Elucidating the role of EPPK1 in lung adenocarcinoma development

TMAs were prepared from the surgical specimens obtained from the Vanderbilt-Ingram Cancer Center Thoracic Biorepository. This biorepository houses tissue samples, along with sex as a biological variable, age, and clinical data from the Vanderbilt University Medical Center (VUMC) and the associated Veterans Affairs Hospital, Tennessee Valley Healthcare System. The TMA dataset comprised 35 cases for normal lung tissues and 295 patients diagnosed with NSCLC, including 140 patients with LUAD and 155 patients with lung squamous cell carcinoma (LUSC). The study received approval from the VUMC Institutional Review Board (IRB approval number: 000616), and all patients provided informed consent before their tumor specimens were collected. The study design did not involve randomization or blinding trials.

Lung cancer cell lines (A549, H2009, H1819, Calu3, H23, H1993, H226, H460, and H1299) and the normal bronchial epithelial cell line (BEAS2B) were obtained from the American Type Culture Collection, as previously described [7]. The cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Thermo Fisher Scientific, Waltham, MA, USA) or RPMI-1640 medium (Thermo Fisher Scientific), supplemented with 1% penicillin–streptomycin and 10% fetal bovine serum.

The expression of EPPK1 was assessed in the lung cancer cell line TMAs, normal lung TMAs, and normal bronchial epithelial cell line microarray. The TMAs were stained using antibodies for EPPK1 (#2–49597, rabbit polyclonal, 1:1000; NOVUS, Colorado, USA) and programmed death-ligand 1 (PD-L1) (#228415, clone 73–10, rabbit monoclonal, 1:500; Abcam, Cambridge, UK). The IHC score for EPPK1 was defined as cytoplasmic staining by multiplying the staining distribution scores of 0 (0%), 0.1 (1–9%), 0.5 (10–49%), 1 (50–100%) by the staining intensity score (0–3) [7]. The patients were classified into two equal groups based on the median score, indicating high and low EPPK1 expression. The IHC score for PD-L1 was defined as membrane staining present in at least 1% of cells, regardless of the staining distribution and intensity [13, 14]. The expression levels of EPPK1 and PD-L1 were evaluated by an experienced pathologist.

We investigated the EPPK1 gene alterations and mRNA expression in 36 different human cancer types using The Cancer Genome Atlas (TCGA) pan-cancer cohort, which encompasses a wide range of human cancer types. The objective was to examine their association with OS (UCSC Xena, RRID:SCR_018938) [15]. Moreover, this cohort allowed us to compare mRNA expression levels between normal and tumor tissues across 17 cancer types. For visualization and analysis, we used TIMER (RRID:SCR_018737), a tool specifically designed for analyzing data from TCGA [16]. To facilitate the evaluation of mRNA expression differences, we converted the expression values to the log2 of transcripts per million and assessed the variation between tumor tissues and normal tissues adjacent to tumor tissues.

Cell protein lysates were extracted, and WB analysis was conducted using a standard protocol [17]. The bands were visualized using specific antibodies against the following proteins: EPPK1 (#1488502, 1:1000, rabbit polyclonal, BioSource, Camarillo, California, USA), γH2AX (#97185, 1:1000, rabbit monoclonal, Cell Signaling Technology, Massachusetts, USA), E-cadherin (#610182, 1:1000, mouse monoclonal, BD Transduction Laboratories, California, USA), Vimentin (#5741, 1:1000, rabbit monoclonal, Cell Signaling Technology), c-MYC (#5605, 1:1000, D84C12, rabbit monoclonal, Cell Signaling Technology), p53 (#2527, 1:1000, 7F5, rabbit monoclonal, Cell Signaling Technology), and β-actin (#2228, 1:5000, mouse monoclonal, Sigma, Missouri, USA).

A total of 5 × 10⁵ BEAS2B control cells and BEAS2B cells exposed to smoking were seeded in a transwell with a 0.4 µm pore size and 500 µL of DMEM. Simultaneously, a 6-well plate (#3450, Corning Life Sciences, Massachusetts, USA) was filled with 1500 µL of DMEM, and the cells were cultured for 3 days before cigarette smoke exposure. Prior to smoke exposure, the medium in the transwell for both control and smoke-exposed cells was aspirated. Smoke-exposed cells were then placed in a smoking machine (Smoke Inhalation Unit 24, Promech Lab, Scania, Sweden), while control cells were kept in ambient conditions in an outer incubator during the smoking exposure period. The cells were exposed to cigarette smoke from three research cigarettes (3R4F; Cooperation Centre for Scientific Research Relative to Tobacco, Kentucky, USA) each weekday for four months. Following the cigarette smoke exposure, 500 µL of the medium was returned to each transwell. Weekly, each cell line was harvested, and the cells were seeded in a transwell again. The smoking density was measured using a Microdust Pro 880 nm Aerosol Monitoring device (Casella, New York, USA), and carbon monoxide levels in the smoking chamber were measured using the Auto Calibrating Carbon Monoxide detector (#UTLC11, UEi, California, USA). All experiments were performed in triplicate and repeated three times.

The cells were seeded in a chamber slide (Nunc Lab-Tek II Chamber Slide system, NUNC) and incubated overnight in a cell incubator. Subsequently, the cells were fixed with 4% formalin and stained with a mounting medium containing DAPI (#H1200, Vectashield mounting medium with DAPI, Vector Laboratories, California, USA) to visualize the nuclei. Immunostaining was performed using an antibody for γH2AX (#97185, 1:400, rabbit monoclonal, Cell Signaling Technology). After staining, the cells were examined in five imaging fields using a fluorescence microscope with 20 × magnification (IX51 Inverted Microscope, Olympus, Massachusetts, USA). The fluorescence intensity was measured, and the analysis was conducted using cell image analysis software (CellProfiler Image Analysis Software, RRID:SCR_007358). Each experiment was performed in triplicate.

Transwell inserts with an 8.0 µm pore size, placed in a 24-well plate (#3422, Corning Life Sciences), were coated with Matrigel (#356234, Corning Life Sciences) to create a basement membrane on both the upper and lower surfaces of the transwell membranes. A total of 5 × 10⁵ cells were seeded in the upper chamber of the transwell with 100 µL of DMEM, while the wells of the 24-well plate were filled with 500 µL of DMEM. After incubating the cells for 24 h, the cells on the upper surface of the transwell membrane were gently scraped off, and the cells on the lower surface of the membrane were fixed with 4% formalin and stained with a mounting medium containing DAPI. Cell images were captured in five fields of view using a fluorescence microscope with a 20x magnification, and the number of cells was quantified using cell image analysis software (ImageJ, RRID:SCR_003070). Each experiment was performed in triplicate.

Cells were seeded at a density of 1 × 10³ cells per well in 96-well plates (Thermo Fisher Scientific, #167008, Nunc MicroWell 96-Well Microplates). The cells were then cultured for 2, 4, and 6 days. To measure cell counts, the CyQUANT assay (Thermo Fisher Scientific, #C7026, CyQUANT Cell Proliferation Assay) was used, and the fluorescence was measured using a plate reader. Each experiment was independently repeated three times, with six wells analyzed per experiment.

RNA was extracted using the RNeasy Plus Mini Kit (#74034; Qiagen, Redwood City, CA, USA). The RNA sequencing libraries contained 300 ng of RNA and were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit (#E7760L, New England Bio Labs, County Road, MA, USA). Fragmentation, cDNA synthesis, end-repair/dA-tailing, adaptor ligation, and PCR enrichment were performed according to the manufacturer’s instructions. The quality of individual libraries was assessed using an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA), and quantification was performed using a Qubit Fluorometer (Thermo Fisher Scientific). The adapter-ligated material was estimated using qPCR before normalization and pooling for sequencing. The libraries were sequenced on a NovaSeq 6000 platform using 150 bp paired-end reads. Base calling was performed using RTA (Illumina, San Diego, CA, USA), and data quality control was conducted using the Vanderbilt Technologies for Advanced Genomics core facility. Poor-quality reads were trimmed, and adapter sequences were removed. The reads were then aligned to the human genome using STAR (STAR, RRID:SCR_004463) [18] and quantified using featureCounts [19]. Poor-quality RNA-Seq experiments were excluded from the analysis. Differential gene expression analysis was performed using the R package DESeq2 (DESeq, RRID:SCR_000154) [20]. A heat map was generated, and gene ontology (GO) enrichment analysis with a false discovery ratio was conducted using the R package to visualize gene expression across the samples. All experiments were performed in triplicate.

To investigate the potential association of EPPK1 with OS across multiple cancers, we analyzed mRNA expression data from TCGA database. Our findings revealed a significant correlation between EPPK1 expression and poor prognosis in various cancers (see Table 1). These results indicate that EPPK1 may serve as a promising prognostic biomarker for these malignancies. Furthermore, we examined the mRNA expression of EPPK1 to assess its relationship with tumor development. By comparing EPPK1 mRNA expression levels between normal and tumor tissues from multiple cancer types in TCGA dataset, we made several noteworthy observations. Firstly, we observed elevated expression of EPPK1 mRNA in LUAD (P < 0.01, see Fig. 1A) and LUSC (P < 0.001, see Fig. 1A) as well as in several other cancers, in comparison to that in normal tissues (see Fig. 1A). These findings highlight the potential significance of EPPK1 in progression across different cancer types.

To evaluate EPPK1 as a potential biomarker, we analyzed its protein expression in 280 patients with NSCLC, including 140 with LUAD and 140 with LUSC. We examined the association among EPPK1 protein expression, clinical variables, and OS. Patient characteristics, such as age, gender, smoking history, histology, pathological stage, and chronic obstructive pulmonary disease status, were summarized by histology and EPPK1 protein expression (Tables 2, 3 and 4). EPPK1 protein was found to be expressed in the cytoplasm of resected lung cancer tissue (Fig. 1B) and lung cancer cell lines (Fig. S1). However, EPPK1 was not expressed in resected normal lung tissue or normal bronchial epithelial cell line microarrays (Fig. S2). We did not observe any relevant interaction effect between EPPK1 and PD-L1 status (Fig. S3) or gene alterations. However, in patients with LUAD, the interaction effect between EPPK1 and stage (early vs. late stage) was associated with OS in the multivariable Cox regression model adjusted for gender and pack-years (P < 0.01, LUAD in Table 5). This indicates that the impact of EPPK1 on OS differed between the early and late stages. In the early stages, there was a significant difference in survival between the EPPK1-high and -low groups (Fig. 1C, log-rank test, P = 0.024 in stage I; P = 0.022 in stages I and II; Fig. 1C). Patients in the EPPK1-high group had a 2.431 times higher likelihood of death than those in the low EPPK1 expression group in stage I (HR with CI = 2.431 [1.097–5.39], Wald test: P = 0.03, Fig. 1C), and a 2.085 times higher likelihood in stages I and II (HR with CI = 2.085 [1.1–3.951], Wald test: P = 0.02 in stages I and II, Fig. 1C). After adjusting for gender in the multivariable analysis, the risk of death in the high-expression EPPK1 group was reduced to 2.008 times higher than that in the low-expression EPPK1 group in stage I (adjusted [adj.] HR with CI = 2.008 [0.890–4.530], Wald test: P = 0.09 in stage I) and 2.619 times higher in stages I and II (adj. HR with CI = 1.619 [0.839–3.128], Wald test: P = 0.15 in stages I and II). In late-stage LUAD, there was no difference in OS between the EPPK1-high and -low groups after adjusting for gender and pack-years using multivariable Cox regression analysis. Additionally, there was no difference in OS when considering the interaction effect between EPPK1 and stage or the main effect of EPPK1 in the LUSC group (adj. P = 0.96, LUSC in Table 5). These findings suggest that EPPK1 expression plays a critical role in cancer development in LUAD and could serve as a prognostic biomarker for early-stage LUAD.

Our clinical data revealed that patients with high expression of EPPK1 had a higher smoking pack-year history than those with low expression (Difference with CI = 10 [0.00–21.00], P = 0.032 in LUAD, Table 3; Difference with CI = 20 [10.00–28.00], P < 0.001 in LUSC, Table 4). To further investigate the impact of smoking on EPPK1 expression, we exposed BEAS2B, a normal bronchial epithelial cell line, to cigarette smoke for 16 weeks. The average smoking density and maximum carbon monoxide level during exposure were 1.8 mg/l and 121 ppm, respectively. As a positive control for smoking exposure, we initially examined γH2AX, a biomarker for DNA double-stranded breaks. We confirmed an increase in γH2AX levels after 16 weeks of smoking exposure using WB and immunofluorescence, validating the DNA damage caused by our smoking method (Fig. 2A, B, mean difference with CI = 11.705 [8.746–14.664], Fig. 2C). Subsequently, we evaluated EPPK1 expression in normal bronchial epithelial cells following 16 weeks of smoking. Our results demonstrated a correlation between EPPK1 expression and smoking (Fig. 2D), consistent with our clinical study where EPPK1 expression significantly correlated with pack-years (Tables 3 and 4).

Based on our findings indicating the critical role of EPPK1 in cancer development and its association with poor prognosis in LUAD and other cancers, we aimed to investigate whether the loss of EPPK1 function could impede cancer progression. Initially, we assessed EPPK1 protein expression in NSCLC cell lines using WB analysis (Fig. 3A). Additionally, we verified EPPK1 protein expression in our cell line microarray through IHC (Fig. S1B). To explore the impact of EPPK1 on LUAD progression, we conducted a loss-of-function study using CRISPR-Cas9 for all LUAD cell lines in Fig. 3A. We successfully generated KO of EPPK1 to investigate epithelial to mesenchymal transition (EMT), its reverse process mesenchymal to epithelial transition (MET) [21], and angiogenesis in the A549 LUAD cell line. Our analysis revealed that the KO of EPPK1 led to increased expression of E-cadherin and p53, while Vimentin, and MYC expression decreased, as demonstrated by WB analysis (Fig. 3B). These findings indicated a MET process, as evidenced by the upregulation of E-cadherin, downregulation of Vimentin, upregulation of tumor suppressor proteins, and decreased expression of oncogenic proteins following EPPK1 silencing. Overall, our results strongly suggested that EPPK1 promotes EMT and lung cancer development.

Based on our data indicating the association of EPPK1 with cancer development and EMT we conducted an analysis of cell growth and invasion following the KO of EPPK1. To elucidate the mechanisms through which EPPK1 regulates cell proliferation, we compared the proliferation of WT and KO cells at 2, 4, and 6 days. Our findings revealed that EPPK1 KO cells had a slower proliferation rate than WT cells (mean difference with CI = 2716.851 [187.670–5246.032], P = 0.037, Fig. 3C). To examine the impact of EPPK1 regulation on cell invasion, we conducted a comparison of cell invasion between WT and KO cells. The results of invasion assays demonstrated that EPPK1 KO cells exhibited a reduced invasion ability compared to that of A549 WT cells (mean difference with CI = 2.110 [1.817–2.402], P < 0.001, Fig. 3D, E).

Based on our accumulated evidence indicating the involvement of EPPK1 in cancer development, EMT, cell proliferation, and cell invasion, we employed RNA sequencing to examine the expression levels of WT and KO EPPK1. Out of the 15,282 genes analyzed, 2,201 (14%) exhibited alterations following EPPK1 KO. Specifically, after KO, the expression of 11 oncogenes, 75 anti-apoptosis genes, and 22 angiogenesis genes was downregulated, while the expression of 8 tumor suppressor genes and 12 anti-cell growth genes was upregulated (Fig. 4A–D). Furthermore, performing GO enrichment analysis on selected molecular functions of differentially expressed genes between WT and KO cells, we observed that WT cells exhibited enhanced regulation of mesenchymal cell proliferation, mesenchymal differentiation, angiogenesis, and cell growth in comparison to those in KO cells (Fig. 5). These results serve to underscore the involvement of EPPK1 in tumorigenesis, EMT, angiogenesis, and cell growth.