Numerous transcriptional alterations are shared between EGFR-mutant lung cancer cells during EGFR inhibition and BRAF-mutant melanoma cells during BRAF inhibition.
Transcriptional changes are critical to a cancer cell’s ability to tolerate targeted therapy; however, it is unclear if changes are common between different cancer types and inhibitors. We performed RNA-seq on EGFR-mutant PC9 NSCLC cells treated with the EGFR inhibitor gefitinib, and BRAF-mutant SKMEL28 melanoma cells treated with the BRAF inhibitor dabrafenib to investigate the scope of transcriptional changes in each system, and to uncover shared mechanisms of drug tolerance. Differential expression analysis revealed 2,179 differentially expressed genes (DEGs) (1,524 up-regulated and 655 down-regulated) in PC9 cells treated with gefitinib for 3 days, and 2,219 DEGs (1,674 up-regulated and 545 down-regulated) in PC9 cells treated with gefitinib for 9 days, compared to untreated controls. In SKMEL28 cells cultured in dabrafenib, 4,926 genes (2,431 up-regulated and 2,495 down-regulated) were differentially expressed at day 3, and 4,581 (2,463 up-regulated and 2,118 down-regulated) genes were differentially expressed at day 9, compared to untreated controls. Additionally, we found 24 Hallmark gene sets were significantly enriched in 9-day gefitinib-treated PC9 cells, and 13 Hallmark gene sets were significantly enriched in 9-day dabrafenib-treated SKMEL28 cells, compared to untreated cells when using a strict cut-off of FDR < 0.05. Interestingly, 11 Hallmark gene sets were significantly enriched in both drug-treated cell lines.
We performed unbiased, hierarchical clustering on the 544 genes (422 up-regulated and 122 down-regulated) we identified as significantly altered in both gefitinib- and dabrafenib-treated cells (Fig. 1a, b). Narrowing our focus to genes that have a log2-fold change of at least 2.0, we found alterations in 150 genes that are shared between cell lines (Fig. 1c). Unique
changes occur in each cell line, with 93 genes significantly altered in PC9 cells treated with gefitinib, and 523 genes significantly altered in SKMEL28 cells treated with dabrafenib. A very small number of genes showed significant changes in opposite directions between cell lines. 18 genes showed significant upregulation in PC9 cells and significant downregulation in SKMEL28 cells, whereas 5 genes showed significant upregulation in SKMEL28 cells, and significant
downregulation in PC9 cells (Fig. 1c).
Principal component analysis (PCA) revealed distinct separation of gefitinib-treated and untreated PC9 cells by the first principal component, which explained over 74% of the variance (Fig. 1d). Interestingly, when we performed PCA on SKMEL28 cells in dabrafenib, the first
principal component did not completely separate samples on treatment status alone (Fig. 1e).
These findings suggest that many transcriptional changes are shared between different cancer types adapting to various targeted therapies, although the exact timing of these changes may be unique to each cell line and drug combination.
Proteomic and phospho-proteomic alterations accompany the development of drug tolerance
We next sought to investigate proteomic and phospho-proteomic changes involved in the development of drug tolerance by performing Reverse Phase Protein Array (RPPA) analysis on PC9 cells cultured in gefitinib for 0, 3, 6, and 9 days. Of the 471 antibodies included in the panel, we saw significant changes in 305 protein and/or phospho-protein levels at one or more time points during gefitinib treatment. We performed unbiased, hierarchical clustering to visualize changes in the 55 phospho-proteins that we identified as significantly altered in gefitinib-treated cells (Fig. 2a). We integrated our transcriptomic and proteomic data sets by comparing the log2-fold change of mRNA transcripts to differences in protein levels, each in 9-day gefitinib-treated versus untreated PC9 cells. We observed a positive correlation of the 336 genes/proteins featured in both data sets (Fig. 2b). The labeled points show both a significant
increase in protein level from RPPA analysis and a log-fold change of >2 in our RNA-seq data set. Fibronectin was selected from these points to illustrate the large increases in both protein (Fig. 2c) and transcript abundance (Fig. 2d). These significant changes observed in numerous protein and phosphoprotein levels in gefitinib-treated PC9 cells complement prior transcriptomic studies and reveal potential mechanisms by which cancer cells continue to cycle in targeted therapy.
Targeted inhibition of mutationally activated oncogenes induces sustained mTOR-pathway suppression.
We began investigating shared effects of inhibiting mutationally activated oncogenes in both EGFR-mutant lung cancer cells and BRAF-mutant melanoma cells by analyzing changes in gene expression at 3-, 6-, and 9-days following addition of the EGFR inhibitor gefitinib or the BRAF inhibitor dabrafenib, compared to untreated controls. Using gene set enrichment analysis (GSEA), we found that mTOR-related gene sets were more enriched in untreated controls relative to gefitinib and dabrafenib-treated cells (Fig. 3a, b). Transcript levels of indicated mTOR-related genes such as RPS6, EIF4G1, and EEF2K, showed significant decreases in PC9 cells in gefitinib and SKMEL28 cells in dabrafenib at multiple time points (Fig. 3c, d).
We then examined individual protein and phospho-protein levels of mTOR pathway members. RPPA analysis revealed that total protein levels of mTOR do not significantly change, but phosphorylated mTOR levels decreased during gefitinib treatment of PC9 cells (Fig. 3e). Downstream targets of mTOR including eIF4E, p70-S6K, and S6 showed significant decreases in the amount of phosphorylated protein and total protein (Fig. 3e). When immunoblotting was performed on SKMEL28 cells in dabrafenib to complement protein array results obtained in PC9 cells treated with gefitinib. We observed a significant decrease in phospho-S6 within 24 hours of dabrafenib treatment (Fig. 3f). Taken together, these results indicate treated susceptible cells
with corresponding targeted therapies results in a sudden, sustained decrease in mTOR signaling.
Targeted inhibition of mutationally activated oncogenes induces IGF-pathway activation.
To better understand how cells continue to proliferate despite a crash in mTOR signaling, we began to search for alternate mitogenic growth pathways that were upregulated in drug-treated cells. GSEA revealed that genes associated with the Insulin-like growth factor (IGF) signaling cascade are significantly up-regulated in PC9 cells treated with gefitinib (Fig. 4a) and SKMEL28 cells treated with dabrafenib for 9 days, relative to untreated controls, (Fig. 4b). In gefitinib- (Fig. 4d) or dabrafenib-treated cells (Fig. 4e), we observed significant increases in individual transcript levels of various IGF-related genes such as IGFBP3, IGFBP5, IGFBP7, and most notably IGF1R. Total protein levels of IGFRB did not appear to significantly change during 9-day gefitinib exposure of PC9 cells (Fig. 4c). However, we observed significant increases in the amount of phosphorylated IGF1R (Y1135/6) in PC9 cells treated with gefitinib for 9 days (Fig. 4c). We also observed significant increases in IGFBP3 at days 3 and 6 of gefitinib treatment (Fig. 4c), coinciding with the increase observed in the earlier IGFBP3 transcript levels. These results obtained in multiple cancer contexts expand on the previous work implicating IGF1R in anti-EGFR drug tolerance.
Targeted inhibition of mutationally activated oncogenes induces PLC/PKC signaling.
We observed increased phosphoinositide-specific phospholipase C (PLC) activity in drug treated cells. PLC is often a key player in transmembrane signaling. GSEA performed using genes associated with GPCR signaling revealed strong enhancement in gefitinib-treated PC9 cells relative to untreated controls (Fig. 5a). Further GSEA studies performed on SKMEL28 cells showed a strong enrichment of genes associated with the regulation of phospholipase activity in the drug-treated group (Fig. 5b). Transcript levels of several genes associated with
increased PLC/PKC signaling such as ADCY6, GNAL, GNAS, PDE1C, and PLCG1 showed significant increases in PC9 cells in gefitinib (Fig. 5d) and SKMEL28 cells in dabrafenib (Fig. 5e) at multiple time points. Phosphorylation of PLCG2 at tyrosine 759 is a marker of the active state, and increased levels were observed in PC9 cells at days 6 and 9 of gefitinib treatment (Fig. 5c). Additionally, we observe increase in PKCA and phospho-PKCA/B (T638/6431) in gefitinib-treated PC9 cells. Together, these findings indicate activation of the PKC/PLC signaling pathway upon gefitinib or dabrafenib treatment, which to our knowledge is the first time this has been documented.
Targeted inhibition of mutationally activated oncogenes induces STAT3 activation and YAP activation.
We also saw activation of two pathways previously observed by others. Our GSEA studies revealed that STAT3 signaling was significantly enriched in PC9 treated with gefitinib (Fig. S1a) or SKMEL28 cells treated with dabrafenib (Fig. S1b) for 9 days when compared to untreated controls. Transcript levels of STAT3-associated genes including JAK1 and BCL6 were also significantly increased in both gefitinib-treated PC9 cells (Fig. S1d) and dabrafenib-treated SKMEL28 cells (Fig. S1e). At days 3 and 9 of gefitinib treatment, total protein levels of STAT3 did not differ significantly from untreated PC9 cells (Fig. S1c). STAT3 is considered active when phosphorylated at tyrosine 705, which induces dimerization and nuclear translocation. Interestingly, we observed a significant increase in the amount of phosphorylated STAT3 (Y705) following 9 days of gefitinib treatment (Fig. S1c). From these results, we conclude that increased STAT3 signaling accompanies the emergence of drug tolerance in various contexts.
We also observed activation of YAP signaling. GSEA studies show a strong enrichment of genes associated with YAP transcriptional signature in PC9 cells treated with gefitinib (Fig. S2a) or SKMEL28 treated with dabrafenib (Fig. S2b), when compared to untreated controls. Total YAP protein levels do not significantly change in PC9 cells during 9-day treatment with
dabrafenib (Fig. S2c). YAP phosphorylated at serine 127 is sequestered in the cytoplasm. Interestingly, we observed a significant decrease in level of phospho-YAP (S127) at each time point during gefitinib treatment (Fig. S2c). Indeed, transcript levels of numerous YAP-related genes are significantly higher in drug-treated cells compared to untreated controls of both PC9 (Fig. S2d) and SKMEL28 (Fig. S2e). Proteomic data also revealed increased protein levels of PAI-1 (Fig. S2c), a common YAP-associated protein, in PC9 cells treated with gefitinib for 3, 6, and 9 days. These results suggest YAP signaling is induced upon targeted blockade in oncogene-dependent cells.
Single-cell RNA-seq reveals simultaneous upregulation of markers of multiple alternate mitogenic signaling pathways.
To determine whether the activation of the multiple alternate mitogenic signaling pathways we observed in bulk experiments (i.e. RNA-seq, RPPA) occurred simultaneously in individual cells, or occurred in separate sub-populations, we performed single-cell RNA seq on PC9 cells in gefitinib and SKMEL28 cells in dabrafenib for 0, 24, 48, or 72 hours. Gene-gene plots showed select marker genes for IGF and STAT3 signaling (Fig. 6a), IGF and YAP signaling (Fig. 6b), IGF and PKC/PLC signaling (Fig. 6c) were simultaneously upregulated in PC9 cells in gefitinib for 72 hours compared to untreated controls. Additionally, we observed simultaneous upregulation of the same pathways in SKMEL28 cells in dabrafenib for 72 hours, compared to untreated SKMEL28 cells (Fig. 6d-f). The simultaneous upregulation of multiple marker genes indicates that the enriched pathways are simultaneously up-regulated in single cells, rather than being up-regulated in different sub-populations prior to pooled analysis.
Autophagy accompanies the development of drug-tolerance and can be disrupted with hydroxychloroquine to reduce the number of drug-tolerant cells.
To investigate potential survival mechanisms utilized by cells in targeted therapy, we investigate the role of autophagy during the development of drug tolerance. We observed enrichment of the GO gene set “Selective Autophagy” in PC9 cells treated with gefitinib for 9 days, compared to untreated controls (Fig. S3a). Increased transcript levels of the common autophagy marker ULK1 were observed in gefitinib-treated PC9 cells (Fig. S3b). scRNA-seq analysis of ULK1 and a second marker, GABARAPL1, also showed up-regulation in gefitinib-treated PC9 cells (Fig. S3d). Additionally, a common protein marker used to assess autophagy, LC3A/B was increased in gefitinib-treated PC9 cells, when measured by RPPA (Fig. S3c). Direct visualization of autophagic flux was performed with the CYTO-ID live cell autophagy detection kit. Increased fluorescent signal, corresponding to the increase in autophagosomes critical for autophagy, was observed in gefitinib-treated PC9 cells (Fig. S3e). PC9 cells in regular media showed little to no evidence of autophagy.
To determine if disrupting autophagy leads to fewer drug tolerant cells, we supplemented gefitinib treatment with hydroxychloroquine. This anti-malarial drug is widely available and has been previously used to disrupt autophagy by lowering lysosomal pH, thereby blocking the key step of lysosome-autophagosome fusion11. Hydroxychloroquine does not exert any growth effects on PC9 cells in regular media (Fig. 7a), but significantly reduces the number of surviving PC9 cells when combined with gefitinib (Fig. 7b). From these results, it seems autophagy is induced upon targeted therapy exposure, and using hydroxychloroquine co-treatment is an effective way to reduce the number of drug tolerant cells.
Disruption of ATG5 results in decreased autophagic flux but does not enhance gefitinib killing of PC9 cells.
To validate the utility of blocking autophagy in cells developing drug tolerance, we disrupted the critical ATG5-ATG12 interaction via CRISPR-induced mutations in exon 2 of ATG512. In unmodified PC9 cells, immunoblotting revealed the 55 kDa ATG5-ATG12 complex, while the CRISPR modified clones we generated appear to only have free ATG5 (Fig. 7c). The ATG5-ATG12 complex is required for autophagy, and indeed, these clones showed a decreased ability to undergo autophagy compared to parental PC9 cells (Fig. 7d). Similar to the results observed with chloroquine co-treatment, CRISPR-ATG5 cells did not show any growth differences when compared to control PC9 cells in regular media (Fig. 7e). Interestingly, contrary to the results observed with chloroquine co-treatment, CRISPR-ATG5 cells survived gefitinib treatment at a higher rate than control PC9 cells (Fig. 7f). Although autophagy appears to be induced by the presence of targeted therapy, and chloroquine co-treatment lead to fewer drug-tolerant cells, genetic studies cast doubt on autophagy’s importance in the development of drug-tolerance.
Cells display features of senescence during the development of drug-tolerance.
Cancer cells surviving challenging environments have been shown to display features of cellular senescence, a process which overlaps with autophagy. PC9 cells were subjected to X-gal staining, a common method used to assess senescence-associated beta-galactosidase (SA- β-gal) activity, a feature unique to senescent cells. We observed the characteristic blue-green colorimetric change resulting from SA-β-gal-mediated cleavage of X-gal in PC9 cells treated with gefitinib, and SKMEL28 cells treated with dabrafenib (Fig. S4a). GSEA revealed an enrichment of a senescence-associated gene set in PC9 cells treated with gefitinib compared to untreated controls (Fig. S4b). Increases in normalized transcript abundance levels of several senescence-associated genes such as CDKN1A (Fig. S4c), and CDKN2B (Fig. S4d) are
observed in PC9 cells treated with gefitinib for 3 and 9 days, compared to untreated PC9 cells. scRNA-seq analysis of CDKN2A and CDKN1B, also showed up-regulation in gefitinib-treated cells PC9 cells (Fig. S4e).
We co-treated PC9 cells in gefitinib with multiple “senolytic” drugs to determine if selectively targeting cells displaying features of cellular senescence would reduce the number of drug tolerant cells. The first compound, navitoclax (ABT-263) is a BCL-2 family protein inhibitor that has been shown to selectively eliminate senescent cells13. Navitoclax did not appear to alter the number of PC9 cells when cultured in regular media (Fig. S5a). However, when added to gefitinib media, navitoclax appeared to significantly reduce the number of surviving cells (Fig. S5b). Rather than broadly targeting BCL-2 family proteins, A1331852 is a purported BCL-XL-specific inhibitor that showed no significant impact on PC9 growth in regular media (Fig. S5c). However, similar to navitoclax, co-treatment with A1331852 significantly reduced the number of PC9 cells surviving gefitinib (Fig. S5d). From these results, we can conclude that inhibitors targeting BCL-2 family proteins, specifically BCL-XL, significantly reduce the number of drug tolerant cells.
Disruption of BCL-XL does not result in enhanced killing of PC9 cells by gefitinib.
To validate the utility of disrupting BCL-XL in cells developing drug tolerance, we used CRISPR/cas9 to create a BCL-XL knockout PC9 cell line (Fig. S5e). There were slightly fewer CRISPR-BCL-XL cells compared to unmodified PC9 cells when cultured in regular growth media for 3 days (Fig. S5f). However, contrary to the results observed with navitoclax and A1331852 co-treatments, CRISPR-BCL-XL cells exhibited no differences in survival during gefitinib treatment compared to unmodified PC9 cells (Fig. S5g). Although senescence appears to be induced by the presence of targeted therapy, and co-treatment with senolytic compounds lead to fewer drug-tolerant cells, genetic studies cast doubt on the exact mechanism by which drug tolerance is blocked.