The prediction of TFs' function and direct targets
Based on our RNA-seq and ATAC-seq data, we used BETA-plus to integrate differentially expressed genes and open accessible peaks [9, 15]. BETA identified 408 upregulated genes (rank product < 0.01, Additional file 6: Table S1) and 133 downregulated genes (Ratioup/down = 3.1). We noticed that the cumulative fraction of upregulated genes was much higher than static background states (P = 0.001, Fig. 1A), but the downregulated genes were insignificant. We thus focused on the 408 genes in the following analysis. Of the 408 significant upregulated genes, 57 were directly targeted by transcription factor AP-1.
Chromatin accessibility can be reflected by other epigenetic markers, such as DNA methylation, histone modifications, and DNase hypersensitivity. To further annotate 408 genes with accessibility peaks, we looked at whether these peaks can be interpreted by other epigenetic modifications. Based on our genome-wide DNA methylation dataset previously generated from the same samples, we extracted methylation levels for loci located within the accessible peaks [9]. For the 18 methylation loci located within 408 peaks, we found all accessible peaks marked by hypomethylation compared with normal controls (Additional file 7: Table S2). Meanwhile, some open accessibility could be confirmed by active histone modification and DNase hypersensitivity, retrieved from UCSC database, such as OSMR, CDCP1, and SHB (Additional file 1: Figure S1A–C). These findings indicated that some psoriasis-associated open accessible peaks were linked with other types of epigenetic markers, but most of them could not be directly interpreted.
For 408 upregulated genes, we found that the accessible peaks contained a consensus binding sequence, mainly targeted by transcription factor AP-1 member JUND, NFE2L2, NFE2, JDP2, FOSL1, BATF, JUN, FOS, and JUNB (Fig. 1B). In addition, compared with nontargeted genes, motifs analysis found potentially collaborating factors: MLX, SMAD3 (t test, P < 0.01, Fig. 1C). The primary AP-1 protein families in mammalian cells are JUN and FOS, forming heterodimer and homodimer through their leucine-zipper domains [16]. Moreover, a variety of pathways by AP-1 mediate the inflammatory response of psoriasis, such as TNF-α and IL-17 [17]. In summary, the BETA algorithm helped us identify several AP-1 targeted genes that might be instrumental for psoriasis development.
KEGG pathway analysis with upregulated target genes
To further explore the functional mechanism of the 408 upregulated genes, we utilized KEGG pathway analysis. These genes were significantly enriched in Influenza A, TNF signaling pathway, IL-17 signaling pathway, Hepatitis C, NOD-like receptor signaling pathway, Epstein-Barr virus infection, and Measles (P < 0.05, Fig. 2A). TNF-mediated chronic inflammation by controlling innate and adaptive immune cells and induced diverse chronic inflammatory diseases, such as psoriasis [18]. Genes enriched in the TNF signaling pathway included NOD2, BCL3, SOCS3, CASP8, RPS6KA4, CCL20, JUNB, TNFAIP3, MLKL, MMP9, PTGS2, CASP7, CEBPB (Fig. 2B, Additional file 8: Table S3).
Furthermore, we found that the upregulated genes were enriched on IL-17 signaling pathway and NOD-like receptor signaling pathway, which were tightly related to psoriasis (Fig. 2A, Additional file 2: Figure S2A, B). The genes enriched in the IL-17 pathway include S100A7A, LCN2, CASP8, CCL20, TBK1, TNFAIP3, MAPK6, MMP9, PTGS2, and CEBPB (Additional file 8: Table S3). And the NOD-like receptor signaling pathway contains NOD2, IFI16, CASP8, OAS2, NLRX1, OAS3, TBK1, GBP5, TNFAIP3, MYD88, NAMPT, PANX1, OAS1, and PYDC1 (Additional file 8: Table S3). Meanwhile, the upregulated genes also enriched in some virus-related signaling pathway, such as Influenza A, Hepatitis C, Epstein-Barr virus infection, and Measles, suggesting that virus infection might be an important triggering factor for psoriasis development [5]. Moreover, these enriched signaling pathways also intersect (Fig. 2C). We found that CASP8 and TNFAIP3 were enriched in all three TNF, IL-17, and NOD-like receptor signaling pathway. MMP9, PTGS2, and CEBPB were implicated within both TNF and IL-17 signaling pathway. While TBK1 was settled in IL-17 and NOD-like receptor signaling pathway.
For the 133 downregulated genes, we found these genes could be targeted by AP-1 factors JUND, BATF, NFE2, JDP2, FOSL1, FOS, JUN, JUNB, and were enriched in “Adipocytokine signaling pathway,” “FoxO signaling pathway,” “AMPK signaling pathway” and some others (Additional file 3: Figure S3A–B).
Target genes are associated with psoriasis histopathology
The histopathological traits are important for psoriasis diagnosis and may vary among patients. To check whether histopathological traits be controlled by different genes, we analyzed all samples by pathological sections and grouped samples into positive and negative status for each pathological feature, including hyperkeratosis and parakeratosis, Munro micro-abscess, acanthosis thickening, vascular dilatation congestion, elongation of rete pegs, granulosa thinning, and lymphocyte infiltration. By retrieving our transcriptome sequencing data of 20 psoriatic skins, we compared expression differences of positive and negative status for each feature. We found that histopathological features can be modulated by different AP-1 targets. For example, expression difference of ATP11B, EIF4E, SQLE, MYO1B, RAP2B, TTC9, HECTD1, STRN, SDR9C7, RAB7A, LIMK2, FRMD6, and CLPX can be detected in hyperkeratosis with parakeratosis (Fig. 3A, Additional file 4: Figure S4A). The expression of these 13 genes also differed in patients with acanthosis thickening (Fig. 3B, Additional file 4: Figure S4B). In addition, the expression of TNFAIP3, ATP11B, MYO1B, SQLE, TTC9, EIF4E, RAP2B, and HECTD1 was increased in vascular dilatation congestion (Fig. 3C, Additional file 5: Figure S5A). The differential genes showed strong overlap, indicating a close link between the three features in psoriasis lesions.
No expression differences were associated with Munro micro-abscess, elongation of rete pegs, or granulosa thinning. Nevertheless, a trend can be perceived that the expression of CDCP1, ATP11B, MYO1B, and EIF4E increases in patients with Munro micro-abscess (Fig. 3D). While in patients with elongation of rete pegs, the expression of PPARD, MID1IP1, LDLR, and ID1 tend to be augmented (Additional file 5: Figure S5B). In patients with granulosa thinning, the expression of NUP210, TNFAIP3, CDCP1, and CEBPG tends to increase (Additional file 5: Figure S5C).
Meanwhile, we evaluated the relationships between several confounder factors (Age, PASI, BMI, Gender, Smoking) and the expressions of genes targeted by AP-1. Among 235 events (47 genes × 5 factors), it was found that 93.19% (219/235) had no statistical significance (P > 0.05, Additional file 9: Table S4). Most important, the genes correlated with confounder factors were not associated with pathological traits mentioned above, suggesting that these genes might not contribute to psoriasis histopathology.
SQLE, STRN, EIF4E, and MYO1B might drive the development of psoriasis
To verify the reliability of 408 upregulated genes in our BETA analysis, we retrieved the publicly accessible expression array dataset with the largest psoriatic and control skin sample size (GSE30999). Among 408 genes in GSE30999, we found 91.42% (373/408) showed significant upregulation (P < 0.01), and 59.80% (244/408) suggested strong statistical significance with both P < 0.01 and |logFC|> 1, indicating that most of the upregulated genes can be confirmed by other datasets.
To further explore the potential pathogenesis of psoriasis regulated by AP-1-mediated targets, we mined and analyzed the differential expression data in the GEO database (GSE80047, GSE53552, GSE41662, GSE30999, GSE14905), all expression data were generated from psoriatic and normal skins. Among 57 AP-1 targeted genes, SQLE, STRN, EIF4E, and MYO1B were consistently upregulated in five public datasets and our BETA analysis (P < 0.01, logFC > 1, Fig. 4), suggesting that AP-1 might modulate disease development through these four genes.