Vorinostat Potentiates 5-Fluorouracil/Cisplatin Combination by Inhibiting Chemotherapy-Induced EGFR Nuclear Translocation and Increasing Cisplatin Uptake

The combination of fluoropyrimidine with platinum analogue is still one of the most widely used chemotherapy regimen in several solid tumors, in different settings. However, toxic side effects and drug resistance, especially in patients with recurrent and metastatic disease, are the major factors in preventing favorable outcomes (1–4).

Several mechanisms of resistance to 5-fluorouracil (5FU) have been described, including high levels of thymidylate synthase (TS) expression, the essential enzyme for de novo synthesis of thymidylate and subsequently DNA synthesis, and the critical target for 5FU (5). Increased levels of TS represent one of the most common and well-described 5FU resistance mechanism (6–9), also in head and neck cancer models (10), and have been correlated with poorer overall patient survival in several tumors (5). Similarly, cisplatin (CDDP) resistance mechanisms are well described and are related to cell targets, uptake, and metabolism (11). Among these events, it was demonstrated that tumor cells treated with CDDP increased both EGFR activation and nuclear translocation as a prosurvival response to cytotoxic treatment (12–14). Moreover, CDDP-resistant cells showed lower basal levels of solute carrier family 31 member 1 transporter (SLC31A1, best known as copper transporter 1, CTR1), responsible for a large fraction of CDDP influx (15). However, several evidences suggest that CTR1 employs different mechanisms for copper and cisplatin transport, indeed the rapid CTR1 conformational changes observed during copper translocation is not coupled with cisplatin uptake (16).

As the landscape of treatment options for solid tumors, and particularly for head and neck squamous cell carcinoma (HNSCC), continues to evolve, attention must be given to the introduction of novel therapies to be combined with existing treatment options to improve the therapeutic index and to overcome the mechanisms of resistance to conventional drug regimens such as the fluoropyrimidines/platinum combination.

Histone deacetylase inhibitors (HDACi) are an emerging group of agents that by targeting histone and nonhistone proteins deacetylation, can modulate gene expression and cellular functions, regulating different altered pathways in cancer, such as apoptosis, cell cycle, and DNA repair (7, 17–19). Although HDACi demonstrated preclinical efficacy as monotherapy or in combination with other anticancer drugs, their clinical efficacy in monotherapy in solid tumors is very poor (20). Vorinostat (suberoylanilide hydroxamic acid), has shown significant preclinical and clinical antitumor activity in both hematologic malignancies and solid tumors (20) and represents the first HDACi approved by the FDA for the treatment of recurrent cutaneous T-cell lymphoma with a good tolerance profile (21). We have previously demonstrated that HDACi, including vorinostat, synergize with fluoropyrimidines, both in vitro and in vivo, in colorectal and breast cancer cells through downregulation of TS and upregulation of thymidine phosphorylase (TP), a critical enzyme mediating 5FU metabolic activation and metabolic transformation of capecitabine to 5FU (22).

Moreover, we have also demonstrated, in HNSCC preclinical models, that the vorinostat-induced antiproliferative effect was paralleled by downregulation of EGFR protein expression and activity (18).

In this study, we demonstrated that the HDACi vorinostat induces, in both in vitro and in vivo models, synergistic antitumor effect in combination with 5FU/CDDP in squamous cancer cell lines by increasing DNA damage and proapoptotic effects. Mechanistically, vorinostat inhibited chemotherapy-induced phosphorylation and nuclear translocation of EGFR, increased the platinated DNA level, and enhanced the platinum intracellular concentration by upregulating the CDDP transporter CTR1.

Human squamous cancer cell lines KB, Hep2, and FaDu were purchased from the ATCC; Cal27 cell line was kindly provided by Dr. J.L. Fishel (Centre A Lacassagne, Nice, France). The cells have been authenticated with short tandem repeat profile generated by LGC Standards.

Green fluorescent protein⁺/luciferase⁺(GFP⁺/luc⁺) Cal27 cell line was obtained by lentiviral infection. Luciferase expression was confirmed by using Luminometer (Applied Biosystems), and GFP coexpression was assessed by fluorescence microscopy.

Cisplatin [or Cis-diammine platinum (II) dichloride] was from Sigma-Aldrich, 5FU from Teva Pharmaceutical Industries Ltd., and vorinostat from Merck and Co., Inc.

Cell viability was evaluated by spectrophotometric dye incorporation assay using sulforhodamine B (Sigma-Aldrich), 72 or 96 hours after treatment, as described previously (6).

Drug combination studies were performed as previously reported by evaluating the combination index (CI) and the dose reduction index (DRI) using CalcuSyn Software (Biosoft; refs. 7, 19). A CI < 0.9, CI = 0.9–1.2, and CI > 1.2 indicates a synergistic, an additive, or an antagonistic effect, respectively (6, 18). DRI₅₀ value represents the order of magnitude (fold) of dose reduction, obtained for the IC₅₀ in combination treatment compared with single-drug treatment.

Immunoblotting with the indicated antibodies was performed as described elsewhere (23). Densitometric analysis was performed using NIH ImageJ software.

RNA was isolated by TRizol (Invitrogen) reagent as described previously (24). Real-Time PCR by ABI Prism 7900 HT Sequence Detection System (Applied Biosystems) was performed using specific TaqMan probes. The TP, TS, CyclinD1, and CTR1 relative mRNA expression levels were calculated using the 2^−ΔΔCt method and were normalized to that of the endogenous control gene β-actin.

Apoptosis was measured by flow cytometry using the Annexin V-FITC, as described previously (25) and by Western blot detection of PARP proteolytic cleavage. Cell-cycle analysis was performed at the indicated timepoints in Cal27 cells treated with vorinostat and/or 5FU or CDDP, alone or in combination, as reported previously (23).

Phosphoprotein detection was performed using the bead-based Bio-Plex Phosphoprotein Assay Kit (Luminex xMAP Technology, Bio-Rad) according to the manufacturer's instructions. Cells were treated with drugs at IC₅₀^72h concentrations for 24 hours. Proteins lysates were prepared using cell lysis buffer containing phenylmethylsulfonylfluoride (Bio-Rad). Data acquisition and analysis were performed using Bio-Plex manager software. Phosphoprotein fluorescence intensities were normalized to total protein expression intensities.

DNA damage was measured by the immunofluorescence assay using a primary antibody specific for γH2AX (Millipore Corporation), as described previously (9). The nuclei were stained using DAPI (4′,6-diamidin-2-phenylindole) or TO-PRO-3 (Life Technologies). The images were obtained using a confocal microscope Zeiss LMS510 (Zeiss).

Proteins extraction was carried out as indicated previously, whereas DNA was extracted using phenol–chloroform method. CDDP adducts were quantified according to a method described by O'Neill and colleagues with some modifications (26). Samples were sonicated, resuspended in 1% HNO, and incubated at 37° C for 16 hours. The samples were analyzed by atomic absorption spectroscopy using a Varian Spectr A200 Atomic Absorption spectrophotometer with a graphite furnace equipped with a platinum hallow cathode and a deuterium lamp for background correction. The temperature program used was as follows: for drying, 85°C (5 seconds), 95°C (40 seconds), and 120°C (10 seconds); for charring, 1,000°C (5 seconds) and 1,000°C (3 seconds); for the atomization, 2,700°C (1 second, read signal) −2,700°C (2 seconds, read signal), and 2,300°C (2 seconds, cleaning). The absorbance was determined at 265.9 nm. Analyses were performed in triplicate. Serial aqueous dilutions from platinum stock solution (H₂PtCl₃H₂O) were used to generate a standard curve and platinum amount was expressed as μg of Pt per g of DNA and μg of Pt per g of protein.

Genomics analysis and visualization platform database (http://r2.amc.nl) were performed applying a signature of genes related to transcriptional activity of nuclear EGFR in TCGA RNA-sequencing data from the TCGA_HNSCC cohort of patients.

Female BALB/c athymic (nu⁺/nu⁺) nude mice (Harlan) were acclimatized at the Animal Care Facility of Biogem S.C.A.R.L. (Ariano Irpino, Italy) in accordance with the institutional guidelines of the Italian Ministry of Health Animal Care and Use Committee.

GFP/luc-transfected Cal27 cells (4 × 10⁴) resuspended in 20 μL of PBS were injected directly into the anterior tongue. Twelve days after injection, mice (n = 10) were randomized into four treatment groups based on bioluminescence signal measured after the first IVIS Imaging (PerkinElmer): (i) control; (ii) vorinostat (100 mg/kg) orally with gavage administration in 100 μL of 10% DMSO/45% PEG/45% physiologic solution; (iii) 5FU (20 mg/kg)/CDDP (1 mg/kg) i.p. in 100 μL of physiologic solution; (iv) vorinostat/5FU/CDDP combination for 5 days a week for 2 weeks. All mice received drugs vehicles. The tumor volumes were monitored by IVIS and the signal intensity (photons/second) was quantified using the Living Image Software 4.1 (PerkinElmer). For the calculation of CI by CalcuSyn, the values of cell kill (CK) for a fixed tumor volume were considered [determined by the log CK (LCK)]. LCK was determined by LCK = (T − C)/(3.3 − T_d), where T and C are the same values as described above; T_d represents the mean control group doubling time required to reach a fixed tumor volume and is expressed in days (8). Tumour growth delay (TGD), determined as %TGD = [(T – C)/C] × 100, where T and C are the mean times expressed in days for the treated or control groups, respectively, to reach a defined tumor volume.

GFP/luc-transfected Cal27 cells (8 × 10⁶) resuspended in 200 μL of PBS were injected subcutaneously into the right flank area of 40 mice. When established tumors became palpable, mice (n = 10) were randomized into four experimental groups: (i) control; (ii) vorinostat (80 mg/kg) orally with gavage administration in 100 μL of 10% DMSO/45% PEG/45% physiologic solution; (iii) 5FU (20 mg/kg)/CDDP (1 mg/kg) i.p. in 100 μL of physiologic solution; and (iv) vorinostat/5FU/CDDP combination for 4 days a week for 2 weeks. All mice received other drugs vehicles. Tumor volume (TV; mm³) was calculated by the formula TV = [length (mm) × width (mm)²]/2, where the width and the length are the shortest and the longest diameters, respectively, as measured by a caliper. One week after the end of treatment, mice were sacrificed by cervical dislocation. The percent change in the experimental groups was compared with that of the vehicle control group using the following equation: [(overall percent change experimental – overall percent change vehicle)/overall percent change vehicle × 100; ref. 8].

The results of the in vitro cell viability assay were expressed as means for at least three independent experiments, which were conducted in quadruplicate (±SD). The results of the apoptotic analysis were expressed as means for at least three independent experiments (±SD), and the statistical significance was determined by one-way ANOVA and Tukey multiple comparisons test. Representative results from a single experiment of quantitative Real-Time PCR, Western blot, IHC, and CDDP levels are presented; additional experiments yielded similar results. Statistical significance of the differences in tumor growth in vivo was determined by the one-way ANOVA. A P < 0.05 was considered to be statistically significant and the specific values are reported or indicated in legends to figures as *, P < 0.05; **, P < 0.01; ***, P < 0.001. All statistical evaluations were performed using GraphPad Prism 7.0 software.

First, we evaluated the antiproliferative effects of either vorinostat or 5FU or CDDP, as single agents in squamous cancer cell lines (KB, Hep2, and Cal27). The p53-mutant (p53-mut) Cal27 cells expressing high EGFR protein levels have been described as intrinsically resistant to CDDP and indeed were the most resistant to CDDP with 72 hours IC₅₀ value 15-fold higher than the other two cell lines. The p-53 wild-type (p53-wt) Hep2 cells, characterized by high TS and only a faint level of TP protein expression, were the less sensitive to 5FU (Supplementary Fig. S1).

We next investigated vorinostat antitumor effect in combination with 5FU and/or CDDP. Simultaneous equipotent increasing doses of 96 hours exposure to vorinostat and either 5FU or CDDP demonstrated strongly synergistic interactions as shown by CIs values lower than 0.8, calculated at 50% (CI₅₀) or 75% (CI₇₅) of cell lethality, in all cell lines (Fig. 1A), confirming the antiproliferative synergistic interaction between HDACi and 5FU or CDDP reported by us and other groups (6, 8, 27–30). More importantly, we demonstrated, for the first time, synergism in the triple-combination setting with the lowest CIs observed in Cal27 CDDP–resistant cells. Interestingly, the evaluation of DRI values, which represent the order of magnitude (fold) of dose reduction, obtained for the IC₅₀ (DRI₅₀) in combination treatment compared with single-drug treatment, showed the best fold reductions in the triple-combination (highlighted in Fig. 1A in gray) with values of 5.71–11.5 for 5FU and 9.8–14.1 for CDDP in all cell lines.

The synergistic antiproliferative effect induced by the triple vorinostat/5FU/CDDP combination correlates with a significant induction of apoptosis evaluated by both Annexin V staining (Fig. 1B) and PARP cleavage (Fig. 1C), as well as strong cell-cycle perturbation with a prominent S-phase arrest (Supplementary Fig. S2C). In details, Cal27 treated for 48 hours with low doses (IC30^72h) of each agent, alone or in combination, were confirmed to be resistant to CDDP as a single agent and only a slight nonstatistically significant increase in apoptosis was observed in vorinostat- or 5FU-treated cells or in double-combination treatments. In contrast, a significant induction of apoptosis (between 1.5- and 1.8-fold increase) was observed in the triple-combination compared with 5FU–CDDP and vorinostat–5FU or vorinostat–CDDP (Fig. 1B). Synergistic induction of apoptosis by vorinostat/5FU/CDDP combination was confirmed in FaDu cells, another p53-mut squamous cancer cell line (Supplementary Fig. S2A) and in p53-wt Hep2 cells (Supplementary Fig. S2B).

Moreover, we measured DNA damage induction of different treatments, by evaluating H2AX phosphorylation (γH2AX; Fig. 2A) or γH2AX nuclear foci formation, in both Cal27 (Fig. 2B; Supplementary Fig. S3) and Hep2 (Fig. 2C) cells. Notably, we demonstrated that simultaneous exposure to equitoxic doses (IC₅₀ at 72 hours) of vorinostat, 5FU, and CDDP resulted in a significant increase in DNA damage compared with single agents or double-combinations, in both cell lines, confirming the synergistic antitumor interactions demonstrated previously.

Because p53 plays a critical role in cell-cycle arrest, DNA damage response, and chemotherapeutics-induced apoptosis, we evaluated in p53-wt Hep2 and p53-mut Cal27 cells, p53 phosphorylation at ser15, ser37, and ser46, by an ELISA-based multiplex immunoassay. Ser15 and ser37 phosphorylations are induced by DNA damage and prevent p53–mdm2 interaction, whereas ser46 phosphorylation regulates transactivation of proapoptotic genes (31). As shown in Fig. 2D, triple-combination increased ser15, ser37, and ser46 phosphorylation in Hep2 compared with control, vorinostat alone, or 5FU/CDDP, in agreement with the synergistic antiproliferative effects and DNA damage induction reported above. These data correlate with the upregulation of p53-wt protein expression induced by vorinostat in Hep2 (Supplementary Fig. S4A), as previously reported by our group and others (6). Surprisingly, in Cal27, although no modification of ser15 and ser46 was observed upon treatments, we demonstrated increased p53 ser37 phosphorylation in triple-combination versus control or other treatments, suggesting, at least in part, a role of p53-mut protein (Fig. 2D). Notably, vorinostat induced destabilization of p53-mut protein via inhibition of the HDAC6-HSP90 chaperone pathway in several cancer models (32). Indeed, we showed p53 protein downregulation upon vorinostat treatment in p53-mut Cal27 and FaDu (Supplementary Fig. S4A).

EGFR overexpression, activity, and nuclear localization are associated with chemotherapy resistance (13, 33). Here, we demonstrated that 24 hours vorinostat treatment downregulated basal and 5FU/CDDP-induced EGFR expression and activity, as well as AKT phosphorylation in Cal27, as shown by Western blot analysis and ELISA-based phosphoprotein assay (Fig. 3A), confirming our previous observations (18). Most importantly, we demonstrated, for the first time, that vorinostat reverted 5FU/CDDP-induced EGFR nuclear translocation (Fig. 3B). Acetyl-alpha-tubulin and acetyl-histone-H3 staining confirmed vorinostat-induced HDAC inhibition, whereas PARP cleavage increase in the nucleus confirmed the proapoptotic activity of vorinostat/5FU/CDDP combination.

We have previously demonstrated that EGFR downregulation in cancer cells by HDACi, including vorinostat, was due to increased protein degradation, prevalently by lysosomes (18, 25). Thus, we evaluated vorinostat effect on 5FU/CDDP-induced EGFR nuclear translocation in the absence or presence of the lysosome-neutralizing agent, ammonium chloride (NH₄Cl) or proteasome inhibitor, bortezomib. As shown in Fig. 3C, concomitant treatment with NH₄Cl, but not with bortezomib, completely reverted vorinostat-mediated inhibition of 5FU/CDDP-induced EGFR nuclear translocation, confirming our previous findings and suggesting that the inhibition of EGFR nuclear translocation observed in Cal27 is due to vorinostat-increased EGFR protein degradation (Fig. 3C).

Furthermore, to functionally evaluate the modulation of nuclear EGFR noncanonical role, we generated a signature of transcription factor and chromatin regulator genes that are either physically interactors of EGFR or are modulated by EGFR, in the nucleus (Fig. 3D; ref. 34). Then, we used head and neck Cancer Genome Atlas (TCGA) database (patient's characteristics are reported in Supplementary Fig. S5B) to validate the relationship between the noncanonical nuclear EGFR signaling and patients' outcome. Interestingly, only genes directly modulated by intrinsic transactivation activity at EGFR carboxyl terminus were highly enriched in the poor versus good outcome patients (Fig. 3E; Supplementary Fig. S5A). However, while cyclin D1 (CCND1), AURKA, ABCG2, and MYC, showed a clear prognostic role in HNSCC, TCGA dataset being overexpressed in low-survival patients, TS (TYMS), PTGS2, and MYBL2 apparently were not correlated with prognosis. Moreover, we demonstrated that vorinostat reverted 5FU/CDDP-mediated induction of TS and CCND1, two genes from the signature (Fig. 3F). Downregulation of basal, as well as 5FU- or 5FU/CDDP-induced mRNA and protein TS expression was also observed in KB and Hep2, both expressing high levels of this enzyme (Supplementary Fig. S6A and S6B). In both cell lines, we confirmed an early increment of TP transcript and protein expression induced by vorinostat alone or in the double- and triple-combination schedules, as compared with very low basal levels observed in untreated cells. Only a slight increase in TP protein expression was observed in Cal27 expressing high basal levels of the protein. Notably, vorinostat treatment with the same dosage failed to significantly modify both TP and TS protein levels in nontumorigenic human cultured keratinocytes (HaCaT) cells (Supplementary Fig. S6A), as previously reported also in ex vivo–treated peripheral blood lymphocytes (7) or in normal epithelial breast cells (8).

To evaluate whether vorinostat might affect CDDP intracellular concentration, we measured both protein- and DNA-bound platinum levels in CDDP–sensitive Hep2 and CDDP–resistant Cal27 cells, treated with CDDP, 5FU/CDDP, vorinostat/CDDP, or vorinostat/5FU/CDDP.

At early timepoint (3 hours), protein-bound platinum levels were similar in both cell lines treated with CDDP alone or in combination with 5FU. Remarkably, although Cal27 were treated with a 20-fold higher dose of CDDP than Hep2, DNA-bound platinum levels were higher in the latter one, according to Cal27-intrinsic CDDP resistance.

Notably, for the first time, we demonstrated that vorinostat and CDDP concomitant treatment significantly enhanced the platinated DNA levels in both cell lines. Indeed, upon 6 hours treatment, a clear increase in platinated DNA and protein-bound platinum levels were further observed in Cal27. In contrast, no further increase in the platinated DNA or protein-bound platinum levels was observed in Hep2 CDDP–sensitive cells where the CDDP accumulation is probably a rapid event with prompt binding to DNA within 3 hours (Fig. 4A). Conversely, at later timepoints (14 and 24 hours), we demonstrated in Cal27, a consistent increase in both protein-bound and DNA-bound platinum levels with vorinostat/CDDP, as well as with vorinostat/5FU/CDDP treatment (Fig. 4B), suggesting a time-dependent effect of vorinostat-induced modulation of both CDDP uptake and DNA binding in CDDP-resistant cells.

Furthermore, we showed that 24 hours vorinostat pretreatment before CDDP enhanced protein- and DNA-bound platinum levels even at earlier timepoints (3–6 hours), compared with non-pretreated cells (Fig. 4C).

These results further suggest that vorinostat needs some time to induce a change in Cal27 cells to impact on CDDP intracellular accumulation and consequently DNA binding. To investigate the mechanism underlying the vorinostat effect, we evaluated the expression of the CTR1, the CDDP major influx transporter (11). As shown in Fig. 4D, we demonstrated that vorinostat increased the CTR1 transcript in a time-dependent manner with an increase observed already after 4 hours up to 24 hours, paralleled by 2-fold induction of CTR1 protein expression, consistent with the timing of increased protein-bound platinum. We confirmed CTR1 protein upregulation also in Hep2 cells (Supplementary Fig. S4B). Notably, the clear induction of CTR1 expression appeared particularly in the mature/active form of CTR1 (top band in the Western blot analysis; Fig. 4D; Supplementary Fig. S4B).

To confirm our results in vivo, we injected GFP/luciferase stably transfected Cal27 cells (Cal27-GFP⁺/luc⁺; Supplementary Fig. S7) in both orthotopic and heterotopic mouse models.

In the orthotopic model, Cal27-GFP⁺/luc⁺ cells were injected into the tongue of 40 mice (35), when bioluminescent tumors became detectable (7 days after implantation) the mice were randomly assigned to receive (5 days/week for 2 weeks) vorinostat (100 mg/kg), 5FU (20 mg/kg)/CDDP (1 mg/kg), the three drugs in combination, or their vehicles. Notably, the 5FU and CDDP dosages correspond to those reported in the literature for in vivo experiments, and were lower than the corresponding doses used in patients; whereas vorinostat dosage was based on our previous study (7). Approximately 28 days after implantation, mice develop feeding issues due to the tongue tumor size. The body weight showed a modest decrease, and the mice were sacrificed at day 35 (inset in Fig. 5A).

Tumor growth was monitored two times/week by photon intensity (Fig. 5B). The major tumor growth inhibition was observed in the vorinostat/5FU/CDDP combination at day 35 (P < 0.0001; Fig. 5A and B). Furthermore, the synergistic effect was confirmed by TGD and CI determination. In detail, TGD reached a peak of 92%, indicating that the rate of tumor growth in the control was approximately 1.65-fold higher than that in the triple-combination setting and CI values were always less than 0.8 indicating a strong synergism (Fig. 5C).

We confirmed the synergistic interaction of the proposed combination in a heterotopic xenograft mouse model of Cal27 cells, reducing drug doses. In detail, the mice were randomly assigned to receive (4 days/week for 2 weeks) vorinostat (80 mg/kg), 5FU (20 mg/kg)/CDDP (1 mg/kg), the three drugs in combination, or their vehicles. The maintenance of body weight (inset in Fig. 5E) and the absence of other acute or delayed toxicity signs indicated a well tolerability of this drugs combination. Notably, 25 days after cell injection, vorinostat/5FU/CDDP treatment induced a significant inhibition of tumor growth compared with that in the untreated (P = 0.005) or 5FU/CDDP (P = 0.038) treated groups (Fig. 5E). Moreover, vorinostat and the three-drug combination reduced the tumor volume by 16.48%, and 61.9%, respectively (Fig. 5F).

Finally, we demonstrated also a prominent increase of DNA damage, measured as H2AX protein phosphorylation, in the triple combination tumor samples compared with the other groups in the orthotopic in vivo model. In the majority of tumor samples from vorinostat- and triple-combination–treated mice, we showed increased PARP cleavage, in line with in vitro data showing increased apoptosis. Moreover, we confirmed in vivo CCND1 and TS protein downregulation, well correlating with reduced tumor growth, and increased CTR1 protein expression in line with the synergistic interaction mechanisms proposed from in vitro experiments (Fig. 5G). Increased acetyl-H3 expression in tumor samples from vorinostat-treated mice was as pharmacodynamics marker of HDAC inhibition (Fig. 5G). Finally, we confirmed CCND1 mRNA downregulation in orthotopic tumor samples (Supplementary Fig. S4C).

Despite significant advances in solid cancer chemotherapy strategies, drug resistance remains a major obstacle to improve cancer patient's outcome. In this study, we highlighted a new therapeutic approach, based on the HDACi, vorinostat, plus two widely used chemotherapy agents, such as 5FU and CDDP. We demonstrated in human squamous cancer cell lines, including intrinsic CDDP-resistant Cal27 cells, that simultaneous exposure to equitoxic doses of vorinostat plus 5FU/CDDP results in strong synergistic antiproliferative and proapoptotic effects related to DNA damage induction, as shown by a significant increase in γH2AX foci, a sensitive marker of DNA double-strand breaks, and p53 phosphorylation. Significantly, the synergistic antitumor interaction of the vorinostat/5FU/CDDP combination was confirmed in vivo in both orthotopic and heterotopic xenograft mouse models of Cal27 cells. The vorinostat effective doses tested in our study can be easily reached in plasma patients (19).

Although the antitumor synergistic interaction between vorinostat and either fluoropyrimidines or platinum was previously shown by us and others, here we showed, for the first time, synergism in the triple-combination approach, with a prominent dose reduction index for both 5FU and CDDP. The observation that vorinostat might enhance, even at doses below the IC₅₀ values, the 5FU/CDDP-induced antitumor effect suggests that the mechanism of the synergism might depend, at least in part, on the ability of vorinostat to modulate the sensitivity of the cells toward 5FU/CDDP, rather than on the concurrent cell killing induced by the three agents.

Indeed, we confirmed, as we and others have previously shown (6–8, 36–39), that the synergistic interaction between vorinostat and 5FU, in both the double- and triple-combination schedule, is due to the modulation of TP and TS, two critical enzymes normally involved in fluoropyrimidines mechanism of action and resistance (40), and described as determinants of CDDP resistance in non–small cell lung cancer (NSCLC) models (41). We also showed that vorinostat inhibited the 5FU/CDDP-induced EGFR-AKT survival pathway in agreement with our previous report demonstrating the downregulation of EGFR and AKT expression and activation upon vorinostat treatment in cancer cells (18).

However, most importantly, we demonstrated, for the first time, that vorinostat reverses 5FU/CDDP-induced EGFR nuclear translocation, a specific mechanism of CDDP and other antitumor approaches resistance (11, 42). The inhibition of 5FU/CDDP-induced EGFR nuclear translocation, might lead to the impairment of EGFR nuclear noncanonical activity, such as gene expression induction, DNA replication, and DNA damage repair inhibition (34), explaining the antitumor synergism observed.

We then generated a signature of EGFR nuclear noncanonical activity genes and demonstrated that two of them, TS and CCND1, were induced by 5FU/CDDP and that this induction was blocked by the concomitant treatment with vorinostat. Interestingly, Kim and colleagues showed that nuclear EGFR and HER2 bind directly to the TS promoter, inducing its transcription (43). Thus, we can speculate that the inhibition of EGFR nuclear translocation is one of the mechanisms involved in the well-described TS downregulation induced by HDACi (8). We reinforce the potential clinical significance of our findings demonstrating, by differential expression analysis from HNSCC TCGA database, that CCND1, AURKA, ABCG2, and MYC transcripts were highly enriched in the poor- compared with good-outcome patients. Notably, these genes were all directly modulated by intrinsic transactivation activity at the carboxyl terminus of EGFR. Although, TS was not significantly associated with poor outcome in TCGA dataset, other evidences from the literature suggested its prognostic and resistance predictive role in HNSCC (44–46). Finally, we provided evidences suggesting that the inhibition of EGFR nuclear translocation was exerted by vorinostat, at least in part, by increasing lysosomal-mediated EGFR protein degradation, confirming our previous observations (18, 25).

We also showed, for the first time, that concomitant or 24 hours pretreatment with vorinostat increase DNA- and protein-bound platinum levels in CDDP-treated cells. Although HDACi-induced DNA relaxation around the histone core might increase DNA accessibility to platinum, as previously shown by our group and others for other DNA-damaging agents (19, 47), here we suggest, particularly in CDDP-resistant Cal27 both in vitro and in vivo xenograft model, that vorinostat increases CDDP-uptake by upregulating CTR1 expression (48), the most important CDDP import channel (15). Notably, it was demonstrated that DNA-bound platinum is the major determinant of CDDP sensitivity (49). Moreover, CTR1 overexpression correlated with CDDP-acquired resistance in HNSCC cells (50). Furthermore, it was also reported that patients with NSCLC whose tumors showed undetectable CTR1 levels also had reduced tissue concentrations of platinum and showed poor responses to neoadjuvant chemotherapy with platinum plus taxane or other drugs (51). Significantly, a recent meta-analysis confirmed that high CTR1 levels predict prolonged survival and an enhanced response to platinum-based chemotherapy in patients with cancer, suggesting that CTR1 might be a potential target to circumvent chemotherapy resistance (52). Although we could not find any correlation in the literature between CTR1 expression and EGFR, particularly nuclear EGFR, interestingly, a recent publication demonstrated that the combined CTR1^High and TS^Low tumor expression status significantly increased the predictive power for the tumor response, progression-free survival, and overall survival, compared with CTR1 or TS alone, in patients with advanced NSCLC treated with the oral fluoropyrimidine S-1 and carboplatin (53).

The tumor response to chemotherapy is a complex process involving many factors; thus, we cannot exclude that other mechanisms can be involved in the strong antitumor synergism we have observed between vorinostat and 5FU/CDDP. Indeed, due to their pleiotropic effect, it has been well described that vorinostat, as well as other HDACi, can abrogate the DNA repair process and inhibit survival mechanisms, thus increasing the cytotoxic effects of DNA-damaging agents and resulting in cell death.

Indeed a number of HDACi have been studied in combination with platinum-based chemotherapy demonstrating synergistic interaction based on DNA-damage induction and DNA repair failure in combination treatment (54–57).

However, to our knowledge, this study is the first to show in squamous cell cancer models, and with novel potential mechanisms (Supplementary Fig. S8), a synergistic antitumor effect both in vitro and in vivo between a HDACi such as vorinostat and a combined widely used chemotherapeutic regimen represented by a fluoropyrimidine and a platinum derivative. Moreover, by highlighting, particularly in intrinsic CDDP-resistant cells, the critical role of both nuclear EGFR and CTR1 modulation induced by vorinostat, we also suggested potential predictive biomarkers to select patients who can benefit from the proposed combined treatment approach.

Several trials investigated the antitumor effect of a combinatory treatment involving HDACi and platinum derivatives. A randomized, phase II, placebo-controlled study on 94 patients with previously untreated stage III B or IV NSCLC demonstrated that vorinostat enhances the efficacy of platinum compounds but it also led to increased toxicity (58). More recently, a phase I trial showed the safety and clinical activity of belinostat 48 hours infusion plus cisplatin and etoposide in patients with advanced solid tumors, suggesting the global lysine acetylation measurement as marker of sufficient HDACi exposure (59).

Furthermore, the triple-combination of vorinostat, the fluoropyrimidine capecitabine, and CDDP, has been shown to be feasible in patients with metastatic or unresectable gastric cancer, but, did not meet the primary endpoint in a phase II study, with more adverse events compared with historical data of fluoropyrimidine-platinum doublet (60). However, the increase in toxicities with the addition of vorinostat was inevitable because the chemotherapy backbone had only a mild dose modification from the standard treatment in the above reported studies. Our data clearly indicated synergistic interaction even at low doses, particularly the significant DRI reported for both 5FU and CDDP, suggesting that, by reducing the doses of chemotherapeutics, we can maintain the same anticancer effects. Thus, further clinical efforts are needed to refine treatment schedules, reduce toxicity, and select the most appropriate patient population for the combination approach to reach a successful clinical translation. In details, a marker-driven categorization of patients is needed to identify chemo-resistant tumors with driver HDACi targetable chemo-escape pathways. Low basal level of the CTR1 channel could indicate a platinum uptake-deficient–resistant phenotype most likely to respond to drug such as HDACis, which can modulate its expression. Similarly, patients with TS-overexpressing tumors could benefit from a combination approach with HDACis and fluoropyrimidines. Finally, we indicated vorinostat-dependent reversion of the chemotherapy-induced EGFR nuclear translocation as possible mechanism for synergistic interaction, on this bases the identification of patients with a basal aberrant nuclear-EGFR signalling could highlight a DNA repair–proficient subset of tumors, which can be targeted by simultaneous administration of HDACis and DNA-damaging agent.

Overall, our findings add new insight into the mechanism of the synergistic interaction between HDACi- and CDDP-based chemotherapy, providing a rationale to clinically explore this combination to overcome dose-limiting toxicity and chemotherapy resistance. The selection of patients for this regimen, based on tumor CTR1 and/or TS expression and/or EGFR localization and/or nuclear EGFR–regulated transcripts, should also be explored.

No potential conflicts of interest were disclosed.

Conception and design: G. Piro, M.S. Roca, A. Budillon, E. Di Gennaro

Development of methodology: G. Piro, M.S. Roca, M.G. Volpe, A. Budillon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Piro, M.S. Roca, F. Bruzzese, F. Iannelli, M.G. Volpe

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Piro, M.S. Roca, F. Iannelli, A. Leone, A. Budillon, E. Di Gennaro

Writing, review, and/or revision of the manuscript: G. Piro, M.S. Roca, A. Leone, A. Budillon, E. Di Gennaro

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Di Gennaro

Study supervision: A. Budillon, E. Di Gennaro

This study was partially supported by the Italian Ministry of Health (Ricerca Corrente funds Prog. 3/6) to A. Budillon. Fondazione Italiana per la Ricerca sul Cancro (FIRC)-AIRC supported with a triennial Fellowship to M.S. Roca (21113) and F. Iannelli (22648).

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.