A key tool of cancer therapy has been targeted inhibition of oncogene-addicted pathways. However, efficacy has been limited by progressive emergence of resistance as transformed cells adapt. Here, we use Drosophila to dissect response to targeted therapies. Treatment with a range of kinase inhibitors led to hyperactivation of overall cellular networks, resulting in emergent resistance and expression of stem cell markers, including Sox2. Genetic and drug screens revealed that inhibitors of histone deacetylases, proteasome, and Hsp90 family of proteins restrained this network hyperactivation. These “network brake” cocktails, used as adjuncts, prevented emergent resistance and promoted cell death at subtherapeutic doses. Our results highlight a general response of cells, transformed and normal, to targeted therapies that leads to resistance and toxicity. Pairing targeted therapeutics with subtherapeutic doses of broad-acting “network brake” drugs may provide a means of extending therapeutic utility while reducing whole body toxicity.

Significance: These findings with a strong therapeutic potential provide an innovative approach of identifying effective combination treatments for cancer. Cancer Res; 78(15); 4344–59. ©2018 AACR.

Despite important progress, cancer remains a serious health problem worldwide. Recent efforts at developing drugs that reduce tumor progression have yielded important successes; however, most tumors eventually develop resistance to drugs including targeted therapies. After an initial response, many tumors develop poorly understood strategies to evade therapy. In some patients, this reflects previously undetected mutations that confer resistance, suggesting a selection process initiated by drug response. However, many tumors achieve resistance through undefined, nongenetic means (reviewed in ref. 1). Furthermore, tumors commonly develop resistance to the initial therapeutic and, simultaneously, to other targeted therapies. This suggests broad changes within tumor cells can occur in response to drug therapy including in the kinome and secretome (2, 3), although the nature of these changes is poorly defined.

Here we use a Drosophila cancer/normal cell model and human cancer cell lines to explore mechanisms by which transformed and normal cells respond to targeted therapies. We demonstrate that a key component of this response is broad alteration of tissues' overall cellular network: the result is activation of a large cross-section of signaling pathways and expression of stem cell markers in a subset of cells. This broad alteration in tumor and normal cells' overall network in turn led to progressive resistance to a broad range of standard-of-care targeted therapies.

Our genetic and drug screens identified specific drug cocktails that restricted this broad response to therapeutics. For example, pairing (i) targeted therapies such as sorafenib, erlotinib, and trametinib with (ii) low, subtherapeutic doses of bortezomib plus vorinostat led to control of cellular networks, reduced “stemness,” reduced whole-animal toxicity, and sustained drug efficacy. We observed similar therapeutic benefits with broadly acting drugs, used as adjunct to targeted therapy, that restrain networks through other means including the Hsp90 inhibitor AUY922 and the HDAC-PI3K dual pathway inhibitor CUDC-907. Our findings uncover a broader principle: targeted therapies induce hyperactivation of the cellular kinase network, an alteration that can be restrained by subtherapeutic doses of broadly acting inhibitors. The result is a significantly improved and sustained therapeutic response.

Antibodies

Antibodies used for Drosophila and human cancer line Western blot analysis were: anti-pRet, anti-pJnk, anti-pAkt, anti-SOX2, anti-KLF4, anti-LIN28, anti-Oct4, anti-Nanog, anti-cMyc, anti-pMOB, anti-cleaved PARP (Cell Signaling Technology), anti-pSrc(Y418) (Invitrogen), anti-pERK, anti-total-ERK (Sigma), plus anti-Actin, anti-E-cadherin, anti-α-catenin, anti-Rho1, anti-Syntaxin, anti-CycD, anti-Argos, anti-β-tubulin (Developmental Studies Hybridoma Bank), anti-actin, anti-GAPDH, and anti-RhoA antibodies were purchased from Santa Cruz Biotechnology. Anti-Rac1 antibody was from BD Biosciences, anti-EGFR from Julia Cordero, anti-activated-β-Catenin from Millipore, and histone modification antibodies from ActiveMotif.

Cancer cell lines

The various human cancer cell lines were obtained from the scientists mentioned in the Acknowledgements section, as well as some from our own previously acquired collection. Authentication was performed by testing for their growth rates, morphology, previously published sensitivity to targeted therapies, for example, Ret-inhibitor sensitivity for MZ and TT cells, EGFR-inhibitor sensitivity for H358, etc. No adverse contamination issues were noticed and therefore Mycoplasma testing was not performed. Typically, a maximum of eight to nine passages were required from time of thawing to completion of individual experiments, except for chronic drug treatment studies where cells were passaged over few months as indicated in experiments.

Chronic drug treatment assays on human cancer lines

Human non–small cell lung cancer (NSCLC) cancer line H358 or melanoma line 239 were grown in 75-cm2 culture flasks. H358 cells were grown in the following conditions: (i) 0.01% DMSO, (ii) 0.5 μmol/L erlotinib, (iii) 1 μmol/L erlotinib, (iv) bortezomib (6 nmol/L) + vorinostat (50 nmol/L), (v) erlotinib (1 μmol/L) bortezomib (6 nmol/L) + vorinostat (50 nmol/L). 239-melanoma cells were grown in the following conditions: (i) 0.01% DMSO, (ii) 0.5 mol/L vemurafenib, (iii) bortezomib (6 nmol/L) + vorinostat (50 nmol/L), (iv) vemurafenib (0.5 μmol/L), bortezomib (6 nmol/L) + vorinostat (50 nmol/L). Initially, cells were seeded in culture flasks at 20% confluency and incubated in RPMI1640 media with the above-mentioned conditions. When cells reached confluency (DMSO controls), they were split and transferred to new culture flask with identical fresh media conditions once a week. Otherwise identical fresh media was provided to each flask once a week. When single drug–treated cells (erlotinib or vemurafenib) started growing similar to DMSO control cells, the drug-resistant cells as well as the triple drug–treated cells were further amplified for three generations in the absence of any drug treatment. This allowed the resistant cells and triple drug cocktail–treated cells to yield sufficient cells for the various assays, that is, phospho-kinome array, Western blot, and MTT viability assay.

Comprehensive statistical analysis

For viability adult/pupal analysis, mean and SEM were calculated and 4–5 vials/experiments, biological replicates, per dose were analyzed and repeated at least two times. Each vial had between 20 and 80 developing embryos and total (N) indicated in legend represents the total number of embryos analyzed. To assess statistical significance of difference between means, t test with Welch correction was performed using PRISM software. The correction was used to account for samples with unequal variances, as each drug or mutant affects overall cellular networks differently, and unequal sample sizes. For MTT on cancer cells, each dose was performed in quadruplicates and mean signal and SEM analyzed.

Fly stocks, genetics, and subcloning

Fly stocks were obtained from Bloomington and VDRC Drosophila stock centers (C. Pfleger, Department of Oncological Sciences, Icahn School of Medicine at Mount Sinai, New York, NY; M. Mlodzik, Department of Cell, Developmental, and Regenerative Biology, Icahn School of Medicine at Mount Sinai, New York, NY). UAS-RetMEN2B flies were generated and published previously (4). Drosophila Erk allele (rolled/CyO-Hu+; Bloomington stock number #386) is fully viable stock over a CyO-RFP marked balancer.

Inhibitor studies in flies

Drugs were obtained from LC laboratories or Selleck Chemicals and dissolved in DMSO as stock solutions ranging from 1 to 200 μmol/L. Drugs were diluted in molten (∼50–60°C) enriched fly food, vortexed, mixed by pipetting, aliquoted into 5-mL vials, and solidified at room temperature to yield the indicated final drug concentrations. Thirty to 60 embryos of each genotype were raised on drug-containing food (500–1,000 μL) in 5-mL vials until they matured as third-instar larvae (wing disc western assay) or allowed to proceed to adulthood (viability assay and wing vein quantitation assay). Five vials per experiment were analyzed and repeated at least three times. At the end of 14–16 days, pupae and eclosed adults were counted. The mean for each condition is represented as columns with the SEM depicted by the error bars.

Kinase tree render analysis

Illustration model was reproduced courtesy of Cell Signaling Technology, Inc. (www.cellsignal.com). “Kinome-render” program from the Najmanovich research group (http://bcb.med.usherbrooke.ca/) was used to generate sorafenib kinase inhibition tree using in vitro targets identified in Karaman and colleagues' (2008) study as input. Tree depicts in vitro targets only, not relative inhibition of targets, using the available Kd data. Fifty-six of the 79 known targets of sorafenib are represented on the “Kinome-Render” generated tree.

MTT assays using cancer lines

All cancer cell lines were cultured in RPMI1640 media, supplemented with 10% BSA and penicillin/streptomycin antibiotics mix. Cells were grown in 75-cm2 sterile polystyrene culture flasks to 80% confluency, trypsinized, and reseeded in equal aliquots into 96-well plates. After 2 days and approximately 50% confluency, media were removed and replaced with DMSO or drug-containing media. Cells were allowed to grow another 6 days (MZ-CRC-1 and TT) or 4 days (all other fast growing cancer lines), after which MTT assay was performed as described previously (4). To compute combination index (CI), the IC50's each individual drug on each cell line was assessed using PRISM software as shown in main figures. Then, CI was analyzed using the following formula (5):

formula

Summary of all the computed CIs are provided in Supplementary Table S1.

Phosphoprotein array analysis

For assessment of kinase activity of human cancer cell lines, we used the PathScan RTK Signaling Antibody Array Kit (catalog no. 7982). Briefly, 100-cm2 tissue culture plates were plated with human cancer cells at 50%–60% confluency in RPMI1640 media with our without drugs and allowed to incubate for 4–5 days. Cells were washed, lysed, quantified, and exposed to phospho arrays as recommended by the manufacturer, and developed according to manufacturer's protocols. Doublet of each signal was quantitated using densitometric analysis on Image J program, and normalized to time matched untreated cells to create PRISM software–generated heatmap.

Western blot quantitation analysis of fly wing discs

For fly Western blot experiments, we gathered tissues from three independent biological replicates (∼15–20 different animals, total = ∼30–40 third-instar discs) pooled into one tube for each treatment condition. These were then analyzed as Western blots using the indicated antibodies. This was done to have enough sample to cover the large number of Western blots for each condition. Third-instar discs of each genotype were dissolved in lysis buffer (50 mmol/L Tris, 150 mmol/L NaCl, 1% Triton-X100, 1 mmol/L EDTA) supplemented with protease inhibitor cocktail (Sigma) and phosphatase inhibitor cocktail (Sigma). For human cell lines, lysis was performed with RIPA buffer. Western blot analysis was developed as described previously (6). Total protein amounts in each lysate was established by performing Bradford assay (Bio-Rad), and equivalent amounts (1–5 μg) of total protein was loaded per lane. Membranes were stripped with SIGMA Restore stripping buffer and reprobed with other antibodies to assess signal under exactly the same loading conditions. Exposed films were scanned and Western signal for each marker (TIFF files) was quantitated using densitometric analysis on Image J program, and normalized to time matched untreated cells/control cells to create PRISM software–generated heatmap.

Western blotting of cancer cell lines

Human cancer cell lines were grown in 100-cm2–well plates in RPMI1640 media each supplemented with 10% heat-inactivated FBS and penicillin/streptomycin antibiotics. Cells were treated for 4–5 days with inhibitors or vehicle (0.1% DMSO). Western blot analysis was developed as described previously (6). Total protein amounts in each lysate were established by performing Bradford assay (Bio-Rad), and equivalent amounts (5–15 μg) of total protein was loaded per lane. Multiple loading controls were analyzed per experimental set, including GAPDH, actin, β-tubulin, histone-H3, and only some were shown in figures.

Whole-mount imaging of fly and wings

For adult wing vein analysis, wings were dissected and kept in 100% ethanol overnight, mounted on slides in 80% glycerol in PBS solution, and imaged by regular light microscopy using Leica DM5500 Q microscope.

Xenograft analysis

A total of 5–10 × 106 TT cells were injected subcutaneously into one flank of male nu/nu mice. Five mice each showing established growing tumors were separated into vehicle or drug treatment groups. A similar range of tumor sizes was selected for each experiment and treatment started when each tumor reached a size of 100 mm3. Vehicle, sorafenib (40 or 60 mg/kg)+bortezomib (0.05 or 0.3 mg/kg), or sorafenib (40 mg/kg)+bortezomib (0.05 mg/kg)+vorinostat (10 mg/kg) were administered by oral gavage (orally) once daily, five times a week. Tumor and body weight measurements were performed three times a week. The difference between initial and final tumor size at each measuring point was used to calculate percentage change in tumor size. Statistical analyses and waterfall plots of mean tumor size changes were performed using unpaired Student t test with Welch correction using PRISM software. Mouse experiments were carried out by Antitumor Assessment Facility at Memorial Sloan Kettering Cancer Center following Public Health Services guidelines, set forth by the Office for Laboratory and Animal Welfare (OLAW) division of the NIH. The work was covered under an approved Institutional Animal Care and Use Committee protocol at the Memorial Sloan Kettering Cancer Center facility (#04-03-009, principal investigator: de Stanchina, Head, Antitumor Assessment Core, Memorial Sloan Kettering Cancer Center, New York, NY).

Multiple endocrine neoplasia type 2B (MEN2B) is an often aggressive disease characterized by a series of morbidities including medullary thyroid carcinoma (MTC), pheochromocytoma, and mucosal neuromas. Most cases are associated with activated, oncogenic RetM918T; previous studies, including our own, have shown that multiple downstream signaling pathways are activated to promote transformation (7–9). RetM918T is modeled by Drosophila RetM955T; we refer to this oncogenic isoform as Ret2B. Genetic modifier studies (7) as well as Western blot analysis confirmed that our Drosophila model (Supplementary Fig. S1A) recapitulated important signaling cascades also observed in vertebrate systems (7, 8). Using this model, we developed a Drosophila whole-animal viability assay to identify potent multi-targeted inhibitors of the Ret signaling cascade (Supplementary Fig. S1B; ref. 4). Using the patched-GAL4 driver to express UAS-Ret2B in multiple developing tissues (ptc>Ret2B) led to approximately 50% of embryos reaching pupal stages but none developing to adults (Fig. 1A). This assay provided a quantitative measure of Ret2B activity including transformation activity (4).

Figure 1.

An approach to identifying drug cocktails for cancer treatment. A,ptc>Ret2B flies were screened against a panel of drugs (Supplementary Fig. S1B). Some drugs improved the number of statistically significant animals that survived to pupal stages (blue columns, asterisk). Only sorafenib yielded statistically significant survival to adult stages (red column, asterisk). ptc>Ret2B flies were rescreened against the same panel of clinically relevant drugs in the presence of sorafenib to identify useful combinations. A subset of combinations, including sorafenib/bortezomib, sorafenib/dasatinib, and sorafenib/wortmann, in improved adult viability compared with sorafenib as single agent (bracket, double asterisk). Significance (P < 0.05; asterisks) of pupa and adult viability was determined by two-tailed Student t test performed with Welch correction using PRISM software. A list of all the Drosophila viability assay tests, mean values, P values, and total number of flies screened (N) is provided in Supplementary Table S1; details on comprehensive statistical analysis section is given in Materials and Methods. B, Reducing a genomic copy of erk, mek, or hdac1, or knockdown of SP1-transcription factor, or coexpressing InRDN further improved viability of flies fed sorafenib/bortezomib combination. The 3-drug cocktail, sorafenib/bortezomib/vorinostat, also strongly improved adult viability of ptc>Ret2B flies compared with flies fed sorafenib/bortezomib (bracket, asterisk). Significance (P values) of adult viability was determined by two-tailed Student t test performed with Welch correction using PRISM software. C, Western blot analysis of whole wing discs of indicated genotype/treatments represented as a heatmap (left). Western blot data for antibodies tested are in Supplementary Fig. S2B. Heatmap is represented as the ratio of signal, analyzed in Image J, of treated tissue:control wild-type tissue. Comparison of 765>Ret2B versus 765>Ret2B; erk−/+ tissues (brackets, asterisk) treated with sorafenib showed that drug efficacy (Fig. 1B; Supplementary Figs. S2A and S2B and S3A) correlated with overall lower levels of the network of proteins tested. At lower sorafenib doses in 765-GAL4 control flies, pERK levels elevated, indicating increase in Ras pathway activity; in addition, overall activity increase was observed in most proteins tested (bracket, asterisk). Right, Western blot analysis of control (765-GAL4) and 765>Ret2B larvae exposed to sorafenib/bortezomib or sorafenib/bortezomib/vorinostat drug combinations. As in Fig. 1C, Western blot data are represented as a heatmap normalized to wild-type tissue treated with DMSO alone. Both tissues exposed to the 3-drug cocktail showed reduced activation of the markers sampled compared with sorafenib/bortezomib combination, including EGFR, pAkt, pErk, Rac1 specifically in control tissues (brackets, asterisks). This indicated the 3-drug cocktail was most potent in restraining pathway hyperactivation in normal as well as Ret-expressing cells.

Figure 1.

An approach to identifying drug cocktails for cancer treatment. A,ptc>Ret2B flies were screened against a panel of drugs (Supplementary Fig. S1B). Some drugs improved the number of statistically significant animals that survived to pupal stages (blue columns, asterisk). Only sorafenib yielded statistically significant survival to adult stages (red column, asterisk). ptc>Ret2B flies were rescreened against the same panel of clinically relevant drugs in the presence of sorafenib to identify useful combinations. A subset of combinations, including sorafenib/bortezomib, sorafenib/dasatinib, and sorafenib/wortmann, in improved adult viability compared with sorafenib as single agent (bracket, double asterisk). Significance (P < 0.05; asterisks) of pupa and adult viability was determined by two-tailed Student t test performed with Welch correction using PRISM software. A list of all the Drosophila viability assay tests, mean values, P values, and total number of flies screened (N) is provided in Supplementary Table S1; details on comprehensive statistical analysis section is given in Materials and Methods. B, Reducing a genomic copy of erk, mek, or hdac1, or knockdown of SP1-transcription factor, or coexpressing InRDN further improved viability of flies fed sorafenib/bortezomib combination. The 3-drug cocktail, sorafenib/bortezomib/vorinostat, also strongly improved adult viability of ptc>Ret2B flies compared with flies fed sorafenib/bortezomib (bracket, asterisk). Significance (P values) of adult viability was determined by two-tailed Student t test performed with Welch correction using PRISM software. C, Western blot analysis of whole wing discs of indicated genotype/treatments represented as a heatmap (left). Western blot data for antibodies tested are in Supplementary Fig. S2B. Heatmap is represented as the ratio of signal, analyzed in Image J, of treated tissue:control wild-type tissue. Comparison of 765>Ret2B versus 765>Ret2B; erk−/+ tissues (brackets, asterisk) treated with sorafenib showed that drug efficacy (Fig. 1B; Supplementary Figs. S2A and S2B and S3A) correlated with overall lower levels of the network of proteins tested. At lower sorafenib doses in 765-GAL4 control flies, pERK levels elevated, indicating increase in Ras pathway activity; in addition, overall activity increase was observed in most proteins tested (bracket, asterisk). Right, Western blot analysis of control (765-GAL4) and 765>Ret2B larvae exposed to sorafenib/bortezomib or sorafenib/bortezomib/vorinostat drug combinations. As in Fig. 1C, Western blot data are represented as a heatmap normalized to wild-type tissue treated with DMSO alone. Both tissues exposed to the 3-drug cocktail showed reduced activation of the markers sampled compared with sorafenib/bortezomib combination, including EGFR, pAkt, pErk, Rac1 specifically in control tissues (brackets, asterisks). This indicated the 3-drug cocktail was most potent in restraining pathway hyperactivation in normal as well as Ret-expressing cells.

Close modal

Sorafenib altered cellular networks in a Drosophila Ret2B model

Mixing drugs into the flies' media, we screened a panel of clinically relevant anticancer drugs for improved ptc>Ret2B viability (Supplementary Fig. S1B and S1C). The panel included clinically approved Ret inhibitors, for example, vandetanib and cabozantinib, as well as drugs that targeted various pathways, like PI3K, MAPK, SRC/Abl, (Supplementary Fig. S1A) that we and others have shown to be important for Ret signaling (4, 10–12). Also included in the panel were inhibitors targeting the proteasome, histone deacetylases, Hsp90, whose effects are more systemic and have shown promise as therapies in different cancer paradigms, including thyroid cancer (13–15). A couple of these broad-acting inhibitors are clinically approved: bortezomib (multiple myeloma) and vorinostat (T-cell lymphoma). Sorafenib exhibited the strongest rescue: a small fraction of embryos was rescued to adult stages (Fig. 1A and B). Sorafenib is a kinase inhibitor with multiple targets including Ret and its downstream effector Raf that is effective against Ret-dependent human thyroid cancer cells and Ret2B-dependent oncogenic signaling in Drosophila tissues (4, 10, 16).

Previous work demonstrated that, at low doses, Raf inhibitors can activate Ras pathway signaling by promoting formation of active complexes (17). Indeed, low-dose sorafenib activated the Ras/MAPK pathway in vivo as assessed by increased wing venation, a phenotype linked to elevated Ras pathway activity (Supplementary Figs. S2A–S2C and S3A–S3C; ref. 18); few animals survived to pupariation (Fig. 1A). Interestingly, reducing gene dosage of the Ras/Raf downstream pathway effector erk (ptc>Ret2B,erk+/−) by 50% (Supplementary Fig. S2C) significantly improved sorafenib efficacy even at low doses, resulting in improved wing venation and reduced toxicity as assessed by pupariation rates (Fig. 1B; Supplementary Figs. S2A and S2B and S3A). The ability of sorafenib to demonstrate efficacy at low doses in the proper genetic background suggests that sorafenib toxicity is at least, in part, due to mechanisms beyond directly promoting active complexes.

Expressing oncogenic Ret throughout the larval wing disc epithelium (765>Ret2B), followed by Western blot analysis indicated that feeding lower doses of sorafenib (50 μmol/L, 100 μmol/L) led to activation of multiple signal transduction proteins including activated, phosphorylated forms of Ret, Erk, Akt, and Src, while this hyperactivation was somewhat restrained at higher doses (Fig. 1C; Supplementary Fig. S3B–S3E). This indicated that lower doses of Raf inhibitors such as sorafenib evoked activation of cellular pathways beyond MAPK signaling. This contributed to sorafenib's toxicity as well as increased wing venation as described previously. In the presence of higher doses of sorafenib, similar to its effects on viability, reducing erk gene dosage (765>Ret2B,erk+/−) improved wing phenotypes and significantly reduced overall phosphorylation levels of most assayed proteins (Fig. 1C, right; Supplementary Fig. S3A–S3E).

The Drosophila system also allowed us to readily isolate large numbers of drug-treated tissues from normal nontumor-containing flies to analyze their cellular protein networks. Higher doses of sorafenib that showed improved viability (ptc>Ret2B) as well as suppression of wing venation (765>Ret2B) also had lower median network activity levels in normal cells (Fig. 1C, left). This suggested that sorafenib has broad effects on cellular protein networks, leading to cellular toxicity and poor efficacy. Restraining these networks genetically can lead to significant improvement of sorafenib's therapeutic profile. We therefore searched for clinically relevant combination therapeutics that led to a similar network restraint.

Drug combinations restrained hyperactivation of cellular protein networks in both normal and transformed cells

We rescreened the same clinically relevant drug library in the presence of sorafenib to identify combinations that could further improve ptc>Ret2B viability. Three drugs combined with sorafenib to increase the viability of these flies (Fig. 1A; Supplementary Fig. S1C). We focused on sorafenib/bortezomib and sorafenib/dasatinib as the most clinically relevant combinations (Fig. 1B; (19, 20). Western blot analyses indicated that each combination kept the median network activity level in transformed cells below the level observed with sorafenib alone (Supplementary Fig. S4A and S4B, brackets, asterisk). This once again suggested that treatment efficacy is linked to the overall activity of the cellular protein networks, and that better treatments would lead to similar restraining of the protein network in the normal cells as well.

Indeed, we found that one combination, sorafenib/bortezomib, which also kept the median protein network in normal cells closer to baseline state compared with sorafenib/dasatinib (Fig. 1C; Supplementary Fig. S4A–S4D). Bortezomib is a proteasome inhibitor approved for multiple myeloma and mantle cell lymphoma. In contrast, another combination, vandetanib/rapamycin, showed the opposite effect: increased toxicity in both ptc>Ret2B flies and control flies (Supplementary Fig. S5A–S5D) as well as higher median level of the protein network in normal cells (Supplementary Fig. S5B). Thus, this further supported the idea that restraining cellular protein networks leads to improved efficacy of anticancer treatments in our fly models.

To further improve sorafenib/bortezomib–mediated rescue we used a dominant genetic modifier screen, focusing on a subset of genes that targeted different kinase pathways as well as other cellular networks, for example, epigenetic and transcriptional regulators. Clinically relevant inhibitors were available for these pathways. Reducing dosage of genes encoding ERK (rolled+/−), MEK (dsor1+/−), or HDAC1 (rpd3+/−) orthologs, SP1 (SP1-RNAi), or expressing a dominant-negative insulin receptor (InRDN) by transgene, improved the viability of ptc>Ret2B flies fed sorafenib/bortezomib (Fig. 1B). The combination sorafenib/dasatinib also benefited from reducing erk gene dosage, again lowering the median protein network level (Fig. 1B; Supplementary Fig. S4A and S4B). This suggested the efficacy of sorafenib/bortezomib combination could be further improved by targeting other cellular networks, for example, MAPK, PI3K, HDACs, and SP1-transcriptional axes.

The requirement for the histone deacetylase Rpd3 was interesting given the broad control of cellular networks by epigenetic complexes. Vorinostat is approved for the treatment of T-cell lymphoma and for multiple myeloma and is one of several pan-HDAC inhibitors in clinical trials for other cancer types (e.g., refs. 21, 22). Addition of low-dose vorinostat enhanced the viability of both ptc>Ret2B flies and, notably, control flies treated with sorafenib/bortezomib (Fig. 1B; Supplementary Fig. S6A). On the basis of Western blot analysis, the three-drug combination sorafenib/bortezomib/vorinostat further reduced cells' overall network activation in both controls and Ret2B- expressing tissues (Fig. 1C, right). Other pan-HDAC inhibitors similarly enhanced viability of control flies when combined with sorafenib/bortezomib (Supplementary Fig. S6A). Vorinostat regulates transcription of a large number of target genes (23), suggesting it restrained cellular response to drugs by controlling changes in transcription. Consistent with this view mithramycin, a chemical inhibitor of SP1-class transcription factors (24), similarly improved efficacy of the sorafenib/bortezomib combination (Supplementary Fig. S6B); previous studies had shown that bortezomib treatment reduced the levels of SP1, a potent transcriptional activator of various HDACs (25). Moreover, knockdown of SP1 had also improved sorafenib/bortezomib combination efficacy in our fly viability assays (Fig. 1B). We concluded that inhibition of HDAC function or other components regulating transcription like SP1, genetically or through drugs, suppressed oncogenic Ret2B signaling in our fly models.

To establish potential translational relevance of our findings, we surveyed previous studies of molecular mechanisms of sorafenib resistance in human tumors. We found that indeed the InsR/IGF (26, 27), EGFR (28, 29), Akt (30, 31), Src kinases (32), and multiple RTK's (29) were all pathways contributing to sorafenib resistance in human cancer cells. In our Drosophila studies restraining these pathways in normal and transformed cells, either genetically or pharmacologically, correlated with increased sorafenib efficacy (Fig. 1B and C; Supplementary Fig. S3B–S4D).

Broadening our work, we examined other Ret pathway kinase inhibitors including trametinib (MEK inhibitor; approved for use in patients with melanoma) and the experimental polypharmacologic drugs AD57 and AD80 (4, 33). Viability of both control (ptc-GAL4) and ptc>Ret2B flies improved in the presence of each targeted therapy when combined with bortezomib and vorinostat (Supplementary Fig. S6A and S6C). We also tested whether other drugs inhibiting broad protein networks cooperated at subtherapeutic doses; substituting bortezomib with Hsp90 inhibitor AUY922 and vorinostat with the pan HDAC inhibitor CUDC-907. In Drosophila, AUY922/vorinostat paired with sorafenib, trametinib, AD57, and AD80 to improve viability of both control and ptc>Ret2B flies (Supplementary Fig. S6D) in a manner similar to bortezomib/vorinostat.

Together, these data support a model in which kinase inhibitors such as sorafenib, trametinib, as well as polypharmacological inhibitors like AD80, etc., promote strong hyperactivation of the overall cellular network in both transformed and normal tissues, limiting efficacy and promoting toxicity. Drug efficacy is strongly improved by restraining broad network hyperactivation through the selective use of genetics or through drugs that broadly control the cellular network including bortezomib and vorinostat. These drugs act as adjunct treatments to provide a “network brake” to the liabilities of sorafenib at doses well below those required for significant efficacy.

Our Western blot data showing that targeted therapeutics induced hyperactivation of kinase networks matches findings from previous studies. For example, multiplexed kinase inhibitor beads and mass spectrometry (MIB/MS) analysis showed that treatment of tumor cells led to upregulation of the majority of RTKs, tyrosine kinases, and MAPKs analyzed within 24 hours (34). We hypothesized that our findings may provide a potential therapeutic solution to this problem. We therefore next assessed whether vertebrate cancer cells would respond similarly to “network brake” cocktails.

“Network brake” cocktails restrained human thyroid cancer networks to oppose tumor progression

The MTC cancer cell line MZ-CRC-1 (35) harbors an activating mutation in Ret analogous to the Drosophila Ret2B mutation. Sorafenib has been previously demonstrated to inhibit MTC cell line growth (36); this efficacy required high-dose treatment (Fig. 2A) or lower doses of polypharmacologic compound AD80 (Fig. 2B). MZ-CRC-1 cells treated with sorafenib for 4 hours showed inhibition of several intracellular pathway effectors including ERK as assessed by phosphoprotein array (Fig. 2C); overall, a low hyperactivation of the signaling network was observed. But four days of chronic sorafenib treatment led to a striking shift: nearly all assayed phosphoproteins were strongly upregulated (Fig. 2C). Hyperactivation of RTKs such as EGFR, HER2/3, MET, FGFR1 were especially intriguing as these are known to promote drug resistance in different cancers (37, 38).

Figure 2.

Network brake drug cocktails restrained human thyroid cancer networks. A, MTT viability assay curves indicate that bortezomib/vorinostat reduced viability of MZ-CRC-1 cells at moderate as well as high doses of sorafenib, including significant reduction of IC50. IC50 are in parentheses; doses of bortezomib (6 nmol/L) and vorinostat (6 nmol/L and 50 nmol/L) are indicated. For each dose–response curve, in this figure as well as subsequent ones, the CI is indicated below and qualitative synergy as per previous studies (5) represented. CI < 0.1 (+++++, very strong synergism), CI = 0.1–0.3 (++++, strong synergism), CI = 0.3–0.7 (+++, synergism), CI = 0.7–0.85 (++, moderate synergism), CI = 0.85–0.9 (+, slight synergism), CI = 0.9–1.10 (±, nearly additive). B, AD80, a lead polypharmacologic drug, efficacy also improved in the presence of bortezomib/vorinostat. Dosing, IC50s, CI, and synergism are indicated. C, Phosphoprotein array results of MZ-CRC-1 cells were treated with indicated drugs in the presence or absence of bortezomib (B, 6 nmol/L), vorinostat (V, 6 nmol/L), and sorafenib (S, 1 μmol/L). Sorafenib treatment alone led to hyperactivation of almost all phosphoprotein markers on the panel; sorafenib/bortezomib/vorinostat was most effective in suppressing overall network hyperactivation (bracket, asterisk). Quantitation of phospho-array analysis represented as PRISM software–generated heatmap. These data are represented as PRISM scatter plot in Fig. 5C. D, Additional broad-acting inhibitors act as network brakes at low subtherapeutic doses (see Fig. 6F for dose ranges). AUY922/vorinostat and bortezomib/CUDC-907 increased efficacy of different targeted therapies, sorafenib, and AD80. IC50s, CI, and synergism are indicated.

Figure 2.

Network brake drug cocktails restrained human thyroid cancer networks. A, MTT viability assay curves indicate that bortezomib/vorinostat reduced viability of MZ-CRC-1 cells at moderate as well as high doses of sorafenib, including significant reduction of IC50. IC50 are in parentheses; doses of bortezomib (6 nmol/L) and vorinostat (6 nmol/L and 50 nmol/L) are indicated. For each dose–response curve, in this figure as well as subsequent ones, the CI is indicated below and qualitative synergy as per previous studies (5) represented. CI < 0.1 (+++++, very strong synergism), CI = 0.1–0.3 (++++, strong synergism), CI = 0.3–0.7 (+++, synergism), CI = 0.7–0.85 (++, moderate synergism), CI = 0.85–0.9 (+, slight synergism), CI = 0.9–1.10 (±, nearly additive). B, AD80, a lead polypharmacologic drug, efficacy also improved in the presence of bortezomib/vorinostat. Dosing, IC50s, CI, and synergism are indicated. C, Phosphoprotein array results of MZ-CRC-1 cells were treated with indicated drugs in the presence or absence of bortezomib (B, 6 nmol/L), vorinostat (V, 6 nmol/L), and sorafenib (S, 1 μmol/L). Sorafenib treatment alone led to hyperactivation of almost all phosphoprotein markers on the panel; sorafenib/bortezomib/vorinostat was most effective in suppressing overall network hyperactivation (bracket, asterisk). Quantitation of phospho-array analysis represented as PRISM software–generated heatmap. These data are represented as PRISM scatter plot in Fig. 5C. D, Additional broad-acting inhibitors act as network brakes at low subtherapeutic doses (see Fig. 6F for dose ranges). AUY922/vorinostat and bortezomib/CUDC-907 increased efficacy of different targeted therapies, sorafenib, and AD80. IC50s, CI, and synergism are indicated.

Close modal

Low-dose bortezomib/vorinostat enhanced the growth-inhibitory effects of sorafenib: low-dose bortezomib plus vorinostat (6 nmol/L each) reduced sorafenib's IC50 20-fold to achieve similar efficacy (Fig. 2A), while the IC50 of AD80 was reduced 10-fold (Fig. 2B). Bortezomib/vorinostat cotreatment also restrained network hyperactivation: EGFR, HER2, MET, FGFR1, and other kinases were at lower levels compared with sorafenib treatment alone (Fig. 2C). That is, the growth-inhibitory effects of bortezomib/vorinostat/sorafenib correlated with the drug cocktail's ability to restrain broad network hyperactivation. Similarly, AUY922/vorinostat as well as bortezomib/CUDC-907, used as adjuncts, also strongly reduced the IC50s of sorafenib as well as AD80 (Fig. 2D; MZ-CRC-1, MTC cell line). Combination Index (CI) computes whether drug combinations work together in a synergistic, additive, or antagonistic manner (5). We found that these drug combinations were working synergistically on MZ cells (CI<0.9; Fig. 2A, B, and D; Supplementary Table S2).

Presence of cancer stem cells are associated with resistance in many cancers (39–41). Stem cell associated transcription factors (STF) like Sox2 are at the apex of multiple pathways of tumorigenesis, many of them regulating kinase networks like the IGF, EGF, and VEGF pathway (42) and cell-cycle progression components (43). Another STF, Oct4, binds the PTEN promoter to regulate Akt signaling in cancer cells (43). Oct4/Sox2 can also regulate cell-cycle progression indirectly through regulation of miRNAs (44). Finally, focal adhesion kinase (FAK) component PEAK1 increases Oct4, Nanog, and cMyc levels to promote 3D tumor sphere growth in pancreatic cancer, indicating a more complex relationship between kinase signaling networks and STFs. Thus, currently there is a significant effort to identify therapies that can regulate stemness in cancer (45) and we next explored the effects of the therapies we identified on STFs.

MZ-CRC-1 cells treated chronically with sorafenib upregulated levels of Sox2 (Fig. 3A), a stem cell fate marker previously demonstrated to promote a broad program of tumorigenesis (42, 46). Importantly, expression of Sox2 was strongly suppressed by addition of bortezomib alone or a bortezomib/vorinostat drug combination (Fig. 3A). Similar results were observed with TT cells, MTC-derived cells that harbor the oncogenic Ret isoform RetC634W: chronic sorafenib treatment had a moderate inhibitory effect on cell growth and signaling while promoting stem cell markers Sox2 (strongly) and c-Myc (moderately; Fig. 3B–D, Supplementary Fig. S7A–S7C, quantitation). Sox2 levels were suppressed by bortezomib and, to a lesser extent, by bortezomib/vorinostat (Fig. 3B and D; Supplementary Fig. S7C, quantitation). Importantly, bortezomib/vorinostat prevented accumulation of nuclear Sox2 (Fig. 3D) in TT cells, indicating that despite modest effects on overall protein levels, the treatment prevents enrichment in the nucleus, where Sox2 is required. Finally, cotreatment led to strong progressive upregulation of cleaved PARP in both MZ-CRC-1 and TT cells (Fig. 3B and C), indicating a significant increase in apoptotic cell death. Finally, in TT cells, the effect on broad kinase networks was similar, as sorafenib cotreatment with bortezomib alone or bortezomib/vorinostat further decreased overall levels of phosphoprotein network (Supplementary Fig. S7A and S7B).

Figure 3.

Network brake drug cocktails restrained induction of stem cell markers. A, MTT viability assay curves indicate that bortezomib/vorinostat reduced viability of TT cells at moderate as well as high doses of sorafenib, including significant reduction of IC50. IC50s are in parentheses; doses of bortezomib (0.75 nmol/L and 2 nmol/L) and vorinostat (0.75 nmol/L and 2 nmol/L) are indicated. IC50s, CI, and synergism are indicated. B, Western blot analysis of MZ-CRC-1 cells demonstrating that bortezomib/vorinostat restrained sorafenib-induced hyperactivation of the cancer stem cell marker Sox2. Levels of other stem cell markers Oct4, Nanog, and cMyc were reduced by the triple combination compared with untreated cells. Drug combinations promoted strong upregulation of the cell death marker cPARP. Data are quantitated in Supplementary Fig. S7C. C, Western blot analysis of TT cells demonstrating that sorafenib induced elevation of stem cell marker cMyc, and bortezomib/vorinostat restrained this activation. The triple-drug combination also reduced levels of Oct4 and Nanog compared with untreated cells. Sorafenib also induced Sox2 levels, which were suppressed moderately by bortezomib and, to a lesser extent, by bortezomib/vorinostat. The triple-drug combination strongly potentiated sorafenib-induced upregulation of cPARP. Data are quantitated in Supplementary Fig. S7C, with pixel densitometric analysis values of Sox2 indicated. D, Immunofluorescence staining showing that sorafenib-treated TT cells upregulated levels of nuclear Sox2. This upregulation was blocked in cells treated with sorafenib (0.5 μmol/L) plus bortezomib (2 nmol/L) plus vorinostat (2 nmol/L). Higher doses of bortezomib/vorinostat (4 nmol/L each and 6 nmol/L each) completely suppressed sorafenib-induced upregulation of Sox2 (Supplementary Fig. S9A). E, Subcutaneous mouse xenograft assays using human TT cells showing median relative tumor growth of: (i) vehicle-treated controls, (ii) sorafenib (S; 40 mg/kg) + bortezomib (B; 0.05 mg/kg)-treated, and (iii) sorafenib (40 mg/kg) + bortezomib (0.05 mg/kg) + vorinostat (V; 10 mg/kg)-treated animals. Eight animals for vehicle treatment and five each for drug treatments were used to analyze relative tumor growth, defined as the difference between median size of tumors at initiation of drug dosing to indicated timepoints. F, Relative growth of each tumor for each treatment class at the end of the experiment in E represented as a column graph plot. The mean relative growth of each class was compared for statistically significant differences using Student t test (Welch correction). Asterisks indicate two animals in S+B+V group showing tumor regression.

Figure 3.

Network brake drug cocktails restrained induction of stem cell markers. A, MTT viability assay curves indicate that bortezomib/vorinostat reduced viability of TT cells at moderate as well as high doses of sorafenib, including significant reduction of IC50. IC50s are in parentheses; doses of bortezomib (0.75 nmol/L and 2 nmol/L) and vorinostat (0.75 nmol/L and 2 nmol/L) are indicated. IC50s, CI, and synergism are indicated. B, Western blot analysis of MZ-CRC-1 cells demonstrating that bortezomib/vorinostat restrained sorafenib-induced hyperactivation of the cancer stem cell marker Sox2. Levels of other stem cell markers Oct4, Nanog, and cMyc were reduced by the triple combination compared with untreated cells. Drug combinations promoted strong upregulation of the cell death marker cPARP. Data are quantitated in Supplementary Fig. S7C. C, Western blot analysis of TT cells demonstrating that sorafenib induced elevation of stem cell marker cMyc, and bortezomib/vorinostat restrained this activation. The triple-drug combination also reduced levels of Oct4 and Nanog compared with untreated cells. Sorafenib also induced Sox2 levels, which were suppressed moderately by bortezomib and, to a lesser extent, by bortezomib/vorinostat. The triple-drug combination strongly potentiated sorafenib-induced upregulation of cPARP. Data are quantitated in Supplementary Fig. S7C, with pixel densitometric analysis values of Sox2 indicated. D, Immunofluorescence staining showing that sorafenib-treated TT cells upregulated levels of nuclear Sox2. This upregulation was blocked in cells treated with sorafenib (0.5 μmol/L) plus bortezomib (2 nmol/L) plus vorinostat (2 nmol/L). Higher doses of bortezomib/vorinostat (4 nmol/L each and 6 nmol/L each) completely suppressed sorafenib-induced upregulation of Sox2 (Supplementary Fig. S9A). E, Subcutaneous mouse xenograft assays using human TT cells showing median relative tumor growth of: (i) vehicle-treated controls, (ii) sorafenib (S; 40 mg/kg) + bortezomib (B; 0.05 mg/kg)-treated, and (iii) sorafenib (40 mg/kg) + bortezomib (0.05 mg/kg) + vorinostat (V; 10 mg/kg)-treated animals. Eight animals for vehicle treatment and five each for drug treatments were used to analyze relative tumor growth, defined as the difference between median size of tumors at initiation of drug dosing to indicated timepoints. F, Relative growth of each tumor for each treatment class at the end of the experiment in E represented as a column graph plot. The mean relative growth of each class was compared for statistically significant differences using Student t test (Welch correction). Asterisks indicate two animals in S+B+V group showing tumor regression.

Close modal

In whole-animal experiments, growth of established TT cell mouse xenografts were potently inhibited with sorafenib/bortezomib/vorinostat compared with vehicle or sorafenib/bortezomib treatments. Treating with all three drugs together inhibited tumor growth and, in some animals, induced tumor regression (Fig. 3E and F; Supplementary Fig. S7D). Of note, sorafenib worked potently at 40 mg/kg in combination with bortezomib/vorinostat, a concentration lower than many previous xenograft studies (47, 48). Higher doses of sorafenib treatment alone, at 60 mg/kg, also induced tumor regression, which was further enhanced in the presence of bortezomib/vorinostat (Supplementary Fig. S7E and S7F). This indicated the usefulness of using bortezomib/vorinostat across a range of sorafenib doses in mouse whole-animal models. Together, these data indicate that, similar to our whole-animal Drosophila data, broadly acting cocktails that include combinations such as bortezomib/vorinostat can act as “network brakes” in at least two thyroid cancer paradigms.

Bortezomib/vorinostat was effective against multiple cancer cell types

To assess the broader applicability of our results, we assembled a panel of cancer cell lines with defined genetic mutations that respond to different targeted kinase inhibitors. The NSCLC cell line H358 exhibits high EGFR activity and is sensitive to erlotinib, an EGFR inhibitor approved for NSCLC (49, 50). Chronic erlotinib treatment led to progressive hyperactivation of the kinase network as assessed by phospho-kinase array analysis; this effect was suppressed and the IC50 of erlotinib reduced significantly in the presence of bortezomib/vorinostat. A similar pattern was observed with the NSCLC cell line H1299, which contains an oncogenic isoform of K-RAS: the approved standard-of-care drug trametinib led to network hyperactivation that was restrained by bortezomib/vorinostat (Fig. 4A–C). In each case, this restraint coincided with an increase in efficacy.

Figure 4.

Network brake drug cocktails restrained lung cancer cell networks. A, H358 cells were treated with indicated drugs in the presence or absence of bortezomib (B; 6 nmol/L), vorinostat (V; 50 nmol/L), and erlotinib (E; 1 μmol/L). Erlotinib induced hyperactivation of various phosphoproteins, which was restrained in the presence of bortezomib/vorinostat (bracket, asterisk). Quantitation of phospho-array analysis represented as PRISM software–generated heatmap. The data are represented as PRISM scatter plot in Fig. 5C. IC50s, CI, and synergism are indicated. B, H1299-NSCLC cells treated with targeted therapy, trametinib, or trametinib in combination with low-dose network brake drugs bortezomib (B, 0.25 nmol/L)/vorinostat (V, 0.25 nmol/L). Heatmap generated as in A. Trametinib induced hyperactivation of many phosphoproteins, which was restrained in the presence of bortezomib/vorinostat (bracket, asterisk). A rare signal showed higher levels in the presence of three drugs (pAKT-308). The data are represented as PRISM scatter plot in Fig. 5C. C, An MTT viability assay demonstrated that bortezomib/vorinostat significantly lowered the IC50 (in parentheses) of erlotinib on H358 cells and trametinib on H1299 cells. The 3-drug combination doses were the same as the ones used in A and B above. D, Western blot analysis of H358 cells shows that the bortezomib/vorinostat combination restrained erlotinib-induced upregulation of Sox2. Erlotinib also upregulated cMyc, which was moderately reduced by bortezomib/vorinostat. Other stem cell markers were kept below untreated levels, including EMT marker vimentin. The cell death marker cPARP was significantly upregulated in the presence of bortezomib and maintained at similar levels to erlotinib treatment alone by bortezomib/vorinostat combination.

Figure 4.

Network brake drug cocktails restrained lung cancer cell networks. A, H358 cells were treated with indicated drugs in the presence or absence of bortezomib (B; 6 nmol/L), vorinostat (V; 50 nmol/L), and erlotinib (E; 1 μmol/L). Erlotinib induced hyperactivation of various phosphoproteins, which was restrained in the presence of bortezomib/vorinostat (bracket, asterisk). Quantitation of phospho-array analysis represented as PRISM software–generated heatmap. The data are represented as PRISM scatter plot in Fig. 5C. IC50s, CI, and synergism are indicated. B, H1299-NSCLC cells treated with targeted therapy, trametinib, or trametinib in combination with low-dose network brake drugs bortezomib (B, 0.25 nmol/L)/vorinostat (V, 0.25 nmol/L). Heatmap generated as in A. Trametinib induced hyperactivation of many phosphoproteins, which was restrained in the presence of bortezomib/vorinostat (bracket, asterisk). A rare signal showed higher levels in the presence of three drugs (pAKT-308). The data are represented as PRISM scatter plot in Fig. 5C. C, An MTT viability assay demonstrated that bortezomib/vorinostat significantly lowered the IC50 (in parentheses) of erlotinib on H358 cells and trametinib on H1299 cells. The 3-drug combination doses were the same as the ones used in A and B above. D, Western blot analysis of H358 cells shows that the bortezomib/vorinostat combination restrained erlotinib-induced upregulation of Sox2. Erlotinib also upregulated cMyc, which was moderately reduced by bortezomib/vorinostat. Other stem cell markers were kept below untreated levels, including EMT marker vimentin. The cell death marker cPARP was significantly upregulated in the presence of bortezomib and maintained at similar levels to erlotinib treatment alone by bortezomib/vorinostat combination.

Close modal

Regarding stem cell markers, erlotinib-treated H358 NSCLC cells showed upregulation of Sox2 and c-Myc. Cotreatment with bortezomib/vorinostat restrained Sox2 activation quite effectively as well as c-Myc levels weakly. In addition, stem cell marker, LIN28, as well as epithelial-to-mesenchymal marker vimentin were both kept below untreated levels with bortezomib/vorinostat cotreatment (Fig. 4D). In H1299 cells, levels of stem cell markers c-Myc, KLF4, and EMT marker vimentin were reduced by treating with trametinib/bortezomib/vorinostat (Supplementary Fig. S8A and S8B) while in the HCC cell line HepG2 stem cell markers Oct4, Nanog, c-Myc, and EMT marker vimentin were reduced by trametinib/bortezomib/vorinostat (Supplementary Fig. S8C and S8D). In each cell line, high levels of cleaved PARP were observed with bortezomib/vorinostat cotreatment indicating increase in apoptotic death.

Broadening our survey, we tested whether bortezomib/vorinostat cotreatment improved the efficacy of kinase inhibitors that targeted deregulated pathways in other cancer cells. Synergy was observed with A549 cells (NSCLC), HepG2, and PLC5 (hepatocellular carcinoma), T47D (ER+ breast cancer), and A375 (melanoma; Fig. 5A and B). The IC50s of each standard-of-care targeted therapy was lowered significantly in the presence of bortezomib/vorinostat, ranging from modest reduction (5-fold in BEZ235-treated T47D cells) to more potent reduction (83-fold in trametinib-treated H1299 cells; Fig. 5B). We found the median level of the phospho-kinase network in the MTC and the NSCLC cells demonstrated a similar trend: presence of bortezomib/vorinostat as adjuncts restrained hyperactivation of the kinase network in each cell type (Fig. 5C), tracking with the increase in efficacy of the triple drug combinations. We conclude that in Drosophila and in human cancer cells, restraining the overall activity of the kinase network led to better drug efficacy.

Figure 5.

Network brake drug cocktails restrained multiple cancer cell networks. A, MTT viability assays of indicated cell lines with indicated drugs and doses. Trametinib (T); BEZ235 (BEZ); bortezomib (B); vorinostat (V). Bortezomib and vorinostat doses are in nmol/L. IC50s are in parentheses, and CI and synergism are indicated. B, IC50s of various kinase inhibitor drugs on different cancer lines were lowered significantly in the presence of indicated dose of bortezomib/vorinostat (from Figs. 2A, 4C, and 5A). S, sorafenib; E, erlotinib; T, trametinib; BEZ, BEZ235; V, vemurafenib. C, Scatter plot (PRISM) summary of phosphoprotein array data on indicated cancer lines treated with sorafenib (S), erlotinib (E), trametinib (T) alone or in combination with bortezomib/vorinostat. Each phosphoprotein signal was compared between treatments, and median level of entire network is indicated with blue line; interquartile range is indicated. Bortezomib/vorinostat treatment consistently reduced median phosphoprotein level of each cancer network. Paired t tests (PRISM) between single- and triple-treated samples for each cell line indicated P values <0.05 (asterisks).

Figure 5.

Network brake drug cocktails restrained multiple cancer cell networks. A, MTT viability assays of indicated cell lines with indicated drugs and doses. Trametinib (T); BEZ235 (BEZ); bortezomib (B); vorinostat (V). Bortezomib and vorinostat doses are in nmol/L. IC50s are in parentheses, and CI and synergism are indicated. B, IC50s of various kinase inhibitor drugs on different cancer lines were lowered significantly in the presence of indicated dose of bortezomib/vorinostat (from Figs. 2A, 4C, and 5A). S, sorafenib; E, erlotinib; T, trametinib; BEZ, BEZ235; V, vemurafenib. C, Scatter plot (PRISM) summary of phosphoprotein array data on indicated cancer lines treated with sorafenib (S), erlotinib (E), trametinib (T) alone or in combination with bortezomib/vorinostat. Each phosphoprotein signal was compared between treatments, and median level of entire network is indicated with blue line; interquartile range is indicated. Bortezomib/vorinostat treatment consistently reduced median phosphoprotein level of each cancer network. Paired t tests (PRISM) between single- and triple-treated samples for each cell line indicated P values <0.05 (asterisks).

Close modal

Together, these results indicate that, across a large number of human cancer cell types, bortezomib/vorinostat cotreatment enhanced efficacy of established therapies by restraining network hyperactivation and reducing levels of protumorigenic markers.

Bortezomib/vorinostat restrained emergence of drug resistance

Cancer cells lose sensitivity to targeted kinase inhibitor therapies by upregulating multiple cellular kinases and RTKs, leading to drug resistance (51). As bortezomib/vorinostat cotreatment restrained hyperactivation of kinase networks, we hypothesized that it would also delay or prevent drug resistance over extended treatment periods.

Cellular resistance to erlotinib in the clinic and in NSCLC human cancer cells such as H358 have been well documented (52, 53). H358 cells became insensitive to erlotinib by approximately 35 days of chronic treatment: parental cells had an IC50 of 0.5 μmol/L, while resistant lines displayed an 8-fold increase to 4 μmol/L (Fig. 6A and B). Chronic cotreatment with erlotinib/bortezomib/vorinostat prevented resistance: H358 cells retained sensitivity to erlotinib similar to parental cells (Fig. 6A and B; IC50 of 0.7 μmol/L).

Figure 6.

Network brake drug cocktails delayed emergence of resistance. A, H358 cells treated chronically with 1 μmol/L erlotinib for 65 days developed resistance to erlotinib (IC50 = 4 μmol/L). Cells treated with erlotinib + bortezomib (6 nmol/L) + vorinostat (50 nmol/L) retained erlotinib sensitivity similar to the parental line. IC50s in parentheses. For these experiments, when single-drug–treated cells (erlotinib or vemurafenib) started growing similar to DMSO control cells, the drug-resistant cells as well as the triple-drug–treated cells were further amplified for three generations in the absence of any treatment. This allowed the drug-resistant cells and triple-drug cocktail–treated cells to grow to sufficient numbers for the various assays, that is, phospho-kinome array, Western blot analysis, and MTT viability assay. B, Phase contrast images provide examples from A. Only erlotinib/bortezomib/vorinostat treatment strongly reduced cell number. See Materials and Methods for how triple-drug–treated cells were amplified and tested. C, Parental, erlotinib resistant, and erlotinib/bortezomib/vorinostat (E+B+V) treated cell lysates analyzed on phosphoprotein arrays. Outlined in red are phosphoproteins whose levels show much higher signals in the resistant line compared with the other two cell treatment conditions. Identities are listed beneath boxes; for example c-Met is a protein known to promote resistance to erlotinib treatment in NSCLC (37). D, Western blot analysis of lysates from erlotinib-resistant H358 NSCLC cells. Resistant lines (res.) upregulated Sox2, RhoA, and active-β-catenin; erlotinib/bortezomib/vorinostat treatment kept these below parental H358 cells, while increasing activity of the tumor suppressor MOB. Active histone marks H3K9-Ac, H3K4-Me3, and H4K5-Ac were upregulated in erlotinib-resistant cells. Erlotinib/bortezomib/vorinostat blocked this upregulation while elevating the repressive mark H3K27Me3. E, Phase contrast images of 239-melanoma cells lines treated with the indicated conditions for 120 days. Cells treated with 0.5 μmol/L vemurafenib developed resistance to the drug and grew confluently similar to the parental line. The 239 cells treated continuously with combined vemurafenib/bortezomib/vorinostat (0.5 μmol/L, 6 nmol/L, and 50 nmol/L, respectively) displayed minimal growth over 120 days of culturing. The combination of bortezomib/vorinostat (6 nmol/L, 50 nmol/L) without targeted therapy vemurafenib grew similar to untreated parental lines. F, Drug–response curve for each network brake drug as single agents—vorinostat, bortezomib, CUDC-907, AUY922—on various different cancer cell lines tested in our study. Shaded gray-boxed area indicates low nanomolar doses used for each of these drugs in these studies. At these doses, individual network brake drugs have very little effect on cancer cell line viability. Relevant viability assay results for network brake drug combinations by themselves on cancer lines provided in Supplementary Fig. S9B.

Figure 6.

Network brake drug cocktails delayed emergence of resistance. A, H358 cells treated chronically with 1 μmol/L erlotinib for 65 days developed resistance to erlotinib (IC50 = 4 μmol/L). Cells treated with erlotinib + bortezomib (6 nmol/L) + vorinostat (50 nmol/L) retained erlotinib sensitivity similar to the parental line. IC50s in parentheses. For these experiments, when single-drug–treated cells (erlotinib or vemurafenib) started growing similar to DMSO control cells, the drug-resistant cells as well as the triple-drug–treated cells were further amplified for three generations in the absence of any treatment. This allowed the drug-resistant cells and triple-drug cocktail–treated cells to grow to sufficient numbers for the various assays, that is, phospho-kinome array, Western blot analysis, and MTT viability assay. B, Phase contrast images provide examples from A. Only erlotinib/bortezomib/vorinostat treatment strongly reduced cell number. See Materials and Methods for how triple-drug–treated cells were amplified and tested. C, Parental, erlotinib resistant, and erlotinib/bortezomib/vorinostat (E+B+V) treated cell lysates analyzed on phosphoprotein arrays. Outlined in red are phosphoproteins whose levels show much higher signals in the resistant line compared with the other two cell treatment conditions. Identities are listed beneath boxes; for example c-Met is a protein known to promote resistance to erlotinib treatment in NSCLC (37). D, Western blot analysis of lysates from erlotinib-resistant H358 NSCLC cells. Resistant lines (res.) upregulated Sox2, RhoA, and active-β-catenin; erlotinib/bortezomib/vorinostat treatment kept these below parental H358 cells, while increasing activity of the tumor suppressor MOB. Active histone marks H3K9-Ac, H3K4-Me3, and H4K5-Ac were upregulated in erlotinib-resistant cells. Erlotinib/bortezomib/vorinostat blocked this upregulation while elevating the repressive mark H3K27Me3. E, Phase contrast images of 239-melanoma cells lines treated with the indicated conditions for 120 days. Cells treated with 0.5 μmol/L vemurafenib developed resistance to the drug and grew confluently similar to the parental line. The 239 cells treated continuously with combined vemurafenib/bortezomib/vorinostat (0.5 μmol/L, 6 nmol/L, and 50 nmol/L, respectively) displayed minimal growth over 120 days of culturing. The combination of bortezomib/vorinostat (6 nmol/L, 50 nmol/L) without targeted therapy vemurafenib grew similar to untreated parental lines. F, Drug–response curve for each network brake drug as single agents—vorinostat, bortezomib, CUDC-907, AUY922—on various different cancer cell lines tested in our study. Shaded gray-boxed area indicates low nanomolar doses used for each of these drugs in these studies. At these doses, individual network brake drugs have very little effect on cancer cell line viability. Relevant viability assay results for network brake drug combinations by themselves on cancer lines provided in Supplementary Fig. S9B.

Close modal

We used phospho-kinase array analysis to explore the mechanisms by which bortezomib/vorinostat blocked drug resistance. Erlotinib-resistant H358 NSCLC cells showed upregulation of a large network of kinases when compared with the parental cell line (Fig. 6C). This included ERBB3, Met, c-Kit, Src, and Stat3, pathway effectors associated with drug resistance (37, 38, 54, 55), suggesting that broad network upregulation contributes to cellular resistance. Consistent with this view, erlotinib-resistant cells that upregulated Met were almost 3-fold more responsive to the Met inhibitor crizotinib (Supplementary Fig. S8E). Erlotinib-resistant cells also showed increased levels of Sox2 and activation of protumorigenic proteins such as β-Catenin and RhoA. The combination erlotinib/bortezomib/vorinostat prevented these increases (Fig. 6D). Interestingly, the three-drug combination elevated levels of phosphorylated Mob, a key effector of the Hippo signaling pathway that suppresses growth (Fig. 6D; Supplementary Fig. S8F and S8G; ref. 56). Similar effects were observed with Mel-239, a melanoma cell line harboring oncogenic BRAFV600E: 90 days of treatment led to resistance to the standard-of-care drug vemurafenib, while bortezomib/vorinostat–treated cells retained sensitivity (Fig. 6E), but we did not explore the molecular mechanism underlying resistance in these cells.

HDAC inhibitors such as vorinostat alter histone modifications to exert broad effects on transcription (57) and can restrain stem cell markers in different cancers (58, 59) as well as oppose tumorigenic states by restoring epithelial markers (60). We examined whether alterations in histone modifications could contribute to the broad network changes provoked by targeted therapies. Erlotinib-resistant lines showed strong upregulation of the active histone mark H3K9-Ac and moderate activation of the active marks H4K5-Ac and H3K4-Me3 (Fig. 6D). Treatment with erlotinib/bortezomib/vorinostat exhibited strongly reduced levels of these active marks and elevated levels of the inactive histone mark H3K27-Me3 (Fig. 6D). An important finding of these studies is that low doses of “network brake” drugs as single agents (Fig. 6F) or in combination with other “network brake” drugs (Supplementary Fig. S9A and S9B) did not significantly affect cancer cell viability. In summary, addition of bortezomib plus vorinostat in our long-term experiments directed the histone code toward lowered transcription, prevented upregulation of overall cellular network activity, and achieved a balance of low protumorigenic and high tumor suppressor signals.

Preclinical and clinical cancer studies show that resistance eventually emerges even with kinase inhibitors with high target selectivity (61). Cancer cells respond to inhibition of oncogene-addicted pathways by finding alternative mechanisms to provide high signaling through these addicted pathways, or by shifting their dependence to other pathways (62). For example, at least six different mechanisms by which cancer cells develop resistance to drugs targeting BRAFV600E have been identified (61, 63). This highlights the adaptability and high degree of connectivity within the kinase signaling network: inhibiting signaling in one part of the network provokes responses in other parts (Fig. 7, model).

Figure 7.

Model: an approach to increase efficacy of targeted cancer therapeutics. Targeted therapeutics, for example, sorafenib, inhibit a number of kinases (red circles; strategy), but over the course of treatment, cancer cells respond by hyperactivation of large number of kinases (effect). This provides cancer cells avenues for resistance to therapy, but is also the source of toxicity in normal cells (outcome). Inclusion of low doses of broad-acting “network brake” drugs (gray), as an adjunct to targeted therapy could directly or indirectly restrain the hyperactivation of different cellular kinase subgroups. The effect is better treatment with high efficacy, low toxicity, and potentially less drug resistance. Generation of rendered kinome tree with target inhibition pattern explained in Materials and Methods.

Figure 7.

Model: an approach to increase efficacy of targeted cancer therapeutics. Targeted therapeutics, for example, sorafenib, inhibit a number of kinases (red circles; strategy), but over the course of treatment, cancer cells respond by hyperactivation of large number of kinases (effect). This provides cancer cells avenues for resistance to therapy, but is also the source of toxicity in normal cells (outcome). Inclusion of low doses of broad-acting “network brake” drugs (gray), as an adjunct to targeted therapy could directly or indirectly restrain the hyperactivation of different cellular kinase subgroups. The effect is better treatment with high efficacy, low toxicity, and potentially less drug resistance. Generation of rendered kinome tree with target inhibition pattern explained in Materials and Methods.

Close modal

Our current findings suggest one path toward addressing these major challenges in cancer therapeutics. Using “network brake” drugs as part of a cocktail would allow standard-of-care kinase inhibitors to be used at lower doses over longer periods. This would reduce the toxicity that may result from both strong on-target inhibition as well as off-target effects. In our analysis, the IC50s of at least five different kinase inhibitors were reduced considerably in the presence of low-dose bortezomib plus vorinostat. Furthermore, sorafenib/bortezomib/vorinostat inhibited tumor growth in xenografted mice at low doses, indicating that sorafenib, and perhaps other targeted therapies, can work potently in vertebrate models if paired with at least one “network brake” cocktail.

Restraining network hyperactivation prevented upregulation of stem cell markers, most prominently Sox2, a factor that controls stem cell fate during development and in cancer progression including EMT (42, 46, 64). In some cancer cells, restraining the network reduced other stem cell markers like Oct4, Nanog, and c-Myc below untreated levels, suggesting these combinations reduce overall “stemness.” A recent study found that therapy-induced upregulation of cellular components such as integrins contribute to “stemness” and drug resistance, a condition that was restrained by cotreatment with bortezomib (65). By allowing targeted therapies to function at lower doses, by restraining hyperactivation of the signaling network, by preventing the upregulation of stem cell markers, and by promoting increased death of cancer cells, “network brake” drugs hold the potential for improving the therapeutic index and longevity of a broad palate of targeted therapies (Fig. 7, model). One practical use of our findings could be for monitoring effectiveness of cancer therapy. Using phospho-kinome arrays or simpler assays such as serial Western blots or single-cell Western blots (66), would provide a window into how cancers cells' signaling network respond to treatment. Treatment-induced broad cellular network hyperactivation would indicate limited therapeutic benefits in the long run.

Our findings indicate that low doses of “network brake” drugs as single agents or in combination with other “network brake” drugs do not significantly affect cancer cell viability. The therapeutic potential of “network brake” drugs at these doses became apparent only when they were paired with optimal doses of the targeted therapies. For example, previous studies have shown that the combination of bortezomib plus vorinostat is effective against multiple myeloma and other cancers, although at significantly higher doses (25, 67, 68). Our studies indicate a potential mechanism as to how these drugs potentiate the effects of targeted therapies at low doses. Cancer cells rely on a subset of the available cellular signaling pathways, a phenomenon termed “oncogene addiction.” The mechanism by which cancer cells undergo apoptotic death when addicted pathways are inhibited is unclear. One theory is that cancer cells are reliant on fewer signaling pathways and therefore inhibition of the addicted pathways sends cancer cells into crisis thereby hastening death (69, 70). We and others have shown that targeted therapies can hyperactivate the overall cellular network; this may allow cancer cells to shift dependence to other pathways, providing a route for resistance (71). Our studies suggest that long-term treatment with “network brake” drug combination directed the histone code towards lowered transcription, which could be one possible way to prevent upregulation of overall cellular network activity. By restraining hyperactivation of the network in response to targeted therapies, the “network brake” drugs may block this alternative path. Broadly-acting drugs such as bortezomib and vorinostat can act on multiple targets; the most relevant target (s) can be difficult to identify. Furthermore, the “network brake” drugs may not always act solely through their intended targets, but our study shows that the presence of these drugs consistently restrains the broad cellular effects elicited by a broad palette of targeted therapies.

This study has identified multiple “network brake” drugs that can improve efficacy of targeted therapies: bortezomib (proteasome), vorinostat, and CUDC-907 (histone deacetylases), mithramycin (Sp1 transcriptome), and AUY922 (Hsp90 inhibitor). Other classes of broadly acting drugs—targeting cell cycle, metabolism, cytoskeleton, proteases, topoisomerases, mitochondria, etc.—may also prove useful. Currently, there are approximately 15,000 cancer clinical trials involving combinations of targeted therapies. One challenge will be to match specific “network brake” cocktails to particular tumors based on the details of the tumor network. Using a whole-animal approach will permit us to address the effects of these powerful drugs on transformed cells and, importantly, on normal untransformed tissue.

No potential conflicts of interest were disclosed.

Conception and design: T.K. Das, R. Cagan

Development of methodology: T.K. Das, J. Esernio

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.K. Das, J. Esernio

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.K. Das, J. Esernio, R. Cagan

Writing, review, and/or revision of the manuscript: T.K. Das, R. Cagan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.K. Das, J. Esernio

Study supervision: T.K. Das, R. Cagan

We thank members of the Cagan laboratory for technical assistance and for helpful discussions. We thank the Bloomington Drosophila Stock Center for Drosophila reagents. Microscopy was performed in part at the Microscopy Shared Resource Facility at the Icahn School of Medicine at Mount Sinai. The authors also thank Ran Brosh, Emily Bernstein, Stuart Aaronson, and Barry Nelkin for human cancer cells. This research was supported by NIH grants R01-CA170495 and R01-CA109730, Department of Defense grant W81XWH-15-1-0111, and American Cancer Society grants 120616-RSGM-11-018-01-CDD (to R.L. Cagan) and 120886-PFM-11-137-01-DDC (to T.K. Das).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
O'Connor
R
. 
A review of mechanisms of circumvention and modulation of chemotherapeutic drug resistance
.
Curr Cancer Drug Targets
2009
;
9
:
273
80
.
2.
Duncan
JS
,
Whittle
MC
,
Nakamura
K
,
Abell
AN
,
Midland
AA
,
Zawistowski
JS
, et al
Dynamic reprogramming of the kinome in response to targeted MEK inhibition in triple-negative breast cancer
.
Cell
2012
;
149
:
307
21
.
3.
Obenauf
AC
,
Zou
Y
,
Ji
AL
,
Vanharanta
S
,
Shu
W
,
Shi
H
, et al
Therapy-induced tumour secretomes promote resistance and tumour progression
.
Nature
2015
;
520
:
368
72
.
4.
Dar
AC
,
Das
TK
,
Shokat
KM
,
Cagan
RL
. 
Chemical genetic discovery of targets and anti-targets for cancer polypharmacology
.
Nature
2012
;
486
:
80
4
.
5.
Chou
TC
. 
Theoretical basis, experimental design, and computerized simulation of synergism and antagonism in drug combination studies
.
Pharmacol Rev
2006
;
58
:
621
81
.
6.
Das
TK
,
Dana
D
,
Paroly
SS
,
Perumal
SK
,
Singh
S
,
Jhun
H
, et al
Centrosomal kinase Nek2 cooperates with oncogenic pathways to promote metastasis
.
Oncogenesis
2013
;
2
:
e69
.
7.
Read
RD
,
Goodfellow
PJ
,
Mardis
ER
,
Novak
N
,
Armstrong
JR
,
Cagan
RL
. 
A Drosophila model of multiple endocrine neoplasia type 2
.
Genetics
2005
;
171
:
1057
81
.
8.
Salvatore
D
,
Domenico
S
,
Massimo
S
,
Michèle
G
,
Gianfranco
F
,
Vieri
G
, et al
Activation of the ret oncogene in human thyroid carcinomas
.
Rend Lincei Sci Fis Nat
1993
;
4
:
367
75
.
9.
Wells
SA
 Jr
,
Santoro
M
. 
Targeting the RET pathway in thyroid cancer
.
Clin Cancer Res
2009
;
15
:
7119
23
.
10.
Das
T
,
Cagan
R
. 
Drosophila as a novel therapeutic discovery tool for thyroid cancer
.
Thyroid
2010
;
20
:
689
95
.
11.
Plaza Menacho
I
,
Koster
R
,
van der Sloot
AM
,
Quax
WJ
,
Osinga
J
,
van der Sluis
T
, et al
RET-familial medullary thyroid carcinoma mutants Y791F and S891A activate a Src/JAK/STAT3 pathway, independent of glial cell line-derived neurotrophic factor
.
Cancer Res
2005
;
65
:
1729
37
.
12.
De Vita
G
,
Melillo
RM
,
Carlomagno
F
,
Visconti
R
,
Castellone
MD
,
Bellacosa
A
, et al
Tyrosine 1062 of RET-MEN2A mediates activation of Akt (protein kinase B) and mitogen-activated protein kinase pathways leading to PC12 cell survival
.
Cancer Res
2000
;
60
:
3727
31
.
13.
Tsumagari
K
,
Abd Elmageed
ZY
,
Sholl
AB
,
Friedlander
P
,
Abdraboh
M
,
Xing
M
, et al
Simultaneous suppression of the MAP kinase and NF-κB pathways provides a robust therapeutic potential for thyroid cancer
.
Cancer Lett
2015
;
368
:
46
53
.
14.
Luong
QT
,
O'Kelly
J
,
Braunstein
GD
,
Hershman
JM
,
Koeffler
HP
. 
Antitumor activity of suberoylanilide hydroxamic acid against thyroid cancer cell lines in vitro and in vivo
.
Clin Cancer Res
2006
;
12
:
5570
7
.
15.
Kim
SH
,
Kang
JG
,
Kim
CS
,
Ihm
SH
,
Choi
MG
,
Yoo
HJ
, et al
Synergistic cytotoxicity of BIIB021 with triptolide through suppression of PI3K/Akt/mTOR and NF-κB signal pathways in thyroid carcinoma cells
.
Biomed Pharmacother
2016
;
83
:
22
32
.
16.
Krajewska
J
,
Handkiewicz-Junak
D
,
Jarzab
B
. 
Sorafenib for the treatment of thyroid cancer: an updated review
.
Expert Opin Pharmacother
2015
;
16
:
573
83
.
17.
Poulikakos
PI
,
Zhang
C
,
Bollag
G
,
Shokat
KM
,
Rosen
N
. 
RAF inhibitors transactivate RAF dimers and ERK signalling in cells with wild-type BRAF
.
Nature
2010
;
464
:
427
30
.
18.
Sawamoto
K
,
Okabe
M
,
Tanimura
T
,
Mikoshiba
K
,
Nishida
Y
,
Okano
H
. 
The Drosophila secreted protein Argos regulates signal transduction in the Ras/MAPK pathway
.
Dev Biol
1996
;
178
:
13
22
.
19.
Kumar
SK
,
James
J
,
Randolph
M
,
Ronald
R
,
Fernando
Q
,
Timothy
M
, et al
Phase 1 study of sorafenib in combination with bortezomib in patients with advanced malignancies
.
Invest New Drugs
2013
;
31
:
1201
6
.
20.
Rao
A
,
Lauer
R
. 
Phase II study of sorafenib and bortezomib for first-line treatment of metastatic or unresectable renal cell carcinoma
.
Oncologist
2015
;
20
:
370
1
.
21.
Kaushik
D
,
Vashistha
V
,
Isharwal
S
,
Sediqe
SA
,
Lin
MF
. 
Histone deacetylase inhibitors in castration-resistant prostate cancer: molecular mechanism of action and recent clinical trials
.
Ther Adv Urol
2015
;
7
:
388
95
.
22.
Duvic
M
,
Madeleine
D
,
Jenny
V
. 
Vorinostat: a new oral histone deacetylase inhibitor approved for cutaneous T-cell lymphoma
.
Expert Opin Investig Drugs
2007
;
16
:
1111
20
.
23.
Chun
P
. 
Histone deacetylase inhibitors in hematological malignancies and solid tumors
.
Arch Pharm Res
2015
;
38
:
933
49
.
24.
Sleiman
SF
,
Langley
BC
,
Basso
M
,
Berlin
J
,
Xia
L
,
Payappilly
JB
, et al
Mithramycin is a gene-selective Sp1 inhibitor that identifies a biological intersection between cancer and neurodegeneration
.
J Neurosci
2011
;
31
:
6858
70
.
25.
Kikuchi
J
,
Wada
T
,
Shimizu
R
,
Izumi
T
,
Akutsu
M
,
Mitsunaga
K
, et al
Histone deacetylases are critical targets of bortezomib-induced cytotoxicity in multiple myeloma
.
Blood
2010
;
116
:
406
17
.
26.
Tovar
V
,
Cornella
H
,
Moeini
A
,
Vidal
S
,
Hoshida
Y
,
Sia
D
, et al
Tumour initiating cells and IGF/FGF signalling contribute to sorafenib resistance in hepatocellular carcinoma
.
Gut
2017
;
66
:
530
40
.
27.
Xu
Y
,
Huang
J
,
Ma
L
,
Shan
J
,
Shen
J
,
Yang
Z
, et al
MicroRNA-122 confers sorafenib resistance to hepatocellular carcinoma cells by targeting IGF-1R to regulate RAS/RAF/ERK signaling pathways
.
Cancer Lett
2016
;
371
:
171
81
.
28.
Blivet-Van Eggelpoël
MJ
,
Chettouh
H
,
Fartoux
L
,
Aoudjehane
L
,
Barbu
V
,
Rey
C
, et al
Epidermal growth factor receptor and HER-3 restrict cell response to sorafenib in hepatocellular carcinoma cells
.
J Hepatol
2012
;
57
:
108
15
.
29.
Ungerleider
N
,
Han
C
,
Zhang
J
,
Yao
L
,
Wu
T
. 
TGFβ signaling confers sorafenib resistance via induction of multiple RTKs in hepatocellular carcinoma cells
.
Mol Carcinog
2017
;
56
:
1302
11
.
30.
Dong
J
,
Zhai
B
,
Sun
W
,
Hu
F
,
Cheng
H
,
Xu
J
. 
Activation of phosphatidylinositol 3-kinase/AKT/snail signaling pathway contributes to epithelial-mesenchymal transition-induced multi-drug resistance to sorafenib in hepatocellular carcinoma cells
.
PLoS One
2017
;
12
:
e0185088
.
31.
Sie
M
,
den Dunnen
WFA
,
Lourens
HJ
,
Meeuwsen-de Boer
TGJ
,
Scherpen
FJG
,
Zomerman
WW
, et al
Growth-factor-driven rescue to receptor tyrosine kinase (RTK) inhibitors through Akt and Erk phosphorylation in pediatric low grade astrocytoma and ependymoma
.
PLoS One
2015
;
10
:
e0122555
.
32.
Zhou
Q
,
Guo
X
,
Choksi
R
. 
Activation of focal adhesion kinase and Src mediates acquired sorafenib resistance in A549 human lung adenocarcinoma xenografts
.
J Pharmacol Exp Ther
2017
;
363
:
428
43
.
33.
Flaherty
KT
,
Robert
C
,
Hersey
P
,
Nathan
P
,
Garbe
C
,
Milhem
M
, et al
Improved survival with MEK inhibition in BRAF-mutated melanoma
.
N Engl J Med
2012
;
367
:
107
14
.
34.
Duncan
JS
,
Whittle
MC
,
Nakamura
K
,
Abell
AN
,
Midland
AA
,
Zawistowski
JS
, et al
Dynamic reprogramming of the kinome in response to targeted MEK inhibition in triple-negative breast cancer
.
Cell
2012
;
149
:
307
21
.
35.
Cooley
LD
,
Elder
FF
,
Knuth
A
,
Gagel
RF
. 
Cytogenetic characterization of three human and three rat medullary thyroid carcinoma cell lines
.
Cancer Genet Cytogenet
1995
;
80
:
138
49
.
36.
Koh
YW
,
Shah
MH
,
Agarwal
K
,
McCarty
SK
,
Koo
BS
,
Brendel
VJ
, et al
Sorafenib and Mek inhibition is synergistic in medullary thyroid carcinoma in vitro
.
Endocr Relat Cancer
2012
;
19
:
29
38
.
37.
Engelman
JA
,
Zejnullahu
K
,
Mitsudomi
T
,
Song
Y
,
Hyland
C
,
Park
JO
, et al
MET amplification leads to gefitinib resistance in lung cancer by activating ERBB3 signaling
.
Science
2007
;
316
:
1039
43
.
38.
Grøvdal
LM
,
Kim
J
,
Holst
MR
,
Knudsen
SLJ
,
Grandal
MV
,
van Deurs
B
. 
EGF receptor inhibitors increase ErbB3 mRNA and protein levels in breast cancer cells
.
Cell Signal
2012
;
24
:
296
301
.
39.
Reya
T
,
Morrison
SJ
,
Clarke
MF
,
Weissman
IL
. 
Stem cells, cancer, and cancer stem cells
.
Nature
2001
;
414
:
105
11
.
40.
Borah
A
,
Raveendran
S
,
Rochani
A
,
Maekawa
T
,
Kumar
DS
. 
Targeting self-renewal pathways in cancer stem cells: clinical implications for cancer therapy
.
Oncogenesis
2015
;
4
:
e177
.
41.
Tan
J
,
Yu
Q
. 
Molecular mechanisms of tumor resistance to PI3K-mTOR-targeted therapy
.
Chin J Cancer
2013
;
32
:
376
9
.
42.
Siegle
JM
,
Basin
A
,
Sastre-Perona
A
,
Yonekubo
Y
,
Brown
J
,
Sennett
R
, et al
SOX2 is a cancer-specific regulator of tumour initiating potential in cutaneous squamous cell carcinoma
.
Nat Commun
2014
;
5
:
4511
.
43.
Singh
DK
,
Kollipara
RK
,
Vemireddy
V
,
Yang
XL
,
Sun
Y
,
Regmi
N
, et al
Oncogenes activate an autonomous transcriptional regulatory circuit that drives glioblastoma
.
Cell Rep
2017
;
18
:
961
76
.
44.
Card
DAG
,
Hebbar
PB
,
Li
L
,
Trotter
KW
,
Komatsu
Y
,
Mishina
Y
, et al
Oct4/Sox2-regulated miR-302 targets cyclin D1 in human embryonic stem cells
.
Mol Cell Biol
2008
;
28
:
6426
38
.
45.
Safa
AR
. 
Resistance to cell death and its modulation in cancer stem cells
.
Crit Rev Oncog
2016
;
21
:
203
19
.
46.
Boumahdi
S
,
Driessens
G
,
Lapouge
G
,
Rorive
S
,
Nassar
D
,
Le Mercier
M
, et al
SOX2 controls tumour initiation and cancer stem-cell functions in squamous-cell carcinoma
.
Nature
2014
;
511
:
246
50
.
47.
Carlomagno
F
,
Anaganti
S
,
Guida
T
,
Salvatore
G
,
Troncone
G
,
Wilhelm
SM
, et al
BAY 43-9006 inhibition of oncogenic RET mutants
.
J Natl Cancer Inst
2006
;
98
:
326
34
.
48.
Salvatore
G
,
De Falco
V
,
Salerno
P
,
Nappi
TC
,
Pepe
S
,
Troncone
G
, et al
BRAF is a therapeutic target in aggressive thyroid carcinoma
.
Clin Cancer Res
2006
;
12
:
1623
9
.
49.
Shepherd
FA
,
Rodrigues Pereira
J
,
Ciuleanu
T
,
Tan
EH
,
Hirsh
V
,
Thongprasert
S
, et al
Erlotinib in previously treated non-small-cell lung cancer
.
N Engl J Med
2005
;
353
:
123
32
.
50.
Bezjak
A
,
Tu
D
,
Seymour
L
,
Clark
G
,
Trajkovic
A
,
Zukin
M
, et al
Symptom improvement in lung cancer patients treated with erlotinib: quality of life analysis of the National Cancer Institute of Canada Clinical Trials Group Study BR.21
.
J Clin Oncol
2006
;
24
:
3831
7
.
51.
Holohan
C
,
Van Schaeybroeck
S
,
Longley
DB
,
Johnston
PG
. 
Cancer drug resistance: an evolving paradigm
.
Nat Rev Cancer
2013
;
13
:
714
26
.
52.
Haber
DA
,
Bell
DW
,
Sordella
R
,
Kwak
EL
,
Godin-Heymann
N
,
Sharma
SV
, et al
Molecular targeted therapy of lung cancer: EGFR mutations and response to EGFR inhibitors
.
Cold Spring Harb Symp Quant Biol
2005
;
70
:
419
26
.
53.
Fong
JT
,
Jacobs
RJ
,
Moravec
DN
,
Uppada
SB
,
Botting
GM
,
Nlend
M
, et al
Alternative signaling pathways as potential therapeutic targets for overcoming EGFR and c-Met inhibitor resistance in non-small cell lung cancer
.
PLoS One
2013
;
8
:
e78398
.
54.
Kanda
R
,
Kawahara
A
,
Watari
K
,
Murakami
Y
,
Sonoda
K
,
Maeda
M
, et al
Erlotinib resistance in lung cancer cells mediated by integrin β1/Src/Akt-driven bypass signaling
.
Cancer Res
2013
;
73
:
6243
53
.
55.
Li
R
,
Hu
Z
,
Sun
SY
,
Chen
ZG
,
Owonikoko
TK
,
Sica
GL
, et al
Niclosamide overcomes acquired resistance to erlotinib through suppression of STAT3 in non-small cell lung cancer
.
Mol Cancer Ther
2013
;
12
:
2200
12
.
56.
Lignitto
L
,
Arcella
A
,
Sepe
M
,
Rinaldi
L
,
Delle Donne
R
,
Gallo
A
, et al
Proteolysis of MOB1 by the ubiquitin ligase praja2 attenuates Hippo signalling and supports glioblastoma growth
.
Nat Commun
2013
;
4
:
1822
.
57.
Chen
HP
,
Zhao
YT
,
Zhao
TC
. 
Histone deacetylases and mechanisms of regulation of gene expression
.
Crit Rev Oncog
2015
;
20
:
35
47
.
58.
Chikamatsu
K
,
Ishii
H
,
Murata
T
,
Sakakura
K
,
Shino
M
,
Toyoda
M
, et al
Alteration of cancer stem cell-like phenotype by histone deacetylase inhibitors in squamous cell carcinoma of the head and neck
.
Cancer Sci
2013
;
104
:
1468
75
.
59.
Cai
MH
,
Xu
XG
,
Yan
SL
,
Sun
Z
,
Ying
Y
,
Wang
BK
, et al
Depletion of HDAC1, 7 and 8 by histone deacetylase inhibition confers elimination of pancreatic cancer stem cells in combination with gemcitabine
.
Sci Rep
2018
;
8
:
1621
.
60.
Tang
HM
,
Kuay
KT
,
Koh
PF
,
Asad
M
,
Tan
TZ
,
Chung
VY
, et al
An epithelial marker promoter induction screen identifies histone deacetylase inhibitors to restore epithelial differentiation and abolishes anchorage independence growth in cancers
.
Cell Death Discov
2016
;
2
:
16041
.
61.
Wilson
TR
,
Fridlyand
J
,
Yan
Y
,
Penuel
E
,
Burton
L
,
Chan
E
, et al
Widespread potential for growth-factor-driven resistance to anticancer kinase inhibitors
.
Nature
2012
;
487
:
505
9
.
62.
Glickman
MS
,
Sawyers
CL
. 
Converting cancer therapies into cures: lessons from infectious diseases
.
Cell
2012
;
148
:
1089
98
.
63.
Haarberg
HE
,
Smalley
KS
. 
Resistance to Raf inhibition in cancer
.
Drug Discov Today Technol
2014
;
11
:
27
32
.
64.
Mani
SA
,
Guo
W
,
Liao
MJ
,
Eaton
EN
,
Ayyanan
A
,
Zhou
AY
, et al
The epithelial-mesenchymal transition generates cells with properties of stem cells
.
Cell
2008
;
133
:
704
15
.
65.
Seguin
L
,
Kato
S
,
Franovic
A
,
Camargo
MF
,
Lesperance
J
,
Elliott
KC
, et al
An integrin β3-KRAS-RalB complex drives tumour stemness and resistance to EGFR inhibition
.
Nat Cell Biol
2014
;
16
:
457
68
.
66.
Sinkala
E
,
Sollier-Christen
E
,
Renier
C
,
Rosàs-Canyelles
E
,
Che
J
,
Heirich
K
, et al
Profiling protein expression in circulating tumour cells using microfluidic western blotting
.
Nat Commun
2017
;
8
:
14622
.
67.
Yu
C
,
Rahmani
M
,
Conrad
D
,
Subler
M
,
Dent
P
,
Grant
S
. 
The proteasome inhibitor bortezomib interacts synergistically with histone deacetylase inhibitors to induce apoptosis in Bcr/Abl+ cells sensitive and resistant to STI571
.
Blood
2003
;
102
:
3765
74
.
68.
Hideshima
T
,
Richardson
PG
,
Anderson
KC
. 
Mechanism of action of proteasome inhibitors and deacetylase inhibitors and the biological basis of synergy in multiple myeloma
.
Mol Cancer Ther
2011
;
10
:
2034
42
.
69.
Pellicano
F
,
Mukherjee
L
,
Holyoake
TL
. 
Concise review: cancer cells escape from oncogene addiction: understanding the mechanisms behind treatment failure for more effective targeting
.
Stem Cells
2014
;
32
:
1373
9
.
70.
Gillies
RJ
,
Verduzco
D
,
Gatenby
RA
. 
Evolutionary dynamics of carcinogenesis and why targeted therapy does not work
.
Nat Rev Cancer
2012
;
12
:
487
93
.
71.
Dar
AC
,
Das
TK
,
Shokat
KM
,
Cagan
RL
. 
Chemical genetic discovery of targets and anti-targets for cancer polypharmacology
.
Nature
2012
;
486
:
80
4
.