 bp paired-end reads.

Single-cell RNA sequencing
Peripheral blood mononuclear cells (PBMCs) were resuspended in fetal calf serum (Invitrogen; Thermo Fisher Scientific)/10% dimethyl sulfoxide and stored in liquid nitrogen for up to 4 weeks. On the day of the experiment, cells were thawed, counted, and loaded on a Chromium Single Cell 39 Chip (103 Genomics, Pleasanton, Calif), as described elsewhere. 11 Libraries were prepared using the Single Cell 39 Reagent Kits v3 (103 Genomics) and sequenced on a HiSeq4000 instrument (Illumina).

Data analysis by single-cell RNA sequencing
Sequence reads were processed, aligned to the GRCh38 reference genome, and annotated to Ensembl (release 86) genes by Cell Ranger v3.0.2 software (103 Genomics). The healthy donor data sets published by Zheng et al 12 (n 5 3) and Schafflick et al 13 (n 5 5) were retrieved from the 103 Genomics portal (support.10xgenomics.com/single-cell-gene-expression/datasets) and the Gene Expression Omnibus repository (www.ncbi.nlm.nih.gov/geo; identifier GSE138266), respectively. The 3 data sets were then merged using Harmony 14 to correct for batch effects. The resulting gene expression matrix was processed with Seurat v3.0. 15 Quality control filters were first applied to remove cells with low (<300) or excessive (>5000) numbers of detected genes. Cells where the percentage of mitochondrial gene reads exceeded 20% were also excluded. After log normalization and data scaling, variation due to the following sources was regressed out: sequencing batch, data origin (Zheng et al, Schafflick et al, and generated in house), smoking status, treatment with biologic agents, and unique molecular identifier.
After principal component analysis and construction of a K-nearest neighbor graph, unsupervised clustering was undertaken with a resolution of 0.4. The resulting cell clusters were visualized by uniform manifold approximation and projection (UMAP). Cluster markers were computed with the FindAllMarkers Seurat function, and cell identities were annotated on the basis of the expression of canonical marker genes. SingleR 16 was used to validate cell identities and to annotate the phenotypes of memory CD4 1 T cells as T H 1, T H 2, or T H 17. The full resource published by Monaco et al 17 was used as a reference data set.

Statistical analysis
Cell abundance and gene expression levels were compared in cases versus controls by the Mann-Whitney test. The significance of overlaps observed in Venn diagrams was computed by the Fisher exact test. All tests were implemented in R v4.02 (www.r-project.org). P < .05 was considered statistically significant.

A prominent T H 2 signature in NL-PPP skin
To explore the immune pathways that are disrupted in PPP, we first carried out bulk RNA sequencing (RNA-Seq) in 3 paired, lesional and NL skin biopsy samples, obtained from the palmar or plantar (acral) skin of affected individuals (Table E1). We identified a total of 1050 differentially expressed genes (log 2 fold change [FC] > |0.5|; false discovery rate [FDR] < 0.05) (see Fig  E1, A, and Table E2, A, in the Online Repository at www. jacionline.org). In keeping with the neutrophilic nature of PPP lesions, these showed a significant enrichment for innate pathways (eg, ''granulocyte adhesion and diapedesis'' and ''IL-8 signaling''; FDR < 10 23 for both) (Fig E1, A).
We obtained similar results when we compared the 3 lesional samples with healthy acral skin donated by 7 volunteers matched for age, sex, and smoking status (Table E1). In fact, we identified 1323 differentially expressed genes showing a very marked enrichment for innate pathways (eg, ''granulocyte adhesion and diapedesis'', FDR < 10 28 ). At the same time, we also uncovered an unexpected overrepresentation of T-cell-related genes (eg, ''T-cell receptor signaling'', FDR < 10 26 ) (Fig E1, B).
To further explore these findings while avoiding the confounding effects of end-stage inflammation (ie, the secondary upregulation of genes that do not contribute to disease pathogenesis), we next compared NL-PPP biopsy samples (n 5 8) with healthy skin (n 5 7). We observed 531 differentially expressed genes (Fig 1, A, Fig  E1, C, and Table E2, B). In keeping with the results of genetic studies and clinical trials, 2,6 we found limited evidence for a sustained upregulation of IL-36 signaling. While IL36A (encoding IL-36a) was overexpressed in NL compared to control skin, the mRNA levels of IL36B, IL36G, and IL1RL2 (encoding, respectively, IL-36b, IL-36g, and IL-36R) were comparable in the 2 groups.
To further explore the significance of these findings, we reanalyzed publicly available skin RNA-Seq data, including data from patients with T H 2-(atopic dermatitis, n 5 27) and T H 17-(plaque psoriasis, n 5 28) mediated conditions, as well as healthy volunteers (n 5 38). 19 We identified 510 genes that were differentially expressed in NL atopic dermatitis skin compared to site-matched control biopsy samples (log 2 FC >|0.5|; FDR < 0.05). When we compared these genes with the 531 that were differentially expressed in NL-PPP skin, we observed a very significant overlap between the 2 data sets (111 shared genes; odds ratio over genomic background, 18.8). While this observation further confirmed the upregulation of T H 2 pathways in PPP skin, the evidence for T H 17 activation was less significant. In fact, there was a limited overlap between the NL-PPP data set and the 630 genes that were differentially expressed in NL psoriasis skin compared to HC (53 shared genes; odds ratio, 5.4) (Fig 1, D).
To further validate these findings, we used real-time PCR to analyze uninvolved acral skin obtained from 8 patients with PPP and 7 HCs (including 5 patients and 3 HCs who had not been included in the RNA-Seq experiment). This confirmed that key T H 2 genes such as IL4R, CCL13/MCP-4 (which activates the CCR3 receptor expressed by T H 2 cells 20 ), and CXCR4 (which is expressed on the surface of T H 2 cells and eosinophils 21 ) were upregulated in NL-PPP skin. A moderate increase of T H 1 (CXCR3) and T H 17 (IL17A) markers was also observed (Fig 1, E).
Taken together, these observations identified a marked signature of T H 2 activation in NL-PPP skin, with evidence for a more modest involvement of T H 17 pathways.

Increased abundance of 2 circulating T-cell subsets in PPP
We next investigated whether systemic immune responses were also deregulated in PPP. We therefore carried out single-cell RNA-Seq (scRNA-Seq) in PBMCs obtained from 7 PPP cases and 4 age-and sex-matched healthy volunteers (Table E1). After capture on a 103 Genomics platform, 39 end sequencing, and quality control, we observed 58,412 viable cells (see Table E4 in the Online Repository at www.jacionline.org). To maximize statistical power, we expanded this data set by including 8 publicly available control samples, which had been processed on the same platform used in our experiment, yielding comparable cell numbers 12,13 (see Fig E2, A, in the Online Repository). To integrate these external HCs in our resource, we undertook batch correction with the Harmony algorithm, 14 obtaining a total of 93,262 cells (Fig 2, A, and Fig E2, B).
When we analyzed the merged data set with Seurat, 15 we identified 13 cell clusters, which we visualized by uniform manifold approximation and projection (UMAP) (Fig 2, A, and Fig E2,  C). The annotation of cell identities (implemented by manual inspection of canonical marker genes [see Table E5 in the Online Repository at www.jacionline.org] and validated with SingleR 16 ) revealed that the clusters corresponded to natural killer cells, monocytes (CD14 1 and CD16 1 subsets), myeloid and plasmacytoid dendritic cells, B cells (memory and naive subsets), and T cells (2 naive, 1 effector, and 3 memory subsets) (Fig 2, A-C, and see Fig E3, A, in the Online Repository). Although unconventional T lymphocytes (mucosal-associated invariant T cells and gd T cells) were also detected, they did not form separate clusters (Fig E3, B and C).
A comparison of cases and controls showed that innate cells (monocytes, natural killer cells, and dendritic cells) were found at similar frequencies in the 2 groups. Conversely, 2 of the T-cell subsets were more abundant among affected individuals. These corresponded to clusters that we had initially labeled as memory CD4 1 T1 (accounting for 27.9% cells in cases vs 23.1% in controls; P 5 .02) and memory CD4 1 T2 (3.6% cells in cases vs 1.8% in controls; P < 10 24 ) cells (Fig 2, D and E). Control cell frequencies were comparable between the samples recruited in house and those retrieved from public databases, showing that the analysis was not skewed by the inclusion of external data sets (Fig E2, D).
To further investigate the identity of the 2 CD4 1 memory populations, we assessed whether they expressed cutaneous lymphocyte-associated antigen (CLA), a well-known skin-homing marker. We found that cells expressing SELPLG (the gene encoding CLA) were a minority among memory CD4 1 T1 lymphocytes, but they were very frequent in the CD4 1 T2 subset   (28.0% vs 51.2% P < .0001). Of note, ITGB7 (encoding the gut-homing receptor integrin b7) was virtually undetectable in the latter population, confirming the specificity of the skinhoming phenotype.
Interestingly, CD4 1 T2 cells also expressed ITGAE, which encodes the CD103 antigen (Fig 2, F). This identifies tissue-resident memory T cells (T RM cells) that have reentered the circulation and are migrating to secondary skin sites. 22 Thus, the memory CD4 1 T1 and CD4 1 T2 clusters correspond to circulating and skinhoming populations, respectively.

T H 17 skewing in the CD4 1 memory T cells of affected individuals
We next investigated the phenotype of CD4 1 memory T cells in affected individuals. Like other researchers, 23,24 we found that it was not possible to separate the different T helper cell subsets into specific subclusters. We therefore used SingleR to annotate T H 1, T H 2, and T H 17 cell identities within the existing CD4 1 T1 and CD4 1 T2 clusters. This revealed a significant enrichment of T H 17 lymphocytes among the memory CD4 1 T cells of PPP cases. The effect was observed in both the circulating (CD4 1 T1) and the skin-homing (CD4 1 T2) populations, but it was especially marked in the latter, where the median T H 17 fraction was 13.9% in cases versus 0.5% in controls (P < 10 24 ) (Fig 3, A, and see Fig E4, A, in the Online Repository at www.jacionline. org). No further abnormalities were consistently observed in both memory CD4 1 compartments (Fig 3, A, and Fig E4, A).
To validate these findings with another methodology, we examined the T H 1, T H 2, and T H 17 transcriptional signatures developed by Cano-Gamez et al. 23 This confirmed that T H 17 gene expression was elevated in both circulating and skinhoming cells of affected individuals, while T H 1 and T H 2 scores were not (Fig E4, B). In keeping with these observations, an analysis of the transcription factors driving T H 1 (TBX21/T-bet), T H 2 (GATA3), and T H 17 (RORG/RORgt) lineage commitment demonstrated that RORG (but not TBX21 or GATA3) was upregulated in the memory CD4 1 T cells of PPP patients (Fig 3, B). Of note, the overexpression of RORG was not replicated in memory CD8 1 T cells (Fig E4, C), which argues against a pathogenic involvement of Tc17 lymphocytes.
Taken together, these observations demonstrate a dominant T H 17 phenotype in the circulating memory CD4 1 T cells of people with PPP.

Increased T H 17 to T H 2 plasticity in the CD4 1 memory T cells of affected individuals
Given the different T-cell responses observed in PPP skin (T H 2 activation) and blood (T H 17 skewing), we investigated the possibility that T helper cell plasticity may contribute to disease pathogenesis. It is now well established that changes in the cytokine environment can modulate the identity of T H 17 cells and induce a shift toward T H 1 or T H 2 phenotypes. 25,26 We therefore sought to determine the extent of T H 17 cell plasticity in patients with PPP and in HCs.
We first queried the scRNA-Seq data generated in circulating and skin-homing CD4 1 memory T lymphocytes. Specifically, we searched for cells that expressed both GATA3 and KLRB1/CD161, which we selected as readily detectable T H 2 and T H 17 markers. This identified a subset of CD4 1 memory T cells that expressed both genes. Unsupervised hierarchical clustering showed that the GATA3 1 /CD161 1 cells were more closely related to T H 17 than T H 2 lymphocytes (Fig 4, A), as the expression of RORG and IL23R was readily detectable in dual-positive cells whereas the IL4R transcript levels were low (Fig 4, B, and see Fig E5, A, in the Online Repository at www.jacionline.org).
Although the simultaneous presence of GATA3 and CD161 has been documented in T H 2A cells (a T H 2 subtype associated with allergic disease 27 ), our dual-positive population did not show well-established T H 2A markers such as PPARG, PTGS2, or HPGDS 27 (Fig 4, B, and Fig E5, B). T H 9 signature genes such as SPI1/PU.1 and BATF 28 were likewise weakly expressed (Fig  4, B, Fig E5, B). Conversely, the GATA3 1 /CD161 1 cells in our data set had the same CCR6 1 /RORG 1 /GATA3 1 /CXCL8 1 phenotype as a T H 17/T H 2 subset observed among patients with asthma. 29 Here, we further explored the correlation between GATA3 1 / CD161 1 T H 17 and T H 2 cells by carrying out a pseudotime analysis of the entire CD4 1 T-cell compartment. Using Slingshot, 18 we found that GATA3 1 /CD161 1 cells appeared later in pseudotime compared to both T H 17 and T H 2 cells (Fig 4, C). Of note, the expression of GATA3 and KLRB1/CD161 continued to rise steadily during pseudotime, reflecting the pattern observed for T H 17 genes such as RORC and IL23R. Conversely, the levels of T H 2 markers such as CXCR4 and PTGDR2 peaked and then fell sharply (Fig 4, D). This is in keeping with the notion that the dual-positive cells differentiate from T H 17 rather than T H 2 lymphocytes.
We next investigated the pathogenic significance of the CD161 1 /GATA3 1 population. We observed that the dualpositive cells were more abundant among the memory CD4 1 T cells of PPP cases compared to those of controls (13.7% vs 6.3%, P 5 .004) (Fig 4, E).
Thus, we have identified a T H 17/T H 2 population that is associated with PPP.

Experimental validation of increased T H 17 and T H 17/T H 2 cell abundance in PPP cases
We next sought to validate the scRNA-Seq findings by flow cytometry analysis of PBMCs obtained from 6 affected individuals and 6 healthy volunteers (including 4 cases and 4 controls that had not been included in the scRNA-Seq experiment). We found that the overall abundance of memory CD4 1 T cells and T H 17 cells was comparable in cases and controls (see Fig E6, A and B, in the Online Repository at www.jacionline. org). However, the frequency of T H 17 cells among skin-homing T lymphocytes was elevated in individuals with PPP (9.8% vs 5.2% in healthy volunteers; P 5 .03) (Fig 5, A), reflecting the pattern observed by scRNA-Seq. Likewise, T H 17/T H 2 cells were more abundant in affected compared to unaffected subjects. This effect was observed in the overall memory CD4 1 T cell compartment (9.1% vs 6.6%, P 5 .009) and also documented in the skin-homing population (6.1% vs 4.3% P 5 .04) (Fig 5, B and C, and Fig E6, C). Thus, dual-positive T H 17/T H 2 cells are overrepresented among the skin-homing CD4 1 T cells of affected individuals and readily detectable in their dermal infiltrates.

DISCUSSION
The purpose of this study was to achieve a better understanding of the immunologic determinants of PPP, a condition that remains poorly understood at the etiological level and recalcitrant to treatment in real-world practice. 1 We applied hypothesis-free transcriptomic approaches to a tightly phenotyped PPP resource that met the rigorous inclusion criteria of the APRICOT clinical trial. 6 We focused on cells (circulating PBMCs) and tissues (NL skin) that were not affected by overt inflammation so that we could survey the immune landscape of the disease in an unbiased fashion. The advantages of this approach are exemplified by the results of the initial RNA-Seq experiment, where the comparison of lesional versus NL biopsy samples detected a predictable upregulation of innate pathways in involved skin. Conversely, the analysis of NL versus control samples revealed an unexpected and highly prominent signature of T-cell activation in uninvolved patient skin. The evidence for the activation of T H 2 lymphocytes was particularly significant, whereas the enrichment of T H 17-related pathways was relatively modest. This argues against the traditional classification of PPP as a clinical variant of plaque psoriasis 1 and highlights hitherto unsuspected similarities with atopic dermatitis.
The pathogenic involvement of T cells was also supported by the results obtained in circulating PBMCs. Because the use of Boolean flow cytometry gates cannot fully recapitulate the immune populations derived by scRNA-Seq cell clustering, there were some discrepancies between the results obtained with the 2 platforms. For example, scRNA-Seq experiments showed an increased frequency of the memory CD4 1 T1 and memory CD4 1 T2 clusters among affected individuals. Although the same trend was observed for the memory CD4 1 T cells detected by flow cytometry, the difference between cases and controls was not statistically significant.
Importantly, our key findings were validated in both platforms. Thus, scRNA-Seq and flow cytometry experiments consistently showed a skewed T H 17 phenotype for patient skin-homing T lymphocytes. They also demonstrated an increased abundance of T H 17/T H 2 (GATA3 1 /CD161 1 ) cells among PPP cases.
To our knowledge, these results provide the first evidence of systemic abnormalities in PPP. They may also explain the common occurrence of extracutaneous, T-cell-mediated comorbidities (eg, psoriatic arthritis, autoimmune thyroid disease) among affected individuals. 3 Our observation of increased T H 17/T H 2 cell abundance in PPP cases also suggests a pathogenic role for T H 17 plasticity. Interestingly, T H 17/T H 2 cells have been detected in the blood and bronchoalveolar lavage of asthmatic individuals, where they have been characterized as IL-4/IL-17-producing cells. 29,30 A similar enrichment in patient populations has been reported for IL-17/IFN-g-producing cells (T H 17/T H 1) in rheumatoid arthritis. Thus, T H 17 cells that are shifted toward T H 1 or T H 2 phenotypes are considered more pathogenic than their unshifted counterparts. 25,31 Intriguingly, T H 17 cell plasticity has also been associated with cigarette smoking, 32 one of the main risk factors for PPP. 2 Thus, several lines of evidence support the notion that the T H 17/T H 2 cells detected in the blood and skin of PPP patients contribute to disease processes.
It has been hypothesized that T H 17/T H 2 and T H 17/T H 1 cells originate in complex inflammatory milieus that cannot be easily recapitulated by in vitro polarization protocols. 25,26 This is in keeping with the multifaceted immune landscape we detected in PPP skin. Although the limitations of bulk RNA-Seq prevented us from characterizing these circuits, our analysis uncovered a clear upregulation of distinct cytokine networks.
The simultaneous activation of multiple immune pathways in PPP skin would also explain the limited therapeutic efficacy of biologic drugs that block single cytokines. 5 In fact, our results suggest that agents inhibiting diverse inflammatory pathways (eg, JAK inhibitors, which have been used with some success in individual PPP cases 33,34 ) might deliver better clinical outcomes than targeted monoclonal antibodies. In this context, single-cell analysis of the signaling hubs that are deregulated in PPP (eg, the JAK1/JAK3 or JAK2/TYK2 complex) holds the promise of identifying novel therapeutic targets for this severe and disabling disease.