Molecular and Cellular Characterization of Pyoderma Gangrenosum: Implications for the Use of Gene Expression

TO THE EDITOR

Pyoderma gangrenosum (PG) is characterized by painful ulcers typically affecting the lower extremities. PG pathogenesis and triggers are poorly understood (Ortega-Loayza et al., 2018). Treatments target systemic inflammation, but clinical response and outcomes remain unpredictable. Further investigations are necessary to understand PG pathobiology; however, little is known about gene expression in PG, including whether important changes localize to the dermis or epidermis and whether nonlesional skin from patients with PG shows sub-clinical signs of the disease. Thus, we analyzed the gene expression signatures of perilesional and nonlesional skin biopsies from patients with PG to characterize the immunologic and cellular response. This study was approved by Oregon Health and Science University’s Institutional Review Board.

We collected paired biopsies of perilesional and nonlesional skin from eight patients with PG and eight healthy controls (HCs) (Supplementary Materials and Methods; Supplementary Table S1); all patients provided written informed consent. Skin samples were collected while ulcers were clinically active (Supplementary Figure S1). Each biopsy specimen was incubated in a solution of aqueous 3.8% ammonium thiocyanate to separate the epidermis from the dermis. RNA was prepared from each tissue for RNA sequencing (Clemmensen et al., 2009) (GenBank: PRJNA590986). After generating alignments and gene counts using STAR (Dobin et al., 2013), gene-wise linear models were employed for differential expression analyses using limma with empirical Bayes moderation (Ritchie et al., 2015) and false discovery rate adjustment (Benjamini and Hochberg, 1995). Discovery tests compared perilesional and nonlesional dermis and epidermis from patients with PG with those of HC. Pathway analysis (Cytoscape using the Reactome F1 plugin) was performed using genes with a fourfold expression difference and false discovery rate P < 0.05 (Supplementary Tables S2‒5) (Shannon et al., 2003). Unregulated genes with fold change > 2 and false discovery rate P < 0.05 were analyzed with Immunological Genome Project software to correlate gene expression with likely cell types present in the perilesional dermis of PG (dermis of perilesional pyoderma gangrenosum [DPPG]) (Heng et al., 2008).

A total of 5,762 genes were significantly differentially expressed in DPPG compared with those in HC, and 5,235 genes were differentially expressed in DPPG compared with those in the dermis of nonlesional PG (fold change > 2, false discovery rate P < 0.05) (Figure 1a). Perilesional epidermis also had significantly differentially expressed genes, most of which were downregulated. The dermis of nonlesional PG and epidermis of nonlesional PG had few differentially expressed genes. Our pathway analysis revealed that differentially expressed genes in DPPG compared with those in the dermis of HC were associated with the signaling of neutrophil degranulation, cytokine–cytokine receptor interactions, the expression of complement cascade, and cell adhesion pathways. Pathway analysis comparing DPPG and dermis of nonlesional PG revealed signaling within similar pathways. Pathways associated with perilesional dermis revealed more clinically meaningful inflammatory pathways than pathways associated with the perilesional epidermis, although epidermal gene expression was also associated with IFN-α/β, cytokine receptors, and adhesion molecule pathways (Figure 1b and Supplementary Figure S2). Immunological Genome Project analysis showed that differentially expressed genes in PG were associated predominantly with myeloid cells—mainly dendritic cells but also granulocytes, macrophages, and monocytes (Figure 1c).

Figure 1. Pathway analysis and gene- associated cell types.

Analysis was based on differentially expressed genes in DPPG and EPPG versus HC. (a) Number of differentially expressed transcripts with FC > 2 and FDR P < 0.05. *Comparisons were selected for pathway analysis. (b) Inflammatory pathways associated with differentially expressed genes in perilesional dermis and epidermis compared with those in healthy dermis and healthy epidermis. Note the increased number of immune pathways found in the perilesional dermis in comparison with that in the perilesional epidermis, suggesting that most of the inflammatory events occur in the dermis. (c) Cell types associated with differentially upregulated genes (FC > 2, FDR P < 0.05) in the perilesional dermis and epidermis of PG compared with those in the HCs (Heng et al., 2008); myeloid cells are the most commonly represented (DCs, GNs, MFs, and Mos). B, BioCarta; BCR, B-cell receptor; CAM, cell adhesion molecule; DC, dermis nonlesional pyoderma gangrenosum; DPPG, dermis perilesional pyoderma gangrenosum; ENPG, epidermis nonlesional pyoderma gangrenosum; EPPG, epidermis perilesional pyoderma gangrenosum; FC, fold change; FCER, Fc epsilon receptor; FCGR, Fcγ receptor; FDR, false discovery rate; GN, granulocyte; HC, healthy control; K, Kyoto Encyclopedia of Genes and Genomes; MF, macrophage; Mo, monocyte; N, National Cancer Institute Pathways Interactions Database; PG, pyoderma gangrenosum; R, reactome; STAT, signal transducer and activator of transcription; TLR, toll-like receptor; Th17, T helper type 17; uPA, urokinase-type plasminogen activator; uPAR, urokinase-type plasminogen activator receptor.

This was followed by targeted analyses of an a priori list of selected genes previously implicated in PG pathogenesis (Supplementary Table S6). Figure 2a displays the cell types associated with these genes. Targeted analysis of cytokine gene expression revealed that DPPG had a cytokine gene expression signature that distinguished it from the dermis of HC, whereas the dermis of nonlesional PG did not (Figure 2b). Interestingly, the cytokine signatures of the epidermis of perilesional PG and epidermis of non-lesional PG were not significantly different from the epidermis of HC (Figure 2c). On the basis of pathway analysis results, select genes were validated using quantitative real-time PCR (Figure 2d) with statistically significant differences for CCL-20 and CXCL-2 (T helper type 17 pathway downstream cytokines).

Figure 2. Cytokine gene expression comparison of DPPG and EPPG with HC ;using 55 a priori selected genes.

(a) This table shows the cell types associated with a priori selected genes implicated in the pathogenesis of PG (GeneCards). (b) Heatmap of selected cytokine, chemokine, and cytokine signaling genes in the perilesional dermis of eight patients with PG. Overexpressed transcripts include genes implicated in Th17 induction/regulation (IL2RA, IL2RB, C5AR1, CXCL8, CSF3, CD40LG), differentiation (IL1B, IL6, IL23A), and biological effects (JAK, signal transducer and activator of transcription gene STAT, TYK2, IFN, TNF, IL17RA, IL36G, matrix metalloproteinase genes MMP). Nonlesional dermis of PG and dermis of HC are indistinguishable. (c) Heatmap of selected cytokine, chemokine, and cytokine signaling in the perilesional and nonlesional epidermis of four patients with PG compared with those in the epidermis of HC. Gene expression Z-scores were calculated using the overall mean and SD of each gene. Expression values used were normalized expression values on the log₂ scale with batch effect removed. (d) Quantitative RT-PCR validation of the main chemokine genes found in the RNA-seq analyses. These graphs show statistically significant overexpression of CCL20 and CXCL2 in the perilesional dermis of PG in comparison with those in the nonlesional dermis of PG and healthy controls by one-way ANOVA. Dermis of HC is shown in Figure 1b, or epidermis of HCs is shown in Figure 1c; n = 3 subjects. Four subjects per group were involved except where noted. DNPG, dermis nonlesional pyoderma gangrenosum; DPPG, dermis perilesional pyoderma gangrenosum; EPPG, epidermis perilesional pyoderma gangrenosum; HC, healthy control; PG, pyoderma gangrenosum; RNA-seq, RNA sequencing; Th17, T helper type 17.

On the basis of the results of our study, we corroborate the role of T helper type 17 inflammatory cytokines in the pathogenesis of PG (Ortega-Loayza et al., 2018; Wang et al., 2018). The relevance of these cytokines is confirmed by successful treatments with biologics; however, not all patients respond to these medications, which suggests that other pathways might be involved. Our analysis also revealed differential expression of Jak and IFN signaling genes (e.g., JAK3, signal transducer and activator of transcription 4 gene STAT4), which is consistent with the described therapeutic effectiveness of Jak inhibitors in PG (Orfaly et al., 2021). Jak inhibitors effectively treat inflammatory bowel disease and inflammatory arthritis, suggesting that these agents can also target PG-associated diseases.

Formed PG ulcers show nonspecific epidermal and superficial necrosis with a mixed inflammatory infiltrate. However, early lesions in PG localize to the dermis with intradermal abscess formation (Weedon, 2010). Our results show that inflammatory gene expression changes occur primarily in the dermis of PG, supporting this pathogenic model of PG. Thus, identifying the cellular profile within this dermal inflammatory response in patients with PG is of utmost importance; single-cell RNA sequencing is the next logical step to deepen our understanding of PG pathogenesis. Although PG is a neutrophilic dermatosis, our results reveal a strong association between differentially expressed genes and dendritic cell signatures, suggesting that the interactions of neutrophils and dendritic cells may be key drivers of this disease. Interestingly, biologic therapeutic interventions in PG are also proven to interfere with dendritic cell activation (Chu et al., 2011).

Overall, we report the following observations in this study:

1. Perilesional dermis of PG shows a cytokine gene expression signature consistent with the disease, whereas perilesional epidermis shows few inflammatory genes/pathways, and nonlesional skin of PG is similar to the skin of HCs (Figure 1). This finding confirms that most inflammatory events occur within the dermis rather than the epidermis.

2. Our pathway analysis implicates several pathways, including complement cascade and trafficking pathways (integrin, cell adhesion molecules), which suggest alternative therapies for PG.

3. Myeloid cells are the predominant cell type in PG and are responsible for the changes of gene expression in perilesional skin of PG.

Limitations of this study include the small sample size and thus the inability to control for variables such as age, sex, associated diseases, medications, or disease duration. Future directions will include comparing neutrophil-rich dermatoses to show that the differences in gene expression are not solely due to the presence of certain immune cell types in PG.

Abbreviations:

DPPG

dermis perilesional pyoderma gangrenosum

HC

healthy control

PG

pyoderma gangrenosum

Data availability statement

Datasets related to this article can be found at https://www.ncbi.nlm.nih.gov/sra/PRJNA590986, Pyoderma Gangrenosum Study hosted at Oregon Health and Science University.