Transcriptomic profiling of pemphigus lesion infiltrating mononuclear cells reveals a distinct local immune microenvironment and novel lncRNA regulators

Skin biopsies were collected from 4 patients with pemphigus, and 4 from age- and sex-matched healthy donors. In the pemphigus group, only blisters or erosions skin lesions were collected. Blood samples were also collected from 4 patients with pemphigus, and 4 from age- and sex-matched healthy donors. All the patients were diagnosed with pemphigus foliaceus or pemphigus vulgaris and had not been treated with systemic therapy before the study. The diagnoses were confirmed with clinical manifestations, histology, Dsg-specific antibody tests and immunohistology criteria. Shanghai Jiao Tong University School of the Medicine Research Ethics Committee approved the study. Written informed consent was obtained from all subjects before involving them in the study.

1cm² sized skin biopsy samples from four patients with pemphigus and four healthy donors were collected and incubated in a buffer containing collagenase IV, hyaluronidase, and DNase-I (Sigma-Aldrich, St. Louis, MO) for digestion at 37 °C for 2 h. After digestion, the samples were passed through a 70 mm cell strainer (BD Biosciences, USA), and single cell suspensions were obtained. Mononuclear cells were isolated from skin tissue single cell suspensions by density separation gradient using Lymphoprep solution (Axis-shield, Norway) and resuspended in RPMI 1640 (Invitrogen, USA) medium supplemented with 1 ml 5% fetal bovine serum (FBS; Sigma-Aldrich, USA) after washed with phosphate buffer saline. 4 ml blood samples were collected from a total of 8 participants (4 pemphigus patients and 4 healthy controls), from which PBMCs were isolated by density separation gradient using Lymphoprep (Stemcell Technologies, Vancouver, Canada) within 4 h since blood collection.

Ranging from 3.0*10⁵ to 8*10⁵ cells, Lymphoprep isolated mononuclear cells derived from each sample were prepared for further experimentation. Total RNA was extracted from the mononuclear cells of a pemphigus lesion and normal skin using Trizol (Invitrogen, USA). Purity and concentration of isolated total RNA were measured using a NanoDrop® UV–Vis spectrophotometer (Thermo Fisher, USA). Sampling and RNA isolation was performed by the same personnel using the same methodology.

The isolated RNA was labeled and hybridized on an Arraystar Human LncRNA Microarray v4.0 (Arraystar, USA) according to instructions on the expression manual by the KangChen Bio-tech Corporation (Shanghai, China). The Arraystar microarray detects a comprehensive collection of 40,173 LncRNAs including 7506 well annotated LncRNAs (Gold Standard) and 32,667 high confidence LncRNAs (Reliable) constructed using highly reputable public transcriptome database (Refseq, UCSC knowngenes, Ensembl). The array also includes a collection of 20,730 protein coding mRNAs supported by Universal Protein Resource (Uniprot) database. Further bioinformatic analysis was carried out in a R environment. Transcriptome data of PBMC was acquired from our previous study [12].

Raw signal intensity was converted into normalized and summarized expression data which was used as input for the linear models for microarray data analysis algorithm (LIMMA) to assess differential expression of genes between pemphigus group and HC. The computing process was done with LIMMA package in R. Genes with log fold-change (logFC) greater than or equal to 1 and p-value < 0.05, were regarded as differentially expressed and selected for further functional enrichment analysis. We used the clusterProfiler package in R (Guangchuang Yu, 2011) to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses on DEGs, respectively. GSEA was performed on the gene expression matrix through the clusterProfiler package and “c7.immunesigdb.v7.4.entrez.gmt” was selected as the reference gene set. A false discovery rate (FDR) < 0.25 and p < 0.05 were considered as significant enrichment.

The co-expression relationship between DEGs was investigated by Pearson’s correlation measures, and modules were detected using WGCNA package in R software (Langfelder & Horvath 2008). Scale-free topology fit index was set as 0.9 as a function of the soft-thresholding power. Edges with weight > 0.1 were selected to construct the co-expression network in Cytoscape (version 3.7.0) software (Broad Institute, Inc., Massachusetts Institute of Technology, and Regents of the University of California).

The FISH assay was performed to detect and localize LINC01588 and NOP14-AS1 in SOMC of pemphigus patients. The probes of LINC01588 and NOP14-AS1 were synthesized by the ServiceBio Company (China) and labeled with fluorescent dye. The Servicebio ™ FISH Kit (Servicebio Company, Wuhan, China) was used to carry out RNA FISH assay according to the procedure provided by the manufacturer.

Skin tissues were fixed and stained with hematoxylin. For Immunohistochemistry analysis, deparaffinized sections were washed with phosphate-buffered saline (PBS) and then treated with 3% hydrogen peroxide for 5 min. The sections were blocked with 10% normal goat serum in Tris–HCl-buffered saline or horse serum in PBS for 1 h and then incubated with primary anti-NCAM1/CD56 antibodies(Clone number: EP2567Y; Abcam, Waltham, USA) at a concentration of 1:200 for 1 h at room temperature or overnight at 4℃. After washing, the sections were incubated with appropriate secondary antibodies (biotin-conjugated IgG; Servicebio, Wuhan, China). The staining intensity was measured in three fields of every section and quantified morphometrically using Image J software.

SIMC Expression profiling data (4 patients and 4 controls, Additional file 1: Table S1) of 40,173 LncRNAs and 20,730 protein coding mRNAs were obtained by leveraging microarray analysis. PBMC expression data (4 patients and 4 controls, Additional file 2: Table S2) were obtained from our previous study [12]. After consolidation and normalization of the microarray data, DEGs were separately screened in each dataset. A total of 11,798 transcripts were differentially expressed at a significant level in SIMC while only 159 transcripts in PBMC comparing patients to control. Out of the DEGs in SIMC, there were 6829 mRNAs (1864 up-regulated and 4965 down-regulated) and 4969 lncRNAs (3515 up-regulated and 1454 down-regulated). As for DEGs of PBMC, there were 79 mRNAs (57 up-regulated and 22 down-regulated) and 60 lncRNAs (33 up-regulated and 27 down-regulated). DEGs are shown by volcano plots (Fig. 1a–d).

Next, those common DEGs in SIMC and PBMC were highlighted, as both types of cells have been proven to produce Dsg-autoantibodies, hence the common DEGs may provide more information on pemphigus-driver genes. Surprisingly, only 17 genes overlapped in two datasets (Fig. 1e), which constitute 12.23% of DEGs in PBMC dataset and only 0.14% of DEGs in SIMC dataset. Then the correlation heatmap were constructed for these 17 genes (Fig. 1f). Examining the statistically significant pair-wise correlation (p-value < 0.05 for Pearson co-efficient), one cluster of positively correlated genes were detected. The cluster included the mRNAs PPFIA3, GTAG1B, HPCA, TP73, and the lncRNAs G013396 and AC007003.1.

We expected the differences of DEGs between the SIMC and PBMC mainly stem from the immune cell subtype composition. To better understand and characterize the differences between SIMC and PBMC, we used the CIBERSORT deconvolution algorithm to identify immune cell infiltration characteristics in pemphigus patients. SIMC from pemphigus patients had a higher NK cells, and neutrophils infiltrating level, compared with control, while PBMC showed no such a difference (Fig. 2a, b, Additional file 3: Fig. S4). An increased number of infiltrating NK cells in pemphigus lesions was confirmed by immunohistochemistry. Dermal layer of pemphigus lesions had significantly more CD56 positive cells compared with normal skin (Fig. 2e).

To further study the immune environment, we generated the chemokines and chemokines receptors based on the global gene expression analysis (Fig. 2c). Notably, CCL19, CCL26, CCL27 and CXCR5 were highly expressed chemokines and chemokine receptors detected in pemphigus group (Fig. 2c). The correlation heatmap of the 16 types of immune cells revealed that B cell abundance had a positive correlation with Tfh cell and NK cell abundances (p < 0.05). Moreover, plasma cell (PC) abundance had a positive correlation with type I macrophage (M1) abundance, and dendritic cell population was positively correlated with type II macrophage (M2) abundance (Fig. 2d).

To further investigate the immunological mechanisms involved in pemphigus lesions, the GESA package in R was used to analyze the SIMC global gene expression data. The results displayed that the related genes were significantly enriched in immune cell activation processes and immune cell differentiation signatures (Fig. 2f). Among top scoring activated gene sets, from GSEA database (https://​www.​gsea-msigdb.​org/​gsea/​index.​jsp), immune cell differentiation signatures, GSE24574 (related to BL6^hi Tfh cells) and GSE37532 (related to regulatory T cell’s inability for maturation) were shown (Fig. 2g).

To better illustrate the unique roles of SIMC, we performed functional enrichment analyses on SIMC and PBMC datasets, respectively. The GO analysis results showed that DEGs in both datasets were mostly enriched in biological process terms. The up-regulated genes in SIMC were mainly enriched in neutrophil aggregation, defense response to fungus, Toll-like receptor 4 binding and RAGE receptor binding, among others (Fig. 3a). On the other hand, in PBMC group, over-represented genes were enriched in immunological processes including cellular response to interleukin-1, response to chemokines, CCR chemokine receptor binding, phagocytic vesicle lumen as well as endocytic vesicle, among others (Fig. 3b). KEGG pathway analysis showed the related genes in PBMC groups, were involved in viral infection related pathways, Chemokine signaling pathway, Cytokine-cytokine receptor interaction, and interleukin-17 (IL-17) signaling pathway (Fig. 3d). In SIMC group, most enriched KEGG terms were fructose and mannose metabolism, VEGF signaling pathway, p53 signaling pathway, PPAR signaling pathway, and IL-17 signaling pathway among others (Fig. 3c).

LncRNAs has emerged as critical regulators in the immune system. To explore its contribution to the immune phenotype of SIMC and PBMC, we looked at the expression level of immune cell specific lncRNAs. 22 immune-cell specific lncRNAs were differentially expressed in skin lesions. These lncRNAs are related to dendritic cells, CD8+ T cells, T helper 17 cells and NK cells, among others. While in PBMC, only TAPT1-AS1, which is specific to activated B cell was found (Table 1). It is well established that pemphigus is an IL-17 related immune response [13‐16]. Therefore, we next focused on Th17 cell specific LncRNA LINC01588. The PPAR signaling pathway score was calculated by averaging the expression values of genes in the PPAR signaling pathway. The PPAR signaling pathway gene-set was acquired from the Nanostring platform. The result showed that LINC01588 had a significant negative correlation with the PPAR signaling pathway (p-value < 0.01) (Fig. 4a). FISH results showed that LINC01588 had a higher expression in pemphigus lesions (Fig. 4b). Higher expression of NOP14-AS1, one of most up-regulated lncRNA in pemphigus lesions, was also confirmed using FISH. Furthermore, NOP14-AS1 was co-expressed with CD4 (Fig. 4e). To explore the relationship between lncRNA and mRNA, we next constructed a lncRNA-mRNA and lncRNA-miRNA-mRNA networks (Fig. 4c, d), based on the data from ENCORI database [17, 18].

The microarray dataset contains over 20 thousand gene expression data. Merely focusing on DEGs may lead to overlooking potentially significant results. Therefore, we next integrated the expression matrices of all 16 samples in both SIMC and PBMC datasets and identified pemphigus-related preserved gene modules in two datasets using weighted gene co-expression network analysis (WGCNA). After batch effect removal (Additional file 3: Fig. S1), scale free topology criterion was applied as follows: the soft threshold power of β was 10 when scale-free topology model fit R² was maximized at 0.85 (Additional file 3: Fig. S2). A total of 19 modules were identified in the network. The parameters were set as follow: a relatively large minimum module size (size = 30), a medium module detection sensitivity (deepSplit = 2), and the cut height for merging modules (height = 0.25). The modules whose eigengenes were correlated above 0.75 were merged. Dendrogram clusters and heatmaps are shown in the attachment (Additional file 3: Fig. S3). The heatmaps of eigengene adjacency and module-trait relationships showed that, in the 19 modules, the brown module was positively correlated with occurrence of pemphigus. None of the modules correlated with sample sources were found (Fig. 5a, b).

A multi-dimensional scaling plot was generated to evaluate the expression of genes in the brown module (Fig. 5c). The brown module includes 1462 mRNAs and 1129 lncRNAs. To screen out the hub genes, we calculated the intramodular connectivity of all genes in the module. According to the multi-dimensional scaling plot, we defined genes of high gene significance (GS > 0.5) and high intramodular connectivity (IC > 0.8) as main contributor genes in the brown module. Then, functional enrichment analysis was used to investigate the module function. The bubble plot illustrates that main contributor genes were markedly enriched in dendritic spine membrane, interleukin-28 receptor complex, dendrite membrane, cluster of actin-based cell projections, neuronal cell body membrane, cell body membrane and neuron projection membrane (adjust p value < 0.01, Fig. 5d). We also constructed the Protein–Protein Interaction Networks (PPIs), to explore molecule interaction in the brown module on protein level. PPIs of these genes were constituted by STRING and visualized by Cytoscape. The genes with combined scores greater than 0.4 were selected for constructing networks. The ten hub genes were chosen using CytoHubba plugin and were as follows: DNAJC17, PRSS3P2, POMGNT1, DTX4, G028733, GPHBP1, G011530, XLOC_001219, PAPD4, and G050942 (Fig. 5f). In the enrichment results, genes such as IFNL, IFNLR (Additional file 4: Table S3) and hub gene DTX4 were type I IFN related. Type I IFN signature has been reported in many autoimmune diseases, such as systemic lupus erythematosus. Hence, we displayed type I IFN signaling pathways associated genes in a heatmap manner. However, most genes were not up-regulated in pemphigus (Fig. 5e). The type I IFN signaling pathways gene-set used above was generated using the Nanostring panel database (https://​www.​nanostring.​com/​products/​ncounter-assays-panels/​panel-selection-tool/​).