Differential genes in lung cancer
RNA-Seq data from 103 normal lung tissues and 998 lung cancer tissues was analyzed. The differentially expressed genes included 189 downregulated genes and 105 upregulated genes at 2-fold changes level as shown in volcano map and heatmap (Fig. 1A and 1B). The 105 upregulated genes were further analyzed in enriched biological process and it had showed drug transport and collagen catabolic processes were enriched in these upregulated genes (Fig. 1C and Table S1), which indicated that these genes were associated with the drug resistance and tumor metastasis activities. Afterwards, we enriched the cellular components related to the 189 downregulated genes and it was found that these genes that were suppressed in lung cancers are typically associated with apical plasma membrane, collagen trimer and apical part of cell components (Fig. 1D and Table S2), which therefore also indicated a possible correlation of these genes with tumor metastasis.
Resistant genes signature in lung cancer
The enriched drug transport process among the upregulated genes implied that these genes rendered the lung cancer resistance to drugs (Fig. 1C and Table S1). However it was necessary to firstly screen the resistance genes from the patients that were under chemo-drugs therapy. To screen the resistance related genes, we had analyzed the differential genes in line with the clinical information. According to the information, we found that there were 629 patients without treatment while 243 patients with treatment among the 998 lung cancer patients (Fig. 1E). The differential genes between non-treatment and treatment groups were analyzed and our findings revealed that there were 32 differential genes, including 13 upregulated genes with 19 downregulated genes (Fig. 1F). The GO analysis for these differential genes revealed that vitamin D receptor signaling and receptor tyrosine kinases signaling pathway were activated in resistant lung cancers (Fig. 1G). As shown in Table S3, vitamin D receptor signaling includes ADRB2, CYP24A1, ID4 genes, whereas receptor tyrosine kinase signaling includes CAV1, EREG, ID4, FGFBP2, ADH1C genes.
To prove that these genes were really correlated to resistance in lung cancer, we analyzed the hazard ratios of first progression survival time between the patients who highly expressed the genes in comparison with patients who lowly expressed the genes. It was found that the HRs of all upregulated genes were greater than 1, which indicated that the patients who highly expressed the upregulated genes had a shorter survival time and implied these genes were positively associated with resistance (Table S4).
Resistant cells display stemness and EMT ability
To validate that the screened resistance genes are actually associated with resistance, we established cisplatin resistant (A549-CR) and taxol resistant (A549-TR) A549 cells, as shown in Fig. 2A. The RT-qPCR assessment of the resistant genes in A549-CR and A549-TR cells revealed 6 positive resistance genes and 3 negative resistance genes (Fig. 2B). The positive resistance genes included CYP24A1, DNAJC12, EREG, NPTX1, PAEP, and TRPM8. Meanwhile, the negative resistance genes included EMP2, HIGD1B and ADH1C. Meta-analysis for HRs of first progression survival between lung cancer patients who highly expressed these genes and who lowly expressed these genes was conducted and it was found that the pooled HRs (95% CIs) of positive resistance genes were 1.52 (1.33, 1.72) while pooled HRs (95% CIs) of negative resistance genes were 0.59 (0.52, 0.66) (Fig. 2C). Taken together, this result demonstrated that these positive resistance genes were supposed to be the resistance gene signature in NSCLC.
Our previous study prevailed that lung cancer stem cells had resistance ability. Herein, we also wondered that whether these acquired resistant cells also possessed cancer stem cell properties. Thus, sphere forming ability of A549-CR or A549-TR cells was also assessed and it was found that these drug resistant cells showed stronger ability to form spheres as compared to A549 cells (Fig. 2D and S2A). Meanwhile, we also detected the stemness related genes expression [19] in A549-TR spheres and found that Bmi1, KLF4, c-Myc, Nanog and Sox2 genes were indeed highly expressed (Fig. 2E). These findings proved that the acquired resistant cells had cancer stem cells properties. Additionally, since enriched differential genes showed metastasis correlation, therefore epithelial mesenchymal transition (EMT) genes were investigated in resistant cells and the RT-qPCR results revealed elevated expression of N-cadherin, Vimentin, Snail, Slug and ZEB1 genes (Fig. 2F). Given this, we may conclude that the resistant cells had the ability of cancer stem cells and metastasis.
Inhibition of ErbB receptor reversed resistance of NSCLC
GO analysis of 32 candidate resistance genes had revealed that receptor tyrosine kinase signaling were supposed to be activated in resistant cells. To further prove the involvement of RTK signaling pathways in drug resistance, we treated the resistant A549 cells with ErbB receptor inhibitor, afatinib[20], in combination with chemo-drugs to determine whether resistance of cancer cells to chemo-drug could be reversed. When A549-TR and H1299 cells were treated with 1 µM afatinib, it was observed that afatinib treatment suppressed the growth of A549-TR or H1299 cells slightly. However, combination of afatinib along with taxol had significantly inhibited viability of A549-TR or H1299 cells compared with taxol alone treatment (Fig. 3A and 3B). Similarly, concurrent treatment of afatinib also slightly suppressed the growth of A549-CR or H1299 cells alone, whereas a combination of afatinib together with cisplatin had remarkably repressed viability of A549-CR or H1299 cells (Figure S3A and S3B). Therefore, it can be concluded that afatinib was capable of attenuating the resistance ability of cells, thus re-sensitizing these resistant cells to taxol and cisplatin drugs.
Moving on, it was still uncertain whether afatinib treatment could inhibit the sphere forming ability of A549-TR cells or H1299 cells. From the sphere forming assay, findings showed that 1 µM afatinib treatment alone did not suppress the sphere forming ability, but this was achieved when afatinib treatment was combined with taxol (Fig. 3C-F). After that, expression of stemness associated genes in A549-TR cells was observed and it was seen that expression of Nanog and SOX2 genes were suppressed by afatinib (Fig. 3G). This result proved that afatinib could attenuate resistance via suppressing stemness of A549-TR cells. Since resistant cells are also correlated with metastasis ability, EMT related genes expression in A549-TR cells treated with afatinib was investigated. However, results obtained showed that afatinib did not inhibit the mRNA level of EMT related genes (Fig. 3H). Overall, it can be concluded that afatinib attenuated resistance via inhibiting the stemness of cancer cells.
EREG Promotes chemoresistance of NSCLC
One of the validated resistance genes, EREG, is the ligand for ErbB receptor. However, EREG function in the chemoresistance regulation remains unclear. The RNA-Seq analysis revealed EREG level was higher in treated patients than untreated patients (Fig. 4A). Consistently, the overall survival time and disease free survival time of EREGhigh population were shorter than EREGlow population (Fig. 4B, 4C), which indicated highly expression of EREG promoted the progression of lung cancer and implied that EREG was correlated with resistance. To figure out whether EREG functions in chemoresistance, the EREG protein level was detected in A549 cells treated with cisplatin or taxol and the result showed that chemo-drugs treatment significantly elevated EREG level (Fig. 4D). This finding proposed that increased EREG level might cause chemoresistance of NSCLC. Subsequently, EREG cytokine was applied to treat the A549 and H1299 cells and the result revealed that the cell viability of EREG plus taxol treated group was higher than the taxol treated group (Fig. 4E). From sphere forming assay, it also revealed that the spheres of EREG plus taxol treated group was significantly higher than the taxol treated group (Fig. 4F). These findings indicated that EREG promoted chemoresistance of NSCLC. Interestingly, EREG treatment significantly increased the mRNA level of stemness associated genes, Bmi1, KLF4, c-Myc, Nanog and Sox2 (Fig. 4G). Taken together, it concluded that EREG was able to promote chemoresistance of NSCLC via increasing the level of stemness associated genes.
Inhibition of ErbB receptor suppressed ERK signaling
It was reported that ERK and AKT signaling was the downstream targets of receptor tyrosine kinase signaling [21–23]. Therefore, p-ERK1/2 and p-AKT activation in A549 and H1299 cells treated with cisplatin or taxol was analyzed. The results revealed that p-ERK1/2 was elevated significantly although p-AKT expression remained unchanged (Fig. 5A). This observation indicated that ERK signaling was primarily involved in drug resistance. Based on our findings so far, it was speculated that afatinib attenuated resistance via inhibition of ERK signaling. To verify this speculation, p-ERK1/2 level in A549-TR and H1299 cells treated with afatinib was tested and indeed, it was found that p-ERK1/2 expression had decreased significantly (Fig. 5B).
To further confirm that afatinib attenuated resistance through ERK signaling, p-EKR1/2 inhibitor (selumetinib) was used to treat A549-TR cells and results obtained showed that selumetinib had suppressed p-ERK1/2 (Figure S4A). Along with this observation, selumetinib was also seen to enhance the sensitivity of A549-TR/CR cells to taxol and cisplatin (Fig. 5C and S4B). We also tested the selumetinib effect on H1299 cells and findings obtained showed that selumetinib had assisted taxol and cisplatin to kill H1299 cells more effectively (Fig. 5D and S4C). Since afatinib treatment inhibited the stemness associated genes expression and its combination with taxol had suppressed sphere forming ability, we therefore subsequently studied whether selumetinib could also suppress sphere forming ability when it was combined with taxol or cisplatin. The results proclaimed that selumetinib did not inhibit sphere forming ability when treated as a standalone drug, but combination of selumetinib and taxol or cisplatin was capable to effectively inhibit sphere forming activities in A549-TR and H1299 cells (Fig. 5E-H and S4D, S4E). In summary, EREG/ErbB played a role in drug resistance mediated by ERK signaling.
Downregulation of EREG re-sensitized NSCLC to chemo-drugs through ERK signaling
EREG was knocked down in A549 and H1299 cells to determine whether inhibition of EREG could really reversed resistance. As shown in Fig. 6A, both shEREG-1 and shEREG-2 had effectively decreased the EREG protein level. Along with EREG knocked down, p-ERK1/2 and survivin expression was also seen to be decreased (Fig. 6A), which indicated that EREG activated ERK signaling. Furthermore, it was found that EREG knocked-down had rendered A549-TR/CR cells to be more sensitive to taxol and cisplatin drugs treatment (Fig. 6B and S5A). From the sphere forming assay, it was also found that shEREG had significantly increased sensitivity of the spheres derived from A549-TR and H1299 cells to taxol or cisplatin treatment (Fig. 6C, 6D and S5C). Similarly, A549-shEREG and H1299-shEREG cells were more sensitive to taxol or cisplatin treatment compared to parental cells (Fig. 6E, 6F and S5B). To further confirm that knocking down EREG affected the stemness, the mRNA level of stemness associated genes was detected and it proclaimed that shEREG significantly decreased these stemness associated genes level (Fig. 6G).Taken together, we concluded that knocking down EREG also reversed drug resistance by inhibition of stemness associated genes through ERK signaling.
Resistance gene signature was associated with disease free survival in other cancers
To study whether this positive resistance gene signature, including CYP24A1, DNAJC12, EREG, NPTX1, PAEP and TRPM8, could also a predictor signature for drug resistance in other cancer types, we analyzed the HRs of disease free survival time for these resistance genes in colon cancer, renal carcinoma, gastric cancer and pancreatic cancer. After that, a meta-analysis was conducted for pooled HRs of these genes. The pooled HRs of these genes in colon cancer was 1.34, which implied that these genes were a signature for drug resistance occurrence in colon cancer (Fig. 6H). Likewise, the pooled HRs of these genes in renal carcinoma and gastric cancer were 1.54 and 1.48 respectively, hence implying that these genes may also be signature genes for drug resistance occurrence in renal carcinoma and gastric cancer (Fig. 6H). On the other hand, the pooled HRs of pancreatic cancer was 1.25, where lower 95% CI is lesser than 1, therefore indicating that it was not significant for this resistance signature to act as a predictor in pancreatic cancer (Fig. 6H). From the meta-analysis, we concluded that this positive resistance gene signature was a predictor for drug resistance in lung cancer, colon cancer, renal carcinoma, gastric cancer and pancreatic cancer.