2.1. Transcriptome Data Analysis

Public datasets GSE101720, GSE19190, and GSE46171 were obtained from the NCBI Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/geo/). Raw data were preprocessed by log₂ transformation and quantile normalization, and the batch effects were corrected using empirical Bayes methods implemented in a sva R package “Combat” [14]. All data analyses were conducted in the RStudio environment (version 4.1.0). Samples with asthmatic tissues were excluded from the datasets (Supplementary Table 1). HC and AR consisted of the nasal mucosa and nasal airway epithelium.

Gene set analyses were performed using the gProfileR (0.7.0) package in R software and data from gene ontology: biological processes (GO: BP, http://geneontology.org/) and REAC (http://www.reactome.org/) databases. Coordinated gene regulation was identified using GSEA (https://www.gsea-msigdb.org/gsea, version 3.0) [15]. A genetic network was constructed using the GeneMANIA software in Cytoscape (version 3.4.1) [16].

2.2. Prediction of Transcription Factor Using In Silico Databases

Transcription factors were predicted using ChIP-X Enrichment Analysis version 3 (ChEA3, https://maayanlab.cloud/chea3/). ChEA3 was performed as previously described [17]. Transcription factor binding sites were analyzed by constructing matrices of immune-related genes analyzed using AliBaba2.1 (http://gene-regulation.com/pub/programs/alibaba2/index.html) [18].

2.3. Cell Culture and Cytokine Treatment

The human nasal epithelial cell line (HNEpC) purchased from Promocell (Heidelberg, Germany) was mainly grown in DetachKit (Promocell) at 37°C in a humidified atmosphere with 5% CO₂. The cells were treated with inflammatory cytokines such as IL-1β for 24 h.

2.4. Cloning and Real-Time PCR

For the cloning of tagged KLF4 coding gene sequences into pcDNA3.1 (-) (Addgene, Cambridge, MA, USA), the KLF4 gene was amplified using total RNA extracted from the HEK293 cell line. PCR was performed using specific primers containing a 5′-extension and XbaI (NEB, Ipswich, MA, USA) and NotI (NEB) restriction sites with CloneAmp HiFi PCR Premix (Thermo Fisher Scientific, San Jose, CA, USA). Subsequently, the amplicons were digested with XbaI and NotI and cloned into the pcDNA3.1 (-) vector using an In-Fusion® cloning system according to the manufacturer’s recommendations.

HNEpC cells were transfected with pcDNA3.1 (-) containing the tagged KLF4 coding sequence using Lipofectamine® 3000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s recommendations. Overexpression effects were validated by quantitative real-time PCR analysis using the iQ™ SYBR Supermix (Bio-Rad, CA, USA). The primers used in this study are summarized in Supplementary Table 2.

2.5. Statistical Analysis

Statistical analysis was performed using R software (version 3.4.0; Vienna, Austria). Statistical analyses were performed using Student’s t-test, permutation t-test, or one-way ANOVA test.

3.1. Differentially Expressed Genes between HC and AR

To acquire robust results with an extended HR and AR sample size, we first selected the independent transcriptome cohorts (e.g., GSE101720, GSE19190, and GSE46171), including 43 cases of HC and 39 cases of AR only composed of the nasal mucosa and nasal airway epithelium using GEO (n = 82). Next, we established the pooled transcriptome data using GEO datasets (Figure 1(a)). Then, we analyzed the differentially expressed genes (DEGs) between the HC and AR samples and classified them as rhinitis signatures “up” (n = 237) or “down” (n = 197) (P < 0.05, permutation t-test with a fold change (FC) > 0.3; Figure 1(b) and Supplementary Table 3), depending on whether they were upregulated or downregulated genes, respectively. Additionally, we sought to identify the underlying key signatures that potentially drive the phenotype difference between the groups. We then evaluated DEG-related signatures using GO: BP (P < 0.01, maximum P value with a minimum set size > 10). Gene set enrichment analysis revealed the enriched expression of immune- and cytokine-related genes in AR. We found that the tissue development-related signatures were frequently contained in the “down” genes (Figure 1(c)). Next, we validated the signatures of the DEGs. By applying the REAC database, we found that immune system-related genes were enriched in AR. In contrast, biological oxidation, glutathione conjugation, and Ca²⁺ pathway-related genes were enriched in HC (Figure 1(d)). Considering these results, we suggest that the immune system plays a pivotal role in AR pathogenesis.

3.2. Immune Response-Related Genes Are Enriched in AR

Next, we focused on the underlying mechanisms involved in the altered expression of immune- and cytokine-related genes. Many immune-related genes were “response to cytokine”-dependent genes. We then sought to identify downstream players that drive the immune response with response to cytokines which had a significant correlation with the identified immune response signature and DEGs (n = 24, P < 0.05, permutation t-test with FC > 0.3, Figure 2(a)) and found that the enrichment score (ES) of the identified genes was significantly increased for AR compared with HC (P < 0.05, Figure 2(b), top). Among the immune-related DEGs, IL1RL1 (FC = 1.12), CD274 (FC = 0.90), and CD44 (FC = 0.40) were the most significantly altered genes between the HC and AR groups (Figure 2(b), bottom). In addition, we evaluated the expression of a previously known biomarker of AR, POSTN [9]. The AR group exhibited high expression of POSTN compared with the HC group (Figure 2(c)).

As the AR samples showed different expression patterns of immune response genes, we performed consensus clustering (median absolute deviation > 0.7, n = 2,804). We analyzed the proportion of ambiguous clustering (PAC) score using the CDF curve and found that the optimal K was 2 with the lowest PAC value (Supplementary Figure 1(a)). Principal component analysis also revealed that AR samples were distributed in two separated fields (Supplementary Figure 1(b)). Based on these results, we classified AR into two robust subtypes, AR1 (n = 21) and AR2 (n = 18). We found that most of the immune- and cytokine-related signatures were significantly altered between AR1 and AR2 (Figure 2(d)), revealing a gradual increase in the expression of top genes and rhinitis marker genes (e.g., IL1RL1, CD274, and CD44) from healthy controls to AR1 and then AR2 (Figure 2(e)) [19, 20]. When we calculated the enrichment score (ES) for each group on the differentially expressed genes of immune cell-related signatures (Supplementary Table 4), we observed significantly higher ES of monocyte and dendritic cells in AR2 samples than other groups (P < 0.001, one-way ANOVA test, Figure 2(f)). Indeed, monocyte and dendritic cells are known to associate with allergic rhinitis and asthma [21, 22]. Moreover, we found that IL1B gene expression, a proinflammatory cytokine induced by monocyte and dendritic cells [23], was enhanced in AR2 compared to other groups (Figure 2(g)). These results indicate that AR samples show heterogeneous molecular subtypes, including AR1 and AR2.

3.3. Profiling of Transcription Factors Associated with Immune Response on HC and AR

As the transcription factor (TF) plays potential driver roles in immune response, we investigated the underlying TFs and the putative regulatory elements of immune-related genes. We identified 1,632 TFs predicted to bind to the promoter region of immune response-related genes (n = 54) using ChEA3 (see, for details, Materials and Methods, Figures 3(a) and 3(b)). Of these, we found that KLF4, GATA2, FOXA3, SPDEF, ZMAT4, and HES4 were significantly altered between the HC and AR groups (P < 0.05, FC > 0.5, Figure 3(c)). Among them, KLF4 showed the highest correlation value with immune-related genes compared to other transcription factors (r = 0.53, P < 3.27 × 10⁻⁷, Supplementary Figure 2). Furthermore, we observed that KLF4 is significantly associated with the ES of monocyte- and dendritic cell-related genes compared to other transcription factors, respectively (r > 0.5, P < 10⁻⁷, Figure 3(d)).

The previous results demonstrated that AR2 was related to immune response, reflecting high expression of IL1B, a proinflammatory cytokine induced by monocyte and dendritic cells (see Figure 2). Thus, we evaluated whether IL1B is associated with KLF4 expression. Next, we confirmed these findings via cell culture experiments, wherein treatment of IL-1β into a human nasal epithelial cell line significantly enhanced KLF4 expression (P = 0.0187, Figure 3(e)). In contrast, other transcription factors such as GATA2, SPDEF, and HES4 showed no effects. We also verified that gene set enrichment analysis revealed an enriched expression of NF-κB and STAT3-related genes in the AR2, KLF4 expression high group, compared to the HC and AR2 groups, respectively (Figures 3(f) and 3(g)). Indeed, IL-1β in immune response is known to induce the activation of the NF-κB and STAT3 pathway [24, 25]. These findings strongly indicated that KLF4 expression was associated with immune response via IL1B expression.

3.4. KLF4 Transcription Factor Is Associated with the Immune Signature of AR

We also constructed a genetic network of immune response-related genes with KLF4 and found that KLF4 had a highly significant interaction with immune response-related genes (IL1RL1, CD274, CD44, etc., 32/54, 59.26%, Figure 4(a)). We also observed that KLF4 binds to the CD274, CD44, and IL1RL1 gene promoter motifs (-2,000 to +100 bp from the transcription start site) (Figure 4(b), top). KLF4 binding domains showed sequences containing core 5′-CACCC-3′ elements (Figure 4(b), bottom). In addition, KLF4 expression gradually increased from HC to AR1 and then AR2 (P < 0.01, Supplementary Figure 3).

To verify whether KLF4 expression can affect the immune response-related genes, we established a KLF4 overexpression system in a human nasal epithelial cell line (HNEpC) that has lower expression levels of KLF4 (P < 10⁻⁶, Figure 4(c)). KLF4 overexpression cells could induce IL1RL1, CD274, and CD44 genes (P < 10⁻⁶, Figure 4(d)). Thus, it is plausible that KLF4 expression enhances immune response-related genes, resulting in immune response-related signatures.