The CHST11 gene is linked to lung cancer and pulmonary fibrosis

2.1 Clinical data on lung cancer and pulmonary fibrosis

Gene Expression Omnibus (GEO) datasets, including GSE108134, GSE11906, GSE40839, GSE44723, and GSE31210, were downloaded from NCBI (https://www.ncbi.nlm.nih.gov/gds). GeneSpring GX software was utilized to process all CEL files.

2.2 Correlation of the pathological features of CHST11 in LUAD

Data regarding pathophysiological features and their relationship to CHST11 expression were obtained from UCSC Xena’s website (https://xenabrowser.net/).

2.3 Correlation between gene expression and prognosis in lung cancer

An analysis of the correlation of gene expression in lung cancer was obtained from the GEPIA2 website (http://gepia2.cancer-pku.cn/#index). Kaplan–Meier plots (https://kmplot.com/analysis/) were used to determine the correlation between gene expression and the prognosis of lung cancer patients. The data on gene expression in lung cancers were retrieved from the Cancer Cell Line Encyclopedia (CCLE; https://sites.broadinstitute.org/ccle/).

2.4 Model for the molecular and pharmaceutical simulation of CHST11

Molecules related to CHST11 were downloaded from the Cbioportal website (https://www.cbioportal.org/). The Cancer Genome Atlas (TCGA)-Firehose Legacy, TCGA-Nature 2014, and TCGA-PanCancer Atlas data were used. Correlated molecules were defined as those with Spearman’s correlations of ±0.3. A Venn diagram was used to group molecules that cross between different datasets. Ingenuity pathway analysis (IPA; https://analysis.ingenuity.com/pa/installer/select) provided an analysis of the signaling pathways and gene ontologies that may be impacted by the molecules. Drug repurposing mining used the Venn diagram method to identify genes linked to CHST11, which were then submitted to the L1000 CDS website for analysis.

2.5 Establishment of a two-way model based on CHST11

pLenti-6.3-CHST11 cDNA and pLKO.1-CHST11 shRNA were obtained from DNASU (Arizona State University, USA) and the National RNAi core facility (Academia Sinica, Taiwan). Lentivirus encapsulation was used to produce the virus in HEK293 cells. Cell lines associated with infection-related lung adenocarcinoma, including A549, H441, H358 and H1975, were infected with the virus by a TOOLSFect transfection reagent TTM-TF01 (BIOTOOLS, Taiwan). The infected cells were further selected for puromycin or blasticidin, and the survival of the infected cells was confirmed by immunoblotting.

2.6 Immunoblotting and qPCR assays

All lung cancer cell lines and human pulmonary fibroblast cell lines were obtained from ATCC (The Global Bioresource Center, USA) and ScienCell (ScienCell Research Laboratories, USA). The cells were cultured according to official guidelines. Culture medium containing 0.05% penicillin–streptomycin antibiotics (Gibco, USA) and fetal bovine serum (Gibco, USA) was used to maintain the tumor cells, which were cultured in a constant temperature at 37°C and 5% CO₂. The cells were treated with different drugs and recombinant proteins, including nicotine (Sigma–Aldrich, USA), particulate matter (PM; 1649b, NIST® SRM®, Gaithersburg, USA), bleomycin (MedChemExpress, USA), recombinant TGF-β protein (Sino Biological, USA), recombinant INFγ protein (Sino Biological, USA) and ITGB inhibitors, including AG-1478, brefeldin A, geldanamycin, importazole and NVP-TAE 684 (MedChemExpress, USA). RNA was collected using an RNA kit (Qiagen, Germany) and translated into cDNA using a TOOLSQuant II Fast RT Kit (Biotools, Taiwan), and mRNA levels were calculated using qPCR. RIPA buffer was used to collect the proteins, and immunoblotting was used to analyze them.

2.7 Proliferation and motility assays for tumor cells

Alamar blue (Sigma–Aldrich, Germany) was used to measure the effect of CHST11 on the proliferation of cells. The appropriate number of cells was seeded on 96-well plates, and the growth trends were assessed over a period of 6 days. Cell migration and invasion were evaluated in a Boyden chamber. In the same manner, appropriate cells were relysed in serum-free medium and seeded in an upper chamber. In the lower chamber, fibronectin (Sigma–Aldrich, Germany) and Matrigel (Corning, USA) were coated with attracting serum medium to evaluate the migration or invasion ability of cells mediated by CHST11. A solution of Giemsa was used for visualization and the polycarbonate membrane filter was exposed under an optical microscope (400×, three randomly selected fields per/well).

2.8 Statistical analysis

Excel was used to perform the statistical analysis using the two-tailed Student’s t-test. The t-test was used to assess the statistical significance of the difference between groups as follows: * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.

3.1 The expression of CHST11 is associated with smoking-related lung cancer

To identify whether an elevation of CHST11 correlates with lung cancer progression, we analyzed several lung cancer cohorts to determine the correlation between CHST11 expression and overall survival. An analysis of the Kaplan–Meier plot demonstrated that the expression of CHST11 was significantly associated with poor prognosis in lung cancer (p = 0.0000053; Figure 1A). Furthermore, the results revealed that CHST11 was only associated with a poor prognosis for subtypes of LUAD (p = 0.0000026) but not lung squamous cell carcinoma (LUSC; p = 0.31). CHST11 is an enzyme involved in the biosynthesis of CS. Among the other enzymes that contribute to chondroitin sulfate modification are CHST3, CHST7, CHST10, CHST12, CHST13, CHST14 and CHST15 [19]. To determine whether CHST11 is the only enzyme involved in LUAD progression, other vital enzymes involved in the biosynthesis of chondroitin sulfate were also examined. Based on Kaplan–Meier plot analysis, CHST11 was the only molecule associated with poor prognosis in LUAD (p = 0.0000026; Figure 1B). Other clinical cohorts (GSE31210) also yielded consistent results (Figure S1).

To determine whether the increased level of CHST11 was due to environmental or genetic changes, two LUAD datasets (TCGA and GSE31210) were used. As shown in Figure 1C, smoker lung cancer patients had higher levels of CHST11 expression than non-smokers (p = 0.0076; Supplementary Table 1), especially those who had smoked for 15+ years. In terms of genetic changes, mutations of the EGFR, KRAS and ALK genes were highly associated with the occurrence of spontaneous lung cancers. As indicated by the analysis of LUAD clinical data (GSE31210), CHST11 levels were decreased in patients who had ALK fusions and EGFR mutations and increased in patients who had KRAS mutations and EGFR/KRAS/ALK triple-negative cases (Figure 1D; Supplementary Table 2). However, CHST11 expression did not differ significantly between KRAS mutations and non-genetic variations. Thus, the level of CHST11 may be related to smoking rather than genetics. Furthermore, data from TCGA indicate that the expression of CHST11 increases with stage progression, especially between stage I and stage IA (p = 0.0299; Figure 2E; Supplementary Table 3). These results suggest that CHST11 may be associated with smoking-induced fibrosis but also participates in the progression of smoking-related lung cancer, particularly LUAD.

3.2 Increased expression of CHST11 is associated with pulmonary fibrosis

There is still controversy regarding the detailed mechanism of how pulmonary fibrosis and lung cancer develop. The presence of lung fibrosis is currently considered one of the risk factors for lung cancer [8]. It may be caused by environmental factors such as smoking and is also one of the factors that contribute to lung cancer [27]. By investigating whether CHST11 is involved in both fibrosis and cancer, we may be able to distinguish between the two symptoms or find a target that can be used in conjunction to treat both conditions. To investigate the relationship between CHST11 and lung fibrosis, a number of clinically relevant GEO datasets with primary lung lesions were utilized, including smoker (GSE108134), chronic obstructive pulmonary disease (COPD; GSE11906) and usual interstitial pneumonia (UIP; GSE40839; Supplementary Table 4). To confirm that these samples were indeed related to pulmonary fibrosis, the levels of fibrosis markers such as fibrosis factor (TGF-β) and activated fibroblast markers (FAP, COL1A1, ACTA2 and FN1) were assessed (Figure S2A). Based on these data, the smoking group had higher CHST11 expression than the non-smoking group (p = 0.0167). The patients with COPD also had higher CHST11 expression than the non-smoking group (p = 0.0575). In pulmonary fibrosis of unknown etiology, such as UPI, CHST11 expression was also found to increase (p = 0.0306; Figure S2A). The heatmap analysis indicated that although there were no significant differences between the COPD groups, the expression of CHST11 was elevated in all pulmonary fibrosis cases, consistent with the smoker and UPI groups (Figure S2B).

To determine whether the upregulation of CHST11 is caused by the stimulation of fibrosis factors, human pulmonary fibroblasts (HPF) were treated with fibrosis inducers, including nicotine, ambient PM and bleomycin. The results demonstrated that diverse stimulators activated the related fibrosis markers (such as a-SMA, FAP, Col I and Col II) as well as CHST11 expression. Overall, these results suggest that CHST11 may also play an essential role in the development of pulmonary fibrosis, in addition to its role as a lung cancer diagnostic marker.

3.3 CHST11 levels affect lung cancer malignancy characteristics

To confirm that CHST11 is involved in lung cancer progression, a CHST11-mediated two-way model was developed. The distribution of CHST11 in cancer cells was evaluated using the cancer model (Figure 2A) (Figure S8). Compared with normal lung fibroblast cells, BEAS-2B lung cancer cells expressed more CHST11 (Figure 2B). To establish a CHST11 two-way model, cells with knockdown and those with overexpression of CHST11 were selected based on their expression of CHST11 (Figure 2C) (Figure S8). For the biological assays, the results showed that when cells overexpressed CHST11, the growth rate increased, while the growth rate of cells with CHST11 knockdown decreased (Figure 2D). The analysis of cell motility also revealed that CHST11 increases the motility of cells, including their ability to migrate and invade. As shown in Figure 2E, knockdown of H1975 with different shCHST11 constructs also reduced the motility of cells. Based on the results presented above, CHST11 can not only function as a diagnostic factor but also participate in lung cancer malignancy.

3.4 CHST11 regulation is involved in the processes of pulmonary fibrosis and cancer development

The molecular mechanism of CHST11 in pulmonary fibrosis or lung cancer has not been studied to date. To reveal the signaling hallmark that CHST11 may be responsible for regulating pulmonary fibrosis and lung cancer progression, a molecular simulation analysis was conducted [28]. This analysis used three TCGA-LUAD datasets: TCGA-Firehose Legacy, TCGA-Nature 2014 and TCGA-PanCancer Atlas. In lung cancer, the Venn diagram was integrated with CHST11 based on a Spearman’s correlation greater than ±0.3 as a valid basis parameter. In total, 1,176 positive and 317 negative genes were collected (Figure 3A; Supplementary Table 5). By utilizing the IPA database, we were able to simulate possible gene ontologies and the mutual regulatory relationships between the molecules (Figure 3B). These signaling hallmarks are illustrated in Figure 3C based on the ranking of the p-values (Supplementary Table 6). The signaling hallmark ranking suggests that CHST11-related molecules participate in both tumor-related signaling and pulmonary fibrosis idiopathic signaling as well as hepatic fibrosis/hepatic stellate cell activation and hepatic fibrosis signaling pathways. These results support the relevance and usefulness of this molecular simulation by emphasizing CHST11’s potential mechanisms in pulmonary fibrosis and LUAD progression.

3.5 The CHST11 gene primarily regulates fibrosis and cancer motility events related to cell motility

Fibrosis is defined as the transformation of epithelial cells into myofibroblasts following the EMT, which is also important in the metastatic progression of tumor cells [11]. To determine whether the increased CHST11 level in pulmonary fibrosis correlates with cellular motility, datasets from GSE44723 were analyzed. The clinical data (GSE44723) showed that CHST11 was expressed at a high level in the fibroblast-rapid group compared with the fibroblast-low group (Figure S3A) (Supplementary Table 7). Cells stimulated by different fibrosis activators also transactivate factors associated with EMT, such as vimentin (Figure S3B), which supports the level of CHST11 associated with EMT processes during fibrosis occurrence. Consistently, TCGA-LUAD data demonstrated that CHST11 was positively correlated (Spearman’s correlations >0.3) with several mesenchymal factors in LUAD, including VIM, TWIST1, TWIST2, SNAI1 and SNAI2, suggesting a role for CHST11 in EMT (Figure S3C). In cancer cells overexpressing CHST11, mesenchymal factors, including snail1, slug, twist, zeb1, zo1, goosecoid, mmp2 and mmp9, are also increased (Figure S3D). Accordingly, the molecular simulations supported that CHST11-related molecules are indeed involved in many EMT-related signaling pathways, including the Wound Healing Signaling Pathway, Integrin Signaling, Signaling by Rho Family GTPases, Regulation of the Epithelial Mesenchymal Transition by Growth Factors Pathway, Axonal Guidance Signaling, Regulation of Actin-based Motility by Rho, Colorectal Cancer Metastasis Signaling, Actin Cytoskeleton Signaling and ILK Signaling (Figure 3C). Thus, CHST11 may contribute to LUAD progression during the EMT process and may be involved in pulmonary fibrosis events.

3.6 CHST11 expression is mainly induced by TGF-β  and INFγ

For an unbiased analysis, EMT signaling was first aligned with molecules that are known to play a role in CHST11 regulation. Four molecules were identified, ERBB2, IFNG, MMP9 and TGF-β (Figure 4A). According to clinical correlation, these molecules are correlated with CHST11 in LUAD. The molecules with Spearman’s correlation values greater than 0.3 were IFNG (R = 0.41) and MMP9 (R = 0.45) (Figure 4B). A comparison of IFNG and MMP9 as signatures with EMT-related molecules revealed a significant correlation with LUAD (R = 0.59) (Figure 4C). To confirm that clinical correlations can be seen in LUAD cell lines, we utilized CCLE datasets. The results indicated a strong correlation between IFNG (R = 0.787) or MMP9 (R = 0.47) and CHST11 in LUAD at the cellular level (Figure 4D; Supplementary Table 8). As a result of treatment with recombinant TGF-β and INFγ proteins, cells were able to further activate the expression of CHST11 (Figure 4E).

To assess whether such regulatory relationships are similar to those of the fibrosis model, GEO datasets related to fibrosis (GSE11906, GSE108134, GSE40839) were used to identify CHST11-related molecules (Figure S4A). Based on the Venn diagram results, approximately 13 molecules are collectively involved in different fibrosis events caused by CHST11. As analyzed by IPA, TGF-β and INFγ share a molecular relationship with CHST11 in both fibrosis and cancer models. It was found that CHST11 expression was increased when HPF cells were stimulated by both TGF-β and INFγ (Figure S4C), which is consistent with the tumor cell response. Additionally, molecules involved in the CHST11 pathway were consistently activated by fibrosis activators (Figure S4D). It should be noted, however, that downstream regulators of CHST11 are distinct from upstream regulators in fibrosis (Figure S4D) and tumor events (Figure 4A). These findings suggest that CHST11 may contribute to pulmonary fibrosis or LUAD progression by regulating EMT, which can be stimulated by TGF-β or INFγ, triggering different downstream CHST11 effectors.

3.7 CHST11-related molecules primarily function through integrin signaling

To identify CHST11-mediated regulation involved in specific signaling, a WordCloud analysis was conducted based on EMT-related signaling (Figure 5A; Supplementary Table 9). Based on correlation analysis, the top 10 repeatable molecules with high correlations in LUAD were IFNG and MMP9 (R = 0.52; Figure 5B). In LUAD, we found a strong clinical correlation (greater than 0.30 Spearman’s correlation) between these highly repetitive EMT molecules and CHST11/IFNG/MMP9. Among the IFNG-related genes were PIK3CG (R = 0.36), PIK3R5 (R = 0.36), ITGB7 (R = 0.59), and RAC2 (R = 0.44). Genes associated with MMP9 included PIK3CG (R = 0.36), PIK3R5 (R = 0.36), GRB2 (R = 0.34), MRAS (R = 0.32), ITGB7 (R = 0.51), RAC2 (R = 0.44) and ACTA2 (R = 0.33) (Figure 5C). According to the CCLE database, these genes were significantly correlated with CHST11, IFNG and MMP9 gene expression (Figure 5D). Based on these observations, we conclude that most of these highly related molecules play important roles in the integrin signaling process. IPA also revealed that CHST11-related molecules are highly correlated with integrin signaling (red indicates a positive correlation; Figure S5). By analyzing Kaplan–Meier plots, it was discovered that the majority of these integrin-related molecules were associated with a poorer prognosis for LUAD, including IFNG (p = 0.0013), MMP9 (p = 0.009), PIK3R5 (p = 0.0000000000063), ITGA6 (p = 0.00017), ITGB1 (p = 0.0099), GRB2 (p = 0.018), MRAS (p = 0.049), RALB (p = 0.00038), RAC2 (p = 0.000000051) and ACTA2 (p = 0.015) (Figure S6). These results demonstrated that CHST11-mediated IFNG and MMP9 could affect cellular EMT by regulating integrin signaling, which may contribute to a poor prognosis in LUAD.

3.8 CHST11 signature-based in-silico drug repurposing

To identify potential pharmaceuticals targeting the CHST11/integrin pathway, 1,176 positive and 317 negative genes that correlate with CHST11 (Figure 3A; Supplementary Table 5) were submitted for analysis as signatures to the ultrafast LINCS L1000 characteristic direction signature search engine. The analysis revealed that approximately 50 drugs were capable of reversing CHST11-positive molecules (Figure 6A; Supplementary Table 10). Most of these drugs have been tested in relative doses in LUAD cell lines, such as A549 and HCC515. The related compounds include sodium dithionite, TG101348, FCCP, WH-4-025, malonoben, IMD 0354, Selamectin, Vorinostat, BRD-A17065207, NVP-TAE684, AG-879, PD-407824, niclosamide, I-BET151, PD-173074, CCCP and CGP 71683 hydrochloride. It is noteworthy that there are a number of drugs that more specifically target integrin signaling-related molecules, including ACTA2, ITGA6, RALB and MMP9 (Figure 6B). Among the drugs in this class that are capable of cross-targeting are 15-deoxy-delta-12, 14-prostaglandin J2 (ITGA6, ACTA2), NVP-TAE684 (ITGA6, MMP9, ACTA2) and brefeldin A (ITGA6, RALB). To investigate the effects of ITGB-related inhibitors on CHST11 and their potential as potential therapeutics, several drugs were tested. Accordingly, the IC₅₀ was determined and considered to be the optimal treatment drug concentration (Figure 6C; Figure S7A). According to the results of qPCR analysis, AG-1478 (Tyrphostin AG 1478), brefeldin A, geldanamycin and importazole effectively inhibited CHST11 or mesenchymal-related factors in H358 cells that expressed the highest levels of CHST11 (Figure 6D), but only geldanamycin and importazole inhibited CHST11 in H1975 cells (Figure S7B), which indicates that different CHST11 levels can respond differently to drugs. Based on these results, these potential drugs could be repurposed by targeting the CHST11 signature and the CHST11/integrin axis to inhibit the progression of LUAD.