Integrative Network Toxicology Identifies NFKB1 and KIF11 as Candidate Genes Associated with 2-Hydroxyphenanthrene in the Context of Psoriasis

Abstract

Background

2-Hydroxyphenanthrene (2-PHE), a polycyclic aromatic hydrocarbon metabolite, has been implicated in immune and inflammatory dysregulation, yet its potential toxicological contribution to psoriasis remains unclear. This study aimed to identify candidate genes and pathways through which 2-PHE might influence psoriasis using a network toxicology framework.

Methods

Differentially expressed genes (DEGs) were identified from the psoriasis dataset GSE30999 and intersected with 2-PHE–predicted targets obtained from SEA, SuperPred, and TargetNet. Functional enrichment, protein–protein interaction (PPI) analysis, and topological algorithms were used to define hub genes. Two-sample Mendelian randomization (MR) was performed to explore suggestive associations with psoriasis. Tissue-classification performance was evaluated by ROC analysis in GSE30999 and GSE41662 for distinguishing lesional from non-lesional skin, while GSEA characterized pathway involvement. Molecular docking simulations assessed potential binding modes between 2-PHE and key proteins.

Results

A total of 4,185 psoriasis-associated DEGs and 219 2-PHE targets yielded 55 overlapping genes enriched in NF-κB signaling, apoptosis, insulin resistance, and AGE–RAGE pathways. PPI analysis identified 13 hub genes, with NFKB1 and KIF11 showing nominally significant associations with psoriasis in MR analyses. Both NFKB1 and KIF11 showed strong capacity to discriminate lesional from non-lesional tissue and were enriched in inflammatory, immune, and cell-cycle–related pathways. Docking predicted favorable binding energies of 2-PHE to NFKB1 (−5.9 kcal/mol) and KIF11 (−9.0 kcal/mol) in silico.

Conclusion

This integrative network toxicology study suggests that NFKB1 and KIF11 are candidate genes potentially involved in the association between 2-PHE and psoriasis, linking environmental exposure to immune activation and cell-cycle dysregulation. However, the genetic evidence remains preliminary, and the lack of colocalization support and the nominal nature of MR findings underscore the need for further validation through functional experiments and larger genetic studies.

Introduction

Psoriasis is a chronic, immune-mediated inflammatory skin disease characterized by well-demarcated erythematous plaques with silvery scales and histopathological features such as keratinocyte hyperproliferation, aberrant differentiation, and dense immune cell infiltration.¹ It affects approximately 2–3% of the global population and imposes a substantial burden on quality of life, healthcare resources, and comorbid conditions including metabolic syndrome, cardiovascular disease, and psychological disorders.² Although biologic therapies targeting TNF-α, IL-17, and IL-23 have significantly improved disease control, psoriasis remains incurable, and many patients experience incomplete responses, loss of efficacy, or treatment-limiting adverse effects.³ These limitations underscore the need to better elucidate the molecular mechanisms underlying psoriasis onset and progression, particularly in relation to environmental exposures that may modulate disease risk or severity.

A growing body of evidence implicates environmental pollutants, particularly polycyclic aromatic hydrocarbons (PAHs), in the pathogenesis and exacerbation of inflammatory skin diseases.⁴ PAHs are ubiquitous contaminants generated by incomplete combustion of organic materials and are prevalent in urban air pollution, tobacco smoke, occupational settings, and certain consumer products. Upon cutaneous exposure, PAHs can penetrate the epidermis, activate xenobiotic-sensing pathways such as the aryl hydrocarbon receptor (AhR), induce oxidative stress through cytochrome P450-mediated metabolic activation, and modulate inflammatory and immune responses.^5,6 These processes have been linked to barrier dysfunction, altered keratinocyte differentiation, and immune dysregulation—hallmarks of psoriatic pathology. However, most previous studies have focused on prototypical PAHs such as benzo[a]pyrene, and comparatively little is known about the molecular effects of specific low-molecular-weight PAH congeners or their metabolites on psoriasis-relevant pathways.

Phenanthrene is one of the most abundant low-molecular-weight PAHs detected in ambient air, soil, and indoor dust, and it is consistently found in human biological matrices.⁷ Upon absorption, phenanthrene undergoes oxidative metabolism via cytochrome P450 enzymes, predominantly CYP1A1 and CYP1B1, to yield mono-hydroxylated metabolites, among which 2-hydroxyphenanthrene (2-PHE) is a major product.⁸ 2-PHE is routinely measured in urine as a biomarker of phenanthrene exposure in environmental epidemiology studies, reflecting recent external exposure and individual metabolic capacity.⁹ Importantly, hydroxylated PAH metabolites are not merely inert detoxification products; they can exhibit distinct biological activities, including pro-oxidant effects, genotoxicity, and endocrine-disrupting properties.¹⁰ To date, however, no epidemiological study has directly investigated whether phenanthrene exposure or urinary 2-PHE levels are associated with psoriasis incidence, severity, or flare frequency. To our knowledge, such evidence does not currently exist in the published literature, and we explicitly acknowledge this gap. Furthermore, direct evidence regarding the cutaneous bioavailability of 2-PHE—specifically, whether it can reach biologically effective concentrations in epidermal keratinocytes or dermal immune cells following environmental or occupational exposure—is lacking. Existing toxicokinetic data for PAHs have primarily focused on parent compounds and systemic circulation, whereas skin-specific disposition of hydroxylated metabolites remains poorly characterized.¹¹ The assertion that 2-PHE may accumulate in skin cells and directly interact with signaling proteins is therefore a hypothesis generated for this computational study, rather than a conclusion drawn from experimental data. Given the potential of hydroxylated PAHs to modulate cellular signaling pathways, and the established role of oxidative stress and immune dysregulation in psoriasis, we reasoned that 2-PHE merits systematic investigation as a candidate environmental factor that might influence psoriatic pathophysiology, while recognizing that this hypothesis requires experimental testing.

Concurrently, advances in high-throughput transcriptomics and large-scale genome-wide association studies (GWAS) have accelerated the discovery of psoriasis-associated genes and loci.^12,13 These studies have identified susceptibility loci implicating immune signaling pathways, including the NF-κB cascade, IL-23/Th17 axis, type I interferon response, and epidermal differentiation complexes.¹⁴ Nevertheless, most mechanistic studies remain descriptive, and few have integrated environmental toxicant exposures with omics-based approaches to delineate potentially relevant pathways. Moreover, while network pharmacology and toxicology have been increasingly applied to drugs and traditional medicines, their application to environmental contaminants in the context of psoriasis is still limited. This gap hampers our understanding of how specific chemicals such as 2-PHE may intersect with psoriasis-associated genetic networks at a molecular level. Another critical gap lies in distinguishing potential causal relationships rather than mere association between candidate genes and psoriasis. Traditional observational studies are susceptible to confounding and reverse causation. Mendelian randomization (MR), leveraging genetic variants as instrumental variables, offers a powerful framework to infer whether genetically proxied expression of a gene may influence disease risk. However, it is important to emphasize that MR analysis of gene expression on psoriasis does not directly establish that 2-PHE exposure affects the gene, that the chemical and eQTL alter expression in the same direction, or that the gene mediates a chemical exposure effect. Instead, MR provides genetic support for prioritizing candidate genes that may warrant further investigation in relation to both psoriasis susceptibility and potential environmental modulation. Yet, this integrative approach has rarely been used to link in silico-predicted toxicant targets to psoriasis risk genes, leaving a key question unanswered: do toxicant-interacting proteins show genetic associations with psoriasis, and are they therefore candidates for further mechanistic study?

In this context, we designed a network toxicology study to identify candidate genes and pathways through which 2-PHE may influence psoriasis pathophysiology. We first identified differentially expressed genes (DEGs) in psoriatic lesions compared with non-lesional skin using the GSE30999 transcriptomic dataset and integrated these with predicted 2-PHE targets derived from multiple public databases. By intersecting DEGs with 2-PHE targets, we aimed to define a set of candidate toxicity-related genes and to characterize their biological functions through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. To further refine key nodes within this network, we constructed a PPI network and applied multiple topological algorithms to identify hub genes that may represent key mediators in the network, under the hypothesis that genes with high centrality are more likely to be functionally relevant. To move beyond correlation, we then applied two-sample MR to explore the suggestive associations of these hub genes on psoriasis onset, followed by colocalization analysis to assess the likelihood that psoriasis-associated GWAS signals and expression quantitative trait loci for key genes share a common causal variant-while noting that such genetic evidence does not directly test the chemical exposure pathway. We also assessed the capacity of the identified genes to discriminate lesional from non-lesional tissue in independent psoriasis cohorts and used gene set enrichment analysis to explore their functional context. Finally, we performed molecular docking simulations to predict the binding affinity and potential interaction mode in silico of 2-PHE with the most promising candidate genes, generating testable hypotheses that link computational toxicology predictions with genetically supported targets.

We hypothesized that 2-PHE may interact with a subset of psoriasis-relevant proteins and that some of these targets, particularly those situated at central positions in the PPI network and supported by suggestive MR evidence, may show associations with psoriasis susceptibility that warrant further investigation. Specifically, we posited that 2-PHE may modulate key regulators such as components of the NF-κB signaling pathway and mitotic machinery, thereby potentially affecting pro-inflammatory and proliferative pathways characteristic of psoriatic lesions. By integrating transcriptomics, network toxicology, genetic epidemiology, and structure-based docking, this study seeks to identify and prioritize mechanistically relevant candidate targets through which 2-PHE might exert effects relevant to psoriasis, providing a computational framework for evaluating the dermatological risks of environmental PAH exposures and generating hypotheses for future experimental validation.

Methods

Data Acquisition and Preprocessing of Psoriasis Datasets

The gene expression profiles of psoriasis were retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) under the accession numbers GSE41662 and GSE30999. Both datasets were generated using the GPL570 platform. GSE30999, comprising 85 paired lesional and non-lesional skin biopsy samples from psoriasis patients, was used as the discovery dataset. GSE41662, containing 24 paired lesional and non-lesional samples from an independent cohort of psoriasis patients, served as the validation dataset for diagnostic performance evaluation of the candidate genes. For independent validation of candidate gene expression, we additionally analyzed the psoriasis dataset GSE13355 (GPL570 platform), which contains 58 paired lesional and non‑lesional skin samples from psoriasis patients, along with 64 normal skin samples from healthy controls. We explicitly accounted for the paired design in our differential expression analysis by applying a paired limma model, which directly compares each lesional sample with its corresponding non-lesional counterpart from the same individual, thereby controlling for inter-individual variation. Raw CEL files were processed and normalized using the affy package. Background correction was performed using the Robust Multi-array Average (RMA) algorithm. Probe sets were mapped to gene symbols based on the manufacturer’s annotation file; probes without corresponding gene symbols were discarded. For multiple probes mapping to the same gene, the median expression value was retained as the final gene-level expression. Missing values were imputed using the impute package.

Identification of DEGs

Paired differential expression analysis between lesional and non-lesional samples within each patient was performed using the limma package with the duplicateCorrelation function. Genes with an adjusted P-value < 0.05 (Benjamini–Hochberg correction) and an absolute log2 fold change (|log2FC|) > 0.5 were considered statistically significant. The volcano plot was generated using the ggplot2 package to visualize the overall distribution of up-regulated and down-regulated genes. Heatmaps of the top 50 DEGs were constructed using the pheatmap package.

Prediction of 2-PHE Target Genes

The potential protein targets of 2-PHE were predicted using three public databases: TargetNet, SuperPred, and SEA. For TargetNet (http://targetnet.scbdd.com/), the SMILES notation of 2-PHE was submitted, and targets with a probability score ≥ 0.5 were retained. For SuperPred (https://prediction.charite.de/), the SMILES string was employed, and predictions with a probability ≥ 0.5 were selected. For the Similarity Ensemble Approach (SEA) database, we retrieved human-related targets with a corrected P-value < 0.05. The overlapping and unique targets from the three sources were merged to obtain a set of 219 high-confidence 2-PHE targets, which were visualized as a protein–compound interaction network using Cytoscape software.

Functional Enrichment Analysis

Functional enrichment analyses—including Gene Ontology (GO) terms spanning biological processes (BP), cellular components (CC), and molecular functions (MF)—as well as Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses, were performed using the clusterProfiler package. The organism-specific background gene set was set to Homo sapiens. An adjusted P-value < 0.05 (Benjamini–Hochberg correction) was applied as the threshold for statistical significance. The top enriched GO terms and KEGG pathways were visualized as horizontal bar charts using the ggplot2 package.

Protein-Protein Interaction Network Construction and Hub Gene Identification

To investigate the functional relationships among the candidate targets, we constructed a protein-protein interaction (PPI) network using the STRING database (https://string-db.org/). A combined interaction score ≥ 0.4 (medium confidence) was used as the threshold for inclusion of interactions. The network was then visualized and analyzed using Cytoscape software (version 3.9.2). To identify central nodes within the network, topological analysis was performed using the cytoHubba plugin in Cytoscape. Six distinct topological algorithms were applied: Degree, Closeness, EPC (Edge Percolated Component), MNC (Maximum Neighborhood Component), DMNC (Density of Maximum Neighborhood Component), and MCC (Maximal Clique Centrality). For each algorithm, the top 20 ranked genes were extracted. The consensus hub genes were identified by intersecting the results from all six algorithms, and the overlap was visualized using an UpSet plot generated with the UpSetR package.

Mendelian Randomization (MR) Analysis

To assess the potential causal associations between the identified hub genes and psoriasis, we performed a two-sample MR analysis. Summary-level genome-wide association study (GWAS) data for psoriasis were retrieved from the UK Biobank dataset (ukb-a-100). For each of the 13 hub genes, we selected genetic instruments using the following criteria: (1) SNPs significantly associated with the respective gene expression at the genome-wide significance threshold (P < 5 × 10⁻⁸); (2) SNPs were clumped using a clumping window of 10,000 kb and an r² threshold of 0.001 to ensure independence; (3) F-statistics for each instrument were calculated to assess instrument strength, with F > 10 considered indicative of no weak instrument bias; (4) SNPs with ambiguous alleles or strand-ambiguous palindromic SNPs were excluded; and (5) only SNPs present in both the exposure and outcome GWAS datasets were retained. The harmonized data ensured that the effect alleles were oriented consistently across exposure and outcome. The MR analysis was conducted using the TwoSampleMR package in R (version 4.2.0). For each hub gene, single nucleotide polymorphisms (SNPs) significantly associated with the respective gene expression was selected as instrumental variables. The MR analysis was conducted using the TwoSampleMR package. Five complementary MR methods were employed to estimate causal effects: inverse variance weighted (IVW), MR Egger, weighted median, simple mode, and weighted mode. The IVW method was used as the primary analysis under the multiplicative random-effects model, while the other methods served as sensitivity analyses to detect potential horizontal pleiotropy and to assess the robustness of the findings. To evaluate the reliability of the MR results, we conducted the following sensitivity analyses: (1) Cochrane’s Q test for heterogeneity among the IVW estimates; (2) the MR-Egger intercept test to detect directional pleiotropy; (3) MR-PRESSO to identify and correct for outlier SNPs; (4) Steiger filtering to confirm the directional relationship between the instrumented gene expression and psoriasis risk; and (5) leave-one-out analysis to examine whether the causal estimates were driven by any single SNP. Given that 13 hub genes were tested, we applied a Bonferroni correction for multiple testing, setting the significance threshold at P < 0.05/13 = 0.0038. Forest plots and scatter plots were generated using the ggplot2 package to visualize the effect estimates of individual SNPs and the overall causal relationships between each gene and psoriasis risk.

Bayesian Colocalization Analysis

To assess whether the observed associations between gene expression and psoriasis risk were driven by a shared causal variant, we performed Bayesian colocalization analysis using the coloc package (version 5.1.0). Summary-level data for NFKB1 expression (eQTL) and psoriasis GWAS (ukb-a-100) were harmonized and extracted using the ieugwasr_to_coloc function. Colocalization analysis was conducted using the coloc.abf function, which evaluates the posterior probability of five distinct hypotheses regarding the existence of a shared causal variant between two traits within a given genomic region. These five hypotheses are: H0 (no association with either trait), H1 (association with trait 1 only), H2 (association with trait 2 only), H3 (association with both traits but driven by distinct causal variants), and H4 (association with both traits driven by a single shared causal variant). The prior probabilities were set to the default values: p1 = p2 = p12 = 1 × 10⁻⁴. We report both the global posterior probability for H4 (PP.H4) across the entire locus and the single-SNP posterior probability (SNP.PP.H4) for the lead variant. The regional association plots visualizing the colocalization signals were generated using the stack_assoc_plot function from the gassocplot2 package.

Gene Set Enrichment Analysis (GSEA)

The analysis was conducted using the clusterProfiler package. For each target gene, the expression levels of all genes in the GSE30999 dataset were ranked based on their Pearson correlation coefficients with the respective target gene. A total of 1,000 permutations were performed to calculate the statistical significance. An enrichment score (ES) was calculated for each gene set, representing the degree to which a gene set was overrepresented at the top or bottom of the ranked gene list. The GSEA enrichment plots were generated using the enrichplot package.

Molecular Docking Analysis

To evaluate the potential direct binding interactions between 2-PHE and the identified core proteins (NFKB1 and KIF11), molecular docking was performed using the CB-Dock2 online server (https://cadd.labshare.cn/cb-dock2/). The three-dimensional crystal structures of NFKB1 and KIF11 were retrieved from the Protein Data Bank (PDB; https://www.rcsb.org/) with the following identifiers: PDB ID 8TQD for NFKB1 (resolution 2.02 Å) and PDB ID 3K5E for KIF11 (resolution 1.97 Å). These structures were selected based on their high resolution and the availability of the complete binding domains relevant to ligand interaction. Prior to docking, the protein structures were prepared by removing all water molecules and heteroatoms, followed by the addition of polar hydrogen atoms and assignment of Gasteiger charges using AutoDock Tools. The chemical structure of 2-PHE in SMILES format was converted to a three-dimensional structure using the Open Babel software and subsequently optimized for docking. Blind docking was performed using the CB-Dock2 server, which automatically identifies potential binding pockets and calculates binding affinities using an integrated AutoDock Vina engine. To validate the docking protocol, we performed re-docking of the co-crystallized ligand for each protein structure and calculated the root-mean-square deviation (RMSD) between the predicted and experimental ligand poses. An RMSD value < 2.0 Å was considered acceptable, confirming the reliability of the docking parameters. The docking scores (kcal/mol) were calculated to estimate the binding free energies of the protein-ligand complexes. The optimal binding pose was selected based on the lowest binding energy and visual inspection of key intermolecular interactions (hydrogen bonds, hydrophobic contacts, and π-π stacking) using PyMOL and Discovery Studio Visualizer.

Statistical Analysis Summary

All statistical analyses were performed using R software (version 4.2.0). For differential expression analysis, paired t-tests were implemented within the limma framework to account for the paired lesional/non-lesional design. For MR analysis, multiple testing correction was applied using the Bonferroni method (α = 0.05/13 = 0.0038). For colocalization analysis, posterior probabilities were interpreted according to established guidelines (PP.H4 > 0.8 as strong, 0.2–0.8 as suggestive, < 0.2 as weak). For all other analyses, a two-sided P < 0.05 or adjusted P < 0.05 was considered statistically significant, unless otherwise specified.

Results

Identification of Psoriasis-Associated DEGs and 2-PHE Target Collection

In this study, we analyzed the dataset GSE30999 (which consists of paired lesional and non-lesional skin biopsies from psoriasis patients) to identify DEGs associated with psoriasis, revealing 4,185 DEGs (Figure 1A). Of these, significantly upregulated and downregulated genes were represented, where red indicates upregulated genes and blue denotes downregulated genes. A hierarchical clustering heatmap was generated to display the expression patterns of the top 50 DEGs across psoriasis lesion (LS) samples and non-lesion (NL) samples, distinctly illustrating differential transcriptional activity between the two groups (Figure 1B). Additionally, we collected 219 predicted targets of 2-PHE from publicly available databases, including SEA, SuperPred, and TargetNet. These targets were visualized as a network, highlighting the interactions of 2-PHE across multiple molecular entities, which may potentially contribute to its role in modulating biological pathways relevant to psoriasis (Figure 1C). This comprehensive analysis provides insights into the molecular basis of psoriasis and lays the foundation for understanding the potential therapeutic mechanistic effects of 2-PHE.

Figure 1.

Overview of psoriasis-related DEGs and predicted targets of 2-PHE. (A) Volcano plot showing DEGs identified from the psoriasis dataset GSE30999. (B) Heatmap of the top 50 DEGs, illustrating expression patterns across LS and NL samples. (C) Network of predicted 2-PHE targets obtained from SEA, SuperPred, and TargetNet databases.

Identification and Functional Enrichment Analysis of Potential Toxicity Targets

To explore the potential toxicological mechanisms of 2-PHE in psoriasis, an intersection analysis was conducted between the 219 predicted 2-PHE targets and 4,185 DEGs identified from the GSE30999 dataset. This analysis revealed 55 overlapping genes, which were considered candidate toxicological targets of 2-PHE (Figure 2A). These 55 genes were subjected to functional enrichment analysis. GO enrichment analysis for BP highlighted their association with histone phosphorylation, regulation of phospholipase C activity, and response to oxidative stress (Figure 2B). CC enrichment showed significant involvement in synaptic structures, such as synaptic membranes and asymmetric synapses (Figure 2C). MF analysis indicated enrichment in protein kinase activity, specifically serine/threonine/tyrosine kinase activity (Figure 2D). KEGG pathway enrichment revealed significant pathways including the NF-kappa B signaling pathway, apoptosis, insulin resistance, and the AGE-RAGE signaling pathway, suggesting a role of these targets in inflammatory and metabolic dysregulation (Figure 2E). These findings provide insights into the molecular mechanisms by which 2-PHE may influence psoriasis pathophysiology.

Figure 2.

Overlap between 2-PHE targets and DEGs, and enrichment analysis of shared genes. (A) Venn diagram showing the intersection between 2-PHE predicted targets and psoriasis-related DEGs. Functional enrichment analysis of the 55 shared genes: (B) Gene Ontology (GO) biological process (BP) enrichment, (C) cellular component (CC) enrichment, (D) molecular function (MF) enrichment, and (E) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. Adjusted p-values indicate the significance of enrichment.

PPI Network Analysis and Identification of Hub Targets

To systematically prioritize candidate targets among the 55 intersecting genes, we constructed a PPI network using the STRING database and visualized the topology using Cytoscape (Figure 3A). The network revealed extensive connectivity among these targets, indicating cooperative biological functions. To identify the most centrally involved targets, we applied six topological algorithms—Degree, Closeness, EPC, MNC, DMNC, and MCC—to rank the top 20 hub genes from the PPI network. The intersection of these six algorithms identified a consensus set of 13 key hub genes (Figure 3B). Notably, the majority of these genes, including STAT3, CXCR4, NFKB1, CDK1, and TOP2A, were located in the highly connected core region of the PPI network. We then examined the expression profiles of these 13 hub genes in the GSE30999 dataset. As shown in the heatmap (Figure 3C), a striking dichotomy was observed between the LS and NL groups. Specifically, while genes such as KIT and NR3C1 were distinctly downregulated in psoriatic lesions, a cluster of genes including TOP2A, CDK1, KIF11, MCL1, NFKB1, and BCL2A1 exhibited marked upregulation in the LS group compared to NL group. This differential expression pattern suggests that 2-PHE may exert its toxic effects in psoriasis by perturbing these key hub genes, potentially impacting pathways related to cell cycle control, apoptosis resistance, and immune modulation.

Figure 3.

PPI network analysis and expression of key hub genes. (A) PPI network constructed for the 55 candidate toxicological targets of 2-PHE using the STRING database, visualized with Cytoscape. (B) Identification of the top 20 hub genes ranked by six topological parameters (degree, closeness, EPC, MNC, DMNC, and MCC). (C) Heatmap showing the expression profiles of the 13 hub genes in LS and NL samples based on the GSE30999 dataset.

MR Analysis of Causal Effects of Key Hub Genes on Psoriasis Onset

As summarized in Table 1, MR estimates across multiple methods—IVW, MR Egger, weighted median, simple mode, and weighted mode—revealed that NFKB1 and KIF11 exhibited nominally significant associations with psoriasis onset. Specifically, the IVW method showed a positive association for KIF11 (OR = 1.005, 95% CI: 1.001–1.009, p = 0.010), while NFKB1 demonstrated consistent nominal significance across the IVW (OR = 1.001, 95% CI: 1.00–1.004, p = 0.041), weighted median (OR = 1.002, 95% CI: 1.00–1.003, p = 0.034), and weighted mode (OR = 1.002, 95% CI: 1.00–1.003, p = 0.084) analyses. Sensitivity analyses were performed to evaluate the robustness of the MR estimates. Cochrane’s Q test revealed no significant heterogeneity among the instrumental variables for NFKB1 (IVW Q_pval = 0.191; MR Egger Q_pval = 0.133) or for KIF11 (IVW Q_pval = 0.380), supporting the consistency of the MR estimates across individual SNPs (Table S1). The MR-Egger intercept test did not detect significant directional pleiotropy for NFKB1 (intercept = 1.58 × 10⁻⁴, p = 0.676). For KIF11, the MR-Egger intercept test could not be performed due to the limited number of instrumental variables (n = 2), which we acknowledge as a constraint on the robustness of this particular finding (Table S2). Collectively, these sensitivity analyses support the reliability of the NFKB1 association, while the KIF11 result remains preliminary and warrants further investigation with additional genetic instruments. However, given that we tested 13 hub genes, we applied a Bonferroni correction for multiple comparisons, which set the significance threshold at P < 0.05/13 = 0.0038. Neither NFKB1 nor KIF11 reached this corrected threshold, indicating that these MR findings should be interpreted as suggestive rather than definitive. The forest plots depicted in Figure 4A and C provide detailed visualizations of the SNP-level causal effects for NFKB1 and KIF11, respectively, where the horizontal bars indicate the 95% confidence intervals for each instrumental variable. Correspondingly, the scatter plots in Figure 4B and D illustrate the dose-response relationships between SNP effects on the exposure gene and SNP effects on psoriasis risk for NFKB1 and KIF11, showing positive slopes across the MR methods that are consistent with a potential causal link, though this remains to be confirmed. Together, these results implicate NFKB1 and KIF11 as candidate risk genes through which 2-PHE may exert its psoriasis-promoting effects, but the evidence is preliminary and requires replication in independent cohorts with larger sample sizes.

Table 1.

MR-Based Estimates of Causal Effects of Genes on Psoriasis Onset

Outcome	Exposure	Method	nsnp	b	Standard Error	pval	OR (95% CI)
Psoriasis	KIF11	Inverse variance weighted	2	0.005	0.002	0.010	1.005 (1.001–1.009)
Psoriasis	NFKB1	MR Egger	6	0.001	0.002	0.588	1.001 (0.997–1.005)
Psoriasis	NFKB1	Weighted median	6	0.002	0.001	0.034	1.002 (1–1.003)
Psoriasis	NFKB1	Inverse variance weighted	6	0.002	0.001	0.041	1.001 (1–1.004)
Psoriasis	NFKB1	Simple mode	6	0.002	0.002	0.300	1.002 (0.998–1.006)
Psoriasis	NFKB1	Weighted mode	6	0.002	0.001	0.084	1.002 (1–1.003)

Figure 4.

Two-sample MR analysis for assessing the causal role of hub genes in psoriasis risk. (A) Forest plot showing the MR effect sizes of SNPs associated with the NFKB1 gene using inverse variance weighted (IVW) and MR-Egger methods. (B) Scatter plot of the SNP effects for NFKB1 on psoriasis. The MR test included IVW, weighted median, MR-Egger, and weighted mode approaches. (C) Forest plot of the MR effect sizes for SNPs associated with the KIF11 gene. (D) Scatter plot of SNP effects for KIF11 on psoriasis using the IVW method.

Colocalization Analysis of Psoriasis-Associated Genetic Signals and NFKB1 Expression Quantitative Trait Loci

As shown in the regional association plots (Figure 5), the top expression quantitative trait locus (eQTL) for NFKB1, rs10013613, was located on chromosome 4 and exhibited a strong association with NFKB1 expression levels (top panel, purple diamond), whereas the lead psoriasis-associated SNP, rs230497, showed moderate significance in the GWAS data (bottom panel). We performed Bayesian colocalization to evaluate whether a shared causal variant drives both the eQTL and GWAS signals. The results, summarized in Table 2, revealed that the single-SNP posterior probability for hypothesis 4 (SNP.PP.H4) at the rs10013613 locus was 0.8967. However, the global posterior probability for colocalization across the entire locus (PP.H4) was only 0.02, indicating that the observed alignment at the lead SNP does not translate into robust evidence for a shared causal variant at the locus level. Thus, contrary to our initial interpretation, the colocalization analysis does not support a shared genetic architecture between NFKB1 expression and psoriasis susceptibility.

Figure 5.

Regional association plots and Bayesian colocalization analysis of the NFKB1 locus.

Table 2.

Bayesian Colocalization Results

Gene	snp	V.df1	z.df1	r.df1	lABF.df1	V.df2	z.df2	r.df2	lABF.df2	internal.sum.lABF	SNP.PP.H4
NFKB1	rs10013613	0.000144	22.25834	0.997011	244.0701	7E-08	2.220886	0.999998	−4.16166	239.9084	0.896743099

Diagnostic Performance and Functional Enrichment Analysis of NFKB1 and KIF11

In GSE30999, the AUC was 0.836 for NFKB1 and 0.949 for KIF11, demonstrating robust predictive performance (Figure 6A). Similarly, in GSE41662, the AUC values were 0.825 for NFKB1 and 0.976 for KIF11, confirming their strong diagnostic capability for psoriasis (Figure 6B). Furthermore, GSEA was conducted to investigate the functional pathways associated with these risk genes. For NFKB1, enriched pathways included RNA metabolic processes, interleukin-mediated signaling, and ribosome biogenesis, highlighting its role in inflammation and immune regulation (Figure 6C). For KIF11, significant enrichment was observed in pathways related to cytoskeletal dynamics, T cell-mediated immunity, and transcription regulation, suggesting its involvement in cellular processes relevant to psoriasis pathogenesis (Figure 6D).

Figure 6.

Validation of diagnostic performance and functional enrichment analysis for psoriasis risk genes NFKB1 and KIF11. ROC curves for NFKB1 and KIF11 in the psoriasis datasets GSE30999 (A) and GSE41662 (B), showing their diagnostic performance. AUC values indicate the predictive accuracy of each gene. GSEA of NFKB1 (C) and KIF11 (D), illustrating pathways enriched for each gene.

Cross‑dataset Validation of NFKB1 and KIF11 Expression in an Independent Cohort

To evaluate the reproducibility of our expression findings, we examined the expression levels of NFKB1 and KIF11 in an independent psoriasis transcriptomic dataset, GSE13355, which comprises 58 paired LS and NL skin samples from psoriasis patients, as well as 64 normal skin samples from healthy controls (Nor). Consistent with our observations in GSE30999, both genes were significantly upregulated in psoriatic lesions compared with paired non‑lesional skin (NFKB1: P < 0.001; KIF11: P < 0.001) and also showed elevated expression relative to normal healthy skin (NFKB1: P < 0.001; KIF11: P < 0.001) (Figure S1). This cross‑dataset consistency supports the robustness of our candidate gene identification and suggests that the differential expression of NFKB1 and KIF11 is not specific to a single cohort.

Molecular Docking Analysis

To evaluate the potential interaction between 2-hydroxyphenanthrene (2-PHE) and psoriasis-associated risk genes, molecular docking simulations were performed for the NFKB1 and KIF11 proteins. The docking of 2-PHE with NFKB1 revealed a binding affinity score of −5.9 kcal/mol, indicating a moderate interaction. The binding pocket consisted of key residues such as Q202, Q198, V171, and H107, which formed hydrogen bonds and hydrophobic interactions with 2-PHE, stabilizing the ligand-protein complex (Figure 7A). For KIF11, a stronger binding affinity was observed, with a docking score of −9.0 kcal/mol. Residues such as I136, E116, R119, Y211, and L214 were identified as critical for ligand stabilization through hydrogen bonds and van der Waals forces (Figure 7B). These docking results suggest a differential binding affinity of 2-PHE for these target proteins, highlighting its potential mechanistic involvement in psoriasis through interactions with NFKB1 and KIF11.

Figure 7.

Molecular docking analysis. (A) Three-dimensional representation of the molecular docking complex between 2-PHE and the NFKB1 protein. (B) Three-dimensional representation of the molecular docking complex between 2-PHE and the KIF11 protein. The ligand is shown in a cyan stick model, and the protein is displayed in a ribbon representation. Key interacting amino acid residues are labeled with black text. Dashed lines denote the intermolecular interactions between the ligand and the protein.

Discussion

This study employed a network toxicology framework integrating transcriptomic analysis, bioinformatics prediction, MR, colocalization, and molecular docking to systematically investigate the potential toxic mechanisms by which 2-PHE may influence psoriasis pathogenesis. By intersecting 2-PHE predicted targets with psoriasis-associated DEGs, we identified a core set of hub genes—most notably NFKB1 and KIF11—that not only exhibited dysregulated expression in psoriatic lesions, showed nominally significant genetic associations with psoriasis onset in MR analyses, and displayed moderate to strong binding affinities to 2-PHE in docking simulations. However, the genetic evidence requires careful interpretation: the MR findings did not survive correction for multiple testing, and colocalization analysis did not support a shared causal variant at the NFKB1 locus. Collectively, these findings suggest, though do not establish, that environmental exposure to 2-PHE might contribute to psoriasis susceptibility or exacerbation through perturbation of NF-κB-driven inflammatory signaling and KIF11-mediated cell cycle regulation. Nevertheless, the current evidence is hypothesis-generating rather than confirmatory, and substantial experimental validation is needed before any causal relationship can be claimed.

Integration with Prior Knowledge on Psoriasis Molecular Pathogenesis

Psoriasis is widely recognized as a chronic, immune-mediated inflammatory skin disorder characterized by aberrant keratinocyte proliferation and immune cell infiltration.^15,16 The transcriptomic landscape of psoriatic lesions has been extensively studied, with prior analyses consistently identifying inflammatory mediators, proliferative kinases, and apoptosis-related genes as central players.^17,18 Our identification of 4,185 DEGs from the GSE30999 dataset aligns with the scale of transcriptional reprogramming reported in prior psoriasis GEO studies, reinforcing the robustness of the dataset as a foundation for downstream network analyses. The upregulation of TOP2A, CDK1, and KIF11 in lesional skin is consistent with reports demonstrating hyperproliferative keratinocyte behavior in psoriasis, where accelerated cell cycle progression is a defining pathological feature.^19,20 Conversely, the downregulation of KIT and NR3C1 echoes earlier findings suggesting impaired stem cell factor signaling and glucocorticoid receptor activity in psoriatic skin, the latter being particularly relevant to the well-documented clinical efficacy of corticosteroid-based therapies.^21–23

The enrichment of our 55 candidate targets in the NF-κB signaling pathway, apoptosis, and AGE-RAGE signaling pathways is biologically coherent and well-supported by established psoriasis literature. NF-κB activation is arguably one of the most extensively validated drivers of psoriatic inflammation, governing the transcription of pro-inflammatory cytokines such as TNF-α, IL-6, IL-17, and IL-23.^24,25 The AGE-RAGE axis, though historically associated with diabetic complications, has increasingly been implicated in skin inflammatory disorders through its capacity to amplify oxidative stress and sustain NF-κB activity.^26,27 The appearance of this pathway in our enrichment results is therefore not incidental but may reflect a mechanistic bridge between environmental polycyclic aromatic hydrocarbon (PAH) exposure and chronic skin inflammation, given that PAH metabolites are known inducers of oxidative stress and advanced glycation end-product formation.²⁸

NFKB1 as a Central Mediator of 2-PHE-Associated Psoriasis Risk

Among the identified hub genes, NFKB1 emerged as a promising candidate target. Its upregulation in psoriatic lesions, nominally significant genetic association with psoriasis onset via multiple MR methods, and moderate molecular docking affinity with 2-PHE (−5.9 kcal/mol) collectively point toward a plausible mechanistic hypothesis. Specifically, we speculate that 2-PHE might interact with the NFKB1 protein, potentially stabilizing or inducing conformational states that augment transcriptional activity, thereby amplifying downstream inflammatory cascades in keratinocytes and immune cells. Importantly, this interaction hypothesis is derived entirely from in silico docking simulations and has not been experimentally confirmed. However, the colocalization result deserves particular caution. Although the lead eQTL SNP rs10013613 showed a high single-SNP probability (SNP.PP.H4 = 0.8967), the global colocalization probability across the entire NFKB1 locus was only 0.02. This discrepancy indicates that the apparent alignment at the lead SNP does not provide sufficient evidence for a shared causal variant, and the observed MR signal may instead reflect linkage disequilibrium or other confounding structures. The lack of colocalization support substantially limits the strength of the genetic evidence linking NFKB1 to psoriasis susceptibility. Thus, while NFKB1 remains a biologically plausible candidate based on its expression pattern and established role in inflammation, the genetic evidence presented here should be viewed as exploratory rather than confirmatory. From a structural biology perspective, the binding of 2-PHE to NFKB1 residues Q202, Q198, V171, and H107 through hydrogen bonds and hydrophobic contacts is plausible given the planar aromatic architecture of 2-PHE, which facilitates intercalation into hydrophobic protein pockets. While the docking affinity of −5.9 kcal/mol falls within a moderate range, it is sufficient to warrant further experimental investigation, particularly under conditions of chronic low-level environmental exposure where sustained receptor engagement rather than high-affinity acute binding may be the operative paradigm.

KIF11 as a Novel Psoriasis Risk Gene Implicated in 2-PHE Toxicity

KIF11, a kinesin motor protein essential for bipolar spindle assembly during mitosis, has been less extensively studied in psoriasis compared to canonical inflammatory mediators.^29,30 Our results, however, make a preliminary case for its relevance. Its marked upregulation in lesional skin, strong diagnostic AUC values, nominally significant association with psoriasis onset by IVW analysis, and notably strong docking affinity with 2-PHE (−9.0 kcal/mol) in silico collectively suggest that KIF11 represents a candidate target worthy of further investigation. The high docking score is particularly striking, as it surpasses the affinity observed for NFKB1 and approaches values typically associated with pharmacologically active compounds, raising the in silico possibility that 2-PHE could potentially influence KIF11 activity if the predicted binding is biologically relevant. KIF11 inhibition has been previously explored as an anticancer strategy, given its indispensability for tumor cell proliferation.³¹ In psoriasis, where keratinocyte hyperproliferation is a core pathological feature, aberrant KIF11 upregulation may accelerate mitotic cycling and contribute to the thickened epidermis characteristic of plaques. The GSEA results linking KIF11 to cytoskeletal dynamics and T cell-mediated immunity further broaden its putative role beyond mere cell division, suggesting involvement in immune cell trafficking and activation within the psoriatic microenvironment. Nevertheless, given that the MR evidence for KIF11 did not survive multiple-testing correction and colocalization was not performed for KIF11 due to the absence of significant eQTL data, these findings remain speculative and require direct experimental validation. The predicted binding of 2-PHE to KIF11 at residues I136, E116, R119, Y211, and L214, as suggested by docking, provides a testable hypothesis but does not constitute evidence of a functional interaction. This hypothesis, however, awaits testing in appropriate cellular and animal models.

Mechanistic Synthesis: A Dual-Target Model

Taken together, the data are consistent with a working model in which 2-PHE might simultaneously engage NFKB1 and KIF11 to promote psoriasis-relevant cellular dysfunctions. On one hand, potential interaction with NFKB1 may potentiate inflammatory gene transcription, elevating cytokine output and sustaining immune cell recruitment to the skin. On the other hand, potential KIF11 engagement may accelerate keratinocyte proliferative cycling, compounding the epidermal hyperplasia that defines psoriatic plaques. These two axes—immune amplification and proliferative dysregulation—are not independent but rather mutually reinforcing in psoriasis pathophysiology, suggesting that even modest perturbation of both targets by a single environmental chemical could have disproportionately large biological consequences (Figure 8). However, this model remains hypothetical at this stage. The genetic evidence underlying this framework is preliminary, and the lack of colocalization support for NFKB1, together with the nominal nature of the MR findings, means that alternative explanations—such as confounding by linkage disequilibrium or other environmental factors—cannot be excluded. Moreover, the model assumes that 2-PHE reaches the relevant cellular compartments at sufficient concentrations, an assumption that lacks experimental support as discussed above.

Figure 8.

Mechanistic hypothesis diagram for this study. Gene names and key terms appearing in bold (eg, NFKB1, KIF11, Immune, Amplification, Proliferative, Dysregulation) are used for emphasis and do not indicate additional statistical or functional significance.

Limitations

Several important limitations should be considered when interpreting the findings of this study. First and foremost, this is a purely computational study. All results—including DEG identification, target prediction, PPI network construction, MR analysis, and molecular docking—are derived from in silico analyses using publicly available datasets and computational algorithms. None of the findings have been validated through wet‑lab experiments, such as cell‑based functional assays, animal models, or biochemical binding studies. Therefore, the results should be interpreted as hypothesis‑generating rather than confirmatory, and any mechanistic or clinical implications remain speculative until experimentally tested. Second, the 2-PHE targets employed in this study were computationally predicted from public databases rather than experimentally validated, introducing the possibility of false-positive interactions. Third, the transcriptomic data were derived from paired lesional and non-lesional samples within the same psoriasis patients (GSE30999) rather than from independent case and healthy control cohorts. This paired design, while useful for controlling inter-individual variation, means that our DEGs primarily reflect local inflammatory and hyperproliferative changes within diseased skin rather than systemic differences between psoriasis patients and healthy individuals. Fourth, while GSE41662 was used for validation of specific biomarkers, larger and more ethnically diverse cohorts would strengthen generalizability. Fifth, the MR analysis was limited by the small number of genetic instruments available for the hub genes, and the nominally significant findings did not survive Bonferroni correction for testing 13 genes. This substantially weakens the causal inference that can be drawn from these data. Sixth, the Bayesian colocalization analysis for NFKB1 yielded a global PP.H4 of only 0.02, indicating no robust evidence for a shared causal variant. This finding undermines one of the key genetic pillars of our original argument and suggests that the observed MR association may be driven by linkage disequilibrium or confounding rather than a genuine causal mechanism. Seventh, molecular docking simulations, while informative, represent static structural approximations and do not account for protein flexibility, solvation effects, or in vivo pharmacokinetic parameters. Eighth, despite the suggestive MR and docking evidence for NFKB1 and KIF11, causal inference from genetic instruments does not directly confirm that 2-PHE exposure operates through the same mechanistic pathways in environmental exposure contexts. Future cell-based and animal model experiments examining 2-PHE-induced transcriptional changes in keratinocytes and immune cells, as well as co-immunoprecipitation or thermal shift assays to confirm direct protein binding, will be essential to experimentally validate the mechanistic hypotheses generated here. Additionally, replication of the MR findings in independent psoriasis GWAS cohorts with larger sample sizes and more genetic instruments for the candidate genes is urgently needed.

Conclusions

In summary, this study provides exploratory evidence that 2-PHE may influence psoriasis pathogenesis through potential targeting of candidate risk genes NFKB1 and KIF11, thereby perturbing inflammatory and proliferative networks characteristic of the disease. However, the genetic evidence is limited: the MR findings did not withstand correction for multiple testing, and colocalization analysis did not support a shared causal variant at the NFKB1 locus. These results should therefore be interpreted as hypothesis-generating rather than confirmatory. While mechanistic confirmation will require further experimental work, our data highlight a plausible, though unproven, gene–environment axis linking PAH exposure to psoriasis and suggest that NFKB1 and KIF11 may warrant further investigation as both mechanistic nodes and potential biomarkers or therapeutic entry points in environmentally influenced psoriasis. Independent replication in larger cohorts and functional validation in appropriate experimental systems are essential before any definitive conclusions can be drawn.

Funding Statement

There is no funding to report.

Data Sharing Statement

The data utilized in this study were sourced from the GEO database, accessible at https://www.ncbi.nlm.nih.gov/geo/.

Ethics Approval and Consent to Participate

The present study made use of publicly available GEO datasets. According to the Declaration of Helsinki (2024 revision) and China’s Measures for the Ethical Review of Life Science and Medical Research Involving Humans (2023; Item 1 and 2 of Article 32), research that relies solely on fully anonymised and openly accessible data is exempt from further ethical review.

Disclosure

All authors declared that there was no conflict of interest.