Carfilzomib and oprozomib synergize with histone deacetylase inhibitors in head and neck squamous cell carcinoma models of acquired resistance to proteasome inhibitors

Abstract

Acquired resistance to proteasome inhibitors represents a considerable impediment to their effective clinical application. Carfilzomib and its orally bioavailable structural analog oprozomib are second-generation, highly-selective, proteasome inhibitors. However, the mechanisms of acquired resistance to carfilzomib and oprozomib are incompletely understood, and effective strategies for overcoming this resistance are needed. Here, we developed models of acquired resistance to carfilzomib in two head and neck squamous cell carcinoma cell lines, UMSCC-1 and Cal33, through gradual exposure to increasing drug concentrations. The resistant lines R-UMSCC-1 and R-Cal33 demonstrated 205- and 64-fold resistance, respectively, relative to the parental lines. Similarly, a high level of cross-resistance to oprozomib, as well as paclitaxel, was observed, whereas only moderate resistance to bortezomib (8- to 29-fold), and low level resistance to cisplatin (1.5- to 5-fold) was seen. Synergistic induction of apoptosis signaling and cell death, and inhibition of colony formation followed co-treatment of acquired resistance models with carfilzomib and the histone deacetylase inhibitor (HDACi) vorinostat. Synergism was also seen with other combinations, including oprozomib plus vorinostat, or carfilzomib plus the HDACi entinostat. Synergism was accompanied by upregulation of proapoptotic Bik, and suppression of Bik attenuated the synergy. The acquired resistance models also exhibited elevated levels of MDR-1/P-gp. Inhibition of MDR-1/P-gp with reversin 121 partially overcame carfilzomib resistance in R-UMSCC-1 and R-Cal33 cells. Collectively, these studies indicate that combining carfilzomib or oprozomib with HDAC or MDR-1/P-gp inhibitors may be a useful strategy for overcoming acquired resistance to these proteasome inhibitors.

Keywords: :

• carfilzomib
• oprozomib
• bortezomib
• vorinostat
• entinostat
• acquired resistance
• MDR-1/P-gp

Introduction

The successful use of bortezomib in the treatment of multiple myeloma and mantle cell lymphoma has spurred interest in applying proteasome inhibitors to other human cancers.^1-5 However, only modest success has been reported in preclinical and early-stage clinical testing of bortezomib in solid tumors, including head and neck squamous cell carcinoma (HNSCC).^6-17 The efficacy of bortezomib in both solid tumor and hematologic malignancies is limited by several factors. Bortezomib exhibits off-target inhibitory activity against several serine proteases, including HtrA2/Omi, cathepsins A and G, chymase, and dipeptidyl peptidase II.¹⁸ These off-target effects likely contribute to the considerable adverse toxicities, particularly peripheral neuropathy, associated with bortezomib treatment.^19-22 Since bortezomib is a reversible inhibitor of the chymotrypsin-like (CT-L) activity of the proteasome,¹ frequent dosing is required to achieve sustained proteasome inhibition, further exacerbating the cytotoxic side effects of the drug. Lastly, many tumors exhibit intrinsic, or inherent, resistance to bortezomib. Moreover, acquired resistance to bortezomib commonly develops, leading once-responsive patients to become refractory to treatment.^3,4,19,23-25

The limitations associated with the use of bortezomib have stimulated the identification and development of second-generation proteasome inhibitors with greater specificities and reduced off-target effects. Carfilzomib is a well-tolerated, irreversible proteasome inhibitor with a high degree of specificity for the CT-L activity of the proteasome.^26-28 The greater specificity, and irreversible nature, of carfilzomib is associated with low rates of peripheral neuropathy in patients.²⁹ In addition, carfilzomib can promote apoptosis in multiple myeloma cells that are resistant to bortezomib,²⁶ and a phase II study indicates that carfilzomib is active in multiple myeloma patients refractory to bortezomib.^30,31 Based on these findings, carfilzomib has been approved for use in multiple myeloma patients who have previously been treated with bortezomib. Carfilzomib is currently undergoing extensive clinical testing in multiple myeloma, as well as evaluation in mantle cell lymphoma, T-cell lymphoma, large B-cell lymphoma, Waldenstrom macroglobulinemia, acute myeloid leukemia, acute lymphoblastic leukemia, small cell lung cancer, renal cell carcinoma, and advanced prostate cancer (http://www.clinicaltrials.gov). Oprozomib is a structurally-related analog of carfilzomib that has the added benefit of being orally bioavailable.³² Currently, oprozomib is being evaluated in clinical trials of multiple myeloma and recurrent solid tumors (http://clinicaltrials.gov). Like carfilzomib, oprozomib is also active against bortezomib-resistant multiple myeloma cells.³³ Thus, carfilzomib and oprozomib offer several unique advantages, including the ability to overcome acquired bortezomib resistance. Both carfilzomib and oprozomib have also been shown to inhibit the growth of HNSCC cells in vitro and HNSCC xenograft tumors in vivo.³⁴ However, it is likely that carfilzomib- and oprozomib-based therapies will also be found to be limited by intrinsic and acquired resistance.

Efforts to combat intrinsic resistance to proteasome inhibitors have sought to identify co-treatment strategies that lead to synergistic or enhanced killing of cancer cells. Several reports have shown that proteasome inhibitors and HDAC inhibitors synergize to overcome intrinsic resistance,^35-39 though considerably less is known about this potential synergy in models of acquired resistance to proteasome inhibitors.⁴⁰ Synergy between bortezomib and STAT3 inhibitors has also been reported.⁴¹ Bortezomib, as well as carfilzomib and oprozomib, also induce expression of antiapoptotic Mcl-1, and the activities of these agents can be enhanced by suppression of Mcl-1 expression or inhibition of Mcl-1 using obatoclax (GX15-070).^8,34,42 All three proteasome inhibitors also induce autophagy, and inhibition of autophagy with chloroquine enhances their cell-killing activities.^34,43 These studies demonstrate that intrinsic resistance to proteasome inhibitors can be reduced by utilizing unique combinations of agents.

While models of acquired resistance to bortezomib have been developed and studied, considerably less is known about acquired resistance to carfilzomib or oprozomib. Acquired resistance to bortezomib has been linked to altered expression of the β5 (PSMB5) proteasome subunit, mutation of the β5 subunit, and overexpression of the multidrug efflux protein MDR-1/P-gp.⁴⁴ Recent evidence indicates that THP1 myeloid cells harboring β5 mutations, as well as CEM/VLB lymphoid cells overexpressing MDR-1/P-gp, also exhibit resistance to carfilzomib and oprozomib.⁴⁵ In addition short peptide inhibitors of MDR-1/P-gp have been shown to reverse carfilzomib resistance in lung and colon adenocarcinoma cell lines.⁴⁶

Understanding the mechanisms that contribute to acquired resistance to carfilzomib and oprozomib, and development of strategies for overcoming this resistance, will be important for optimizing the clinical use of these promising proteasome inhibitors. To address these issues, we developed models of acquired resistance to carfilzomib in the solid tumor malignancy HNSCC. We find that the carfilzomib acquired resistance models demonstrate a high level of cross-resistance to oprozomib, but only moderate resistance to bortezomib. This is significant, as it indicates that, just as carfilzomib/oprozomib are useful therapeutics for bortezomib-resistant disease, bortezomib treatment may be useful for treating tumor cells and patients that have become refractory to carfilzomib and oprozomib. We further show that the HNSCC-acquired resistance models have elevated levels of MDR-1/P-gp, which contributes to the resistance phenotype. Importantly, we also show that carfilzomib and oprozomib display synergy with HDAC inhibitors in the acquired resistance models, providing a further strategy for overcoming acquired resistance to these proteasome inhibitors.

Results

HNSCC models of acquired carfilzomib resistance

Models of acquired resistance to proteasome inhibitors have not been reported in HNSCC. Moreover, acquired resistance to the next generation proteasome inhibitors carfilzomib and oprozomib has been reported only for hematopoietic cells, not solid tumor malignancies. We have previously reported that HNSCC cell lines undergo apoptotic cell death following 48 h treatment with low nanomolar concentrations of carfilzomib and oprozomib.³⁴ To generate models of acquired carfilzomib resistance, we exposed two HNSCC cell lines, UMSCC-1 and Cal33, to gradually increasing concentrations of carfilzomib. After a period of 12 mo, the resistant cell lines grew with normal kinetics in medium containing 1 μM carfilzomib. These models of acquired carfilzomib resistance were designated R-UMSCC-1 and R-Cal33, in contrast to the sensitive parental lines designated P-UMSCC-1 and P-Cal33.

To quantitatively assess resistance, parental and resistant lines were treated for 48 h with varying concentrations of carfilzomib, or other agents, and IC₅₀ values were determined (Table 1; Fig. S1A and B). While parental P-UMSCC-1 and P-Cal33 cells demonstrated carfilzomib IC₅₀ values of 11.2 nM and 17.3 nM, respectively, R-UMSCC-1 and R-Cal33 exhibited carfilzomib IC₅₀ values of 2294 nM and 1112 nM, respectively. Thus, relative to their parental counterparts, R-UMSCC-1 cells were 205-fold more carfilzomib-resistant, and R-Cal33 cells were 64-fold more resistant. Further evaluation determined that the carfilzomib-resistant lines were also highly resistant to oprozomib (Table 1; Fig. S1A and B). In addition, the R-UMSCC-1 and R-Cal33 cell lines exhibited cross-resistance to the structurally distinct proteasome inhibitor bortezomib. However, the resistance to bortezomib was considerably less than that observed for carfilzomib, with R-UMSCC-1 and R-Cal33 exhibiting only 29-fold and 8-fold bortezomib resistance, respectively (Table 1; Fig. S1A and B). The resistance of R-UMSCC-1 and R-Cal33 to carfilzomib-, oprozomib-, and bortezomib-induced apoptosis was confirmed via flow cytometric analysis of Annexin V-stained cells (Fig. S2). Furthermore, in contrast to the parental cell lines, treatment of R-UMSCC-1 and R-Cal33 cells with the three proteasome inhibitors did not result in caspase-3 activation or cleavage of poly(ADP-ribose) polymerase (PARP; Fig. S3).

Table 1. IC₅₀ values of carfilzomib, oprozomib, bortezomib, cisplatin, paclitaxel, vorinostat, and entinostat against parental or CFZ-resistant UMSCC-1 and CAL33 cell lines

	P-UMSCC-1	R-UMSCC-1	P-Cal33	R-Cal33
CFZ (nM)	11.2	2294	17.3	1112
OPZ (nM)	88.2	4041	59.3	1872
BTZ (nM)	8.3	243.3	5.1	40.2
CP (μM)	2.1	10.4	2.6	4
PTX (nM)	33.5	> 5000	5.2	3837
VOR (μM)	4.9	24.2	5.6	12.1
ENT (μM)	19.4	86.7	20.4	75.4

CFZ, carfilzomib; OPZ, oprozomib; BTZ, bortezomib; CP, cisplatin; PTX, paclitaxel; VOR, vorinostat; ENT, entinostat. Cells were treated for 48 h with 0.1% DMSO or the indicated drug. Following treatment, MTT assays were performed. Data points represent the average of values obtained in 3 independent experiments.

We also examined whether the R-UMSCC-1 and R-Cal33 lines exhibited cross-resistance to conventional chemotherapy drugs and other targeting agents (Table 1; Fig. S1A and B). The chemotherapies most commonly used in the treatment of HNSCC are platinum-based or taxane compounds. Only modest 5-fold and 1.5-fold resistance to cisplatin was seen in R-UMSCC-1 and R-Cal33, respectively, when compared with the parental cell lines. By contrast, both cell lines had a high degree of resistance to paclitaxel (at least 150-fold). When challenged with the HDAC inhibitors vorinostat or entinostat, R-UMSCC-1 and R-Cal33 demonstrated low levels of resistance ranging from 1.5- to 5.0-fold.

Carfilzomib synergizes with vorinostat to promote apoptosis in proteasome inhibitor-resistant HNSCC cells

Proteasome inhibitors, primarily bortezomib, and HDAC inhibitors have been shown to synergize against both hematologic and solid tumor malignancies, in models that are not acquired resistance models. Therefore, we sought to determine whether vorinostat, an inhibitor of class I and II HDACs, might synergize with carfilzomib against R-UMSCC-1 and R-Cal33, acting to overcome the acquired resistance to carfilzomib in these cell lines. As shown in Figure 1A and B, the combination of carfilzomib and vorinostat potently synergized to kill both R-UMSCC-1 cells and R-Cal33, as demonstrated by combination index (CI)⁴⁷ values substantially lower than 1.0 at multiple different doses. However, despite the potency of this synergism, the resistance of these cells was not fully overcome, as efficacious killing by the combination required high nanomolar concentrations of carfilzomib. Nonetheless, the synergistic activity of the carfilzomib/vorinostat combination was partially effective at overcoming acquired resistance to carfilzomib.

Figure 1. Carfilzomib and the HDAC inhibitor vorinostat act synergistically to induce cell death in proteasome inhibitor-resistant HNSCC cell lines. Triplicate wells of R-UMSCC-1 (A) and R-Cal33 (B) cells were treated for 48 h with carfilzomib (CFZ) or vorinostat (VOR) alone, or with fixed ratios of the CFZ/VOR combination, followed by performance of MTT assays. Combination indexes (CI) were calculated according to the method of Chou and Talalay.⁴⁷ Error bars represent standard deviations. Representative data from 3 independent experiments is shown.

To confirm the synergism of carfilzomib and vorinostat against our HNSCC models of acquired resistance, flow cytometric analysis of Annexin V/PI staining was performed (Fig. 2A and B). Treatment of the parental cell lines, P-UMSCC-1 and P-Cal33, with either carfilzomib alone or the carfilzomib/vorinostat combination resulted in Annexin V-positive staining in ≥88% of cells. By contrast, treatment of R-UMSCC-1 and R-Cal33 with carfilzomib alone resulted in 20.6% and 7.5% Annexin V-positive cells, respectively, while treatment with vorinostat alone resulted in less than 21% Annexin V-positivity. The combination of carfilzomib and vorinostat, however, resulted in greater than 80% Annexin V-positivity in both of the resistant cell lines, confirming the synergistic activity of the combination in the resistance models. In addition, while carfilzomib alone and vorinostat alone failed to effectively promote caspase-3 activation and PARP cleavage in the resistant cell lines, treatment with the combination led to marked activation of caspase-3 and cleavage of PARP (Fig. 3). Further evidence of carfilzomib/vorinostat synergism in the acquired resistance models was obtained in clonogenic survival assays (Fig. 4A and B). Treatment of R-UMSCC-1 and R-Cal33 with carfilzomib plus vorinostat led to synergistic inhibition of colony formation at two different doses (constant ratio of drugs).

Figure 2. Carfilzomib and vorinostat synergize to induce apoptosis in proteasome inhibitor-resistant HNSCC cells. P-UMSCC-1 and R-UMSCC-1 cells (A) and P-Cal33 and R-Cal33 cells (B) were treated for 48 h with single concentrations of CFZ or VOR, or CFZ in combination with VOR, followed by flow cytometric analysis of Annexin V/PI staining. Numbers indicate the percentage of Annexin V-positive cells. Similar results were seen in 3 independent experiments.

Figure 3. Enhanced induction of apoptosis signaling by the carfilzomib/vorinostat combination in proteasome inhibitor-resistant cells. R-UMSCC-1 and R-Cal33 cells were treated for 24 h with 0.1% DMSO, CFZ alone (1.8 μM for R-UMSCC-1; 0.8 μM for R-Cal33), VOR alone (6 μM for R-UMSCC-1; 4 μM for R-Cal33), or the combination of CFZ plus VOR. Whole cell lysates were subjected to immunoblotting for PARP and cleaved PARP (c-PARP), active caspase-3 subunits, or β-actin. Similar results were seen in 3 independent experiments, with representative blots shown.

Figure 4. Inhibition of colony formation by the combination of carfilzomib and vorinostat. R-UMSCC-1 (A) and R-Cal33 (B) cells were treated with 0.1% DMSO or the indicated concentrations of CFZ alone, VOR alone, or the combination of CFZ and VOR. After treatment for 12 h, cells were harvested, and equal numbers of cells were re-plated in 6-well plates in media containing 10% FBS, but lacking drugs. Following 10–12 d of growth, cells were stained with crystal violet and colonies counted and photographed (left). Graphs (right) depict the mean of values obtained in 3 experiments, and error bars represent standard deviations.

Diverse proteasome inhibitors and HDAC inhibitors synergize against HNSCC models of acquired proteasome inhibitor resistance

We next sought to determine whether our findings of synergism in the acquired resistance models would be extended to combinations of other proteasome inhibitors and other HDAC inhibitors. Oprozomib in combination with vorinostat also promoted synergistic killing of R-UMSCC-1 and R-Cal33 cells (Fig. 5A and B). Moreover, highly potent synergism was also observed when carfilzomib was combined with entinostat, a selective inhibitor of the class I HDACs HDAC1 and HDAC3 (Fig. 6A and B). These findings indicate that acquired proteasome inhibitor resistance can be combated with multiple different combinations of proteasome inhibitors and HDACis.

Figure 5. Oprozomib (OPZ) synergizes with vorinostat in resistant HNSCC cells. Triplicate wells of R-UMSCC-1 (A) and R-Cal33 (B) cells were treated for 48 h with OPZ or VOR alone, or with fixed ratios of the OPZ/VOR combination, followed by performance of MTT assays and determination of CI values. Error bars represent standard deviations. Representative data from three independent experiments is shown.

Figure 6. Carfilzomib synergizes with the HDAC inhibitor entinostat (ENT) to kill resistant HNSCC cells. Triplicate wells of R-UMSCC-1 (A) and R-Cal33 (B) cells were treated for 48 h with CFZ or ENT alone, or with fixed ratios of the CFZ/ENT combination, followed by determination of CI values. Error bars represent standard deviations. The experiments were performed three times with similar results.

Carfilzomib/vorinostat synergy in resistant HNSCC cells is mediated by upregulation of Bik

To determine the mechanism responsible for carfilzomib/vorinostat synergy in the acquired resistance models, we investigated the expression status of pro-apoptotic Bik protein. Bik is known to be upregulated in HNSCC cells following treatment with bortezomib, carfilzomib, or oprozomib, and plays an important role in mediating cell death by these proteasome inhibitors.^8,34 However, as depicted in Figure 7, the R-UMSCC-1 and R-Cal33 resistance models demonstrated only modest, if any, upregulation of Bik when treated with either carfilzomib alone or vorinostat alone. In contrast, co-treatment of the cells with carfilzomib and vorinostat resulted in markedly enhanced upregulation of Bik relative to that seen with either agent alone. To determine whether Bik upregulation was important for synergism by the drug combination, R-Cal33 cells were transfected with a nonspecific siRNA or with siRNA directed against Bik mRNA. Suppression of Bik expression using siRNA partially suppressed synergism by carfilzomib/vorinostat, as determined by assessment of Annexin V-positive cells (Fig. 8A and B). Thus, Bik upregulation contributes to carfilzomib/vorinostat synergism against the acquired resistance models.

Figure 7. The carfilzomib/vorinostat combination induces potent upregulation of proapoptotic Bik. R-UMSCC-1 and R-Cal33 cells were treated for 24 h with 0.1% DMSO, CFZ alone (1.8 μM for R-UMSCC-1; 0.8 μM for R-Cal33), VOR alone (6 μM for R-UMSCC-1 alone; 4 μM for R-Cal33), or the combination of CFZ plus VOR. Treated cells were subjected to immunoblotting for Bik or β-actin. Similar results were seen in 3 independent experiments.

Figure 8. Bik mediates synergism by carfilzomib/vorinostat in proteasome inhibitor-resistant cells. (A) R-Cal33 cells were seeded in 24-well plates, grown overnight, then transfected for 6 h with nonspecific siRNA or Bik siRNA. The transfected cells were then treated for 36 h with the indicated concentrations of CFZ alone, VOR alone, or the CFZ/VOR combination. Following treatment, the floating and attached cells were collected and combined, then subjected to flow cytometric analysis of Annexin V/PI staining. A representative experiment is shown, with numbers indicating the percentage of Annexin V-positive cells. The experiment was performed 3 times with similar results. (B) Data from the three independent experiments performed in (A) were combined and graphed. Columns represent the mean of values obtained in the 3 experiments, and error bars represent standard deviations.

MDR-1/P-gp expression is elevated in acquired resistance models and contributes to resistance

To investigate the mechanism(s) of resistance in the acquired resistance models, the expression levels of anti-apoptotic Bcl-2 family members was compared between parental and resistant cell lines (Fig. 9A). No apparent differences were detected in Mcl-1 expression between the parental and resistant cell lines. Bcl-2 expression was found to be upregulated in R-UMSCC-1 relative to P-UMSCC-1, suggesting that Bcl-2 may contribute to the acquired resistance in UMSCC-1 cells. On the other hand, no elevation of Bcl-2 expression was detected in R-Cal33 cells. Immunoblotting for MDR-1/P-gp revealed no detectable expression in both parental cell lines (Fig. 9A). However, marked expression of MDR-1/P-gp was observed in both of the resistance models (Fig. 9A).

Figure 9. Elevated expression of MDR-1/P-gp in resistant HNSCC cell lines contributes to the acquired resistance. (A) P-UMSCC-1, R-UMSCC-1, P-Cal33, and R-Cal33 cells were subjected to immunoblotting for MDR-1/P-gp, Mcl-1, Bcl-2, or β-actin. (B) R-UMSCC-1 and R-Cal33 cells were plated in triplicate wells, then treated for 48 h with varying concentrations of CFZ in the absence or presence of the MDR-1/P-gp inhibitor reversin 121 (5 μM). Cell viabilities were determined using trypan blue exclusion assays, and IC₅₀ values were determined. Data points represent the mean of 3 independent experiments, and error bars the SEM.

To determine the importance of MDR-1/P-gp to carfilzomib resistance in R-UMSCC-1 and R-Cal33 cells, reversin 121, an MDR-1/P-gp inhibitor, was employed. The addition of reversin 121, alone or in combination with proteasome inhibitor, resulted in modest downregulation of the MDR-1/P-gp protein (Fig. S4). Importantly, treatment with reversin 121 lowered the IC₅₀s for carfilzomib 11-fold and 8-fold in R-UMSCC-1 and R-Cal33 cells, respectively (Fig. 9B). This suggests that elevated expression of MDR-1/P-gp is likely to play a significant role in acquired resistance to carfilzomib.

As R-UMSCC-1 cells, but not R-Cal33 cells, were characterized by upregulation of Bcl-2, we also determined the impact of ABT-737, a small molecule inhibitor of Bcl-2/Bcl-X_L, on the sensitivity of the resistant cells to carfilzomib. Cells were incubated with carfilzomib in the absence or presence of a nontoxic concentration of ABT-737 (2.5 μM). As shown in Figure 10, ABT-737 caused an approximately 50% reduction in the IC₅₀ of R-UMSCC-1 cells for carfilzomib, but did not affect the sensitivity of R-Cal33 cells. This indicates that Bcl-2 upregulation can also play a role in acquired resistance to carfilzomib.

Figure 10. ABT-737 enhances the sensitivity of R-UMSCC-1, but not R-Cal33, to CFZ. R-UMSCC-1 and R-Cal33 cells were plated in triplicate wells in 96-well plates (5000 cells/well). Cells were treated for 48 h with varying concentrations of CFZ in the absence or presence of ABT-737 (2.5 μM). Cell viabilities were determined by MTT assays.

To determine the potential influence of Bcl-2 or MDR-1/P-gp upregulation on cell growth, we compared the growth rates of parental UMSCC-1 and Cal33 cells with their resistant counterparts. No differences were detected in the growth rates of the parental vs. resistant lines (Fig. S5).

Discussion

HNSCC is the seventh leading cause of cancer deaths worldwide.⁴⁸ Current therapy for HNSCC includes surgery, radiation, and/or chemotherapy. These treatment options often result in disfigurement, difficulty in speaking and swallowing, and cytotoxic side effects. Additionally, treatment success for late stage HNSCC is limited, with 5-y survival rates of roughly 50 percent. These grim facts highlight the need to develop new therapeutic approaches for the treatment of this disease. One potential approach involves the use of proteasome inhibitors, which have demonstrated considerable clinical success in the treatment of multiple myeloma and mantle cell lymphoma. In preclinical studies, HNSCC cell lines and xenograft tumors have been shown to be sensitive to the proteasome inhibitors bortezomib, carfilzomib, and oprozomib.^6-10 The modest clinical activity of bortezomib seen in early-stage testing in HNSCC patients has raised both disappointment and optimism.^11-17 While the modest level of this activity is clearly unsatisfactory, there is hope that it may be possible to enhance the clinical efficacies of proteasome inhibitors in HNSCC, and other solid tumors, through rationally designed co-targeting strategies. The development of such strategies will require a more thorough understanding of the mechanisms contributing to intrinsic and acquired resistance to proteasome inhibitors, particularly the second-generation proteasome inhibitors. Our studies reveal a mechanism contributing to acquired resistance to carfilzomib/oprozomib, as well as a co-targeting strategy, and accompanying mechanism, that can be used to overcome this resistance.

The effectiveness of proteasome inhibitors as therapeutic agents is limited by intrinsic resistance. Cancer cells often exhibit aberrant activation of signaling proteins and pathways that promote cellular survival and intrinsic resistance to proteasome inhibitors and other anti-cancer agents. In addition, because the proteasome regulates expression levels of a large number of proteins with diverse functions, proteasome inhibitors induce the expression of proteins with proapoptotic and antiapoptotic roles. Thus, bortezomib, carfilzomib, and oprozomib upregulate proapoptotic Bik and Noxa, which act to mediate the killing activities of these agents against HNSCC cells.^8,34,49 At the same time, these agents also upregulate antiapoptotic Mcl-1 and phospho-STAT3 and promote pro-survival autophagy, which act to blunt the killing activities against these cells. In in vitro studies, co-targeting of Mcl-1, phospho-STAT3, or autophagy has been shown to reduce the intrinsic resistance of HNSCC cells to bortezomib, carfilzomib, and oprozomib, and, therefore, represents a reasonable strategy for improving the therapy of these agents.^8,34,41

Acquired resistance to proteasome inhibitors also represents a very significant clinical problem, causing patients to become refractory to treatment. Although acquired resistance has primarily been studied with bortezomib in hematologic malignancies, it is anticipated that acquired resistance to carfilzomib and oprozomib will also be observed as clinical evaluation of these agents proceeds. One approach that has proven successful for overcoming acquired resistance to bortezomib, a boronic acid-based dipeptide, is to utilize a structurally distinct proteasome inhibitor. Indeed, carfilzomib and oprozomib, both of which are epoxyketone proteasome inhibitors, have been shown to be active against patient multiple myeloma cells that have become bortezomib-resistant.^26,30,31,33 Our results demonstrate that the converse also may be true. The models of acquired carfilzomib/oprozomib resistance we developed were found to exhibit only moderate resistance to bortezomib. These findings raise the possibility that patients who become refractory to frontline treatment with carfilzomib or oprozomib will retain responsiveness to bortezomib.

Our studies also identified co-targeting of HDACs and the proteasome as a strategy for combating acquired carfilzomib/oprozomib resistance in HNSCC cells. While several prior studies have demonstrated synergism between proteasome and HDAC inhibitors,^35-40 the utility of such combinations in models of acquired resistance to proteasome inhibitors is largely unknown. We also demonstrated the generality of this approach for achieving synergism, as different combinations of carfilzomib or oprozomib with vorinostat or entinostat were all effective. The molecular basis for this synergism was found to be partially dependent on upregulation of proapoptotic Bik. By contrast, downregulation of MDR-1/P-gp, which was found to be overexpressed in the acquired resistance models, does not appear to play a role in the synergism between proteasome and HDAC inhibitors, since pretreatment with HDACi alone did not affect MDR-1/P-gp expression levels (data not shown).

The overexpression of MDR-1/P-gp we observed in the acquired resistance models only partly explains the resistance of these cells. Only partial attenuation of carfilzomib resistance was seen when MDR-1/P-gp was inhibited using reversin 121. In addition, while far greater overexpression of MDR-1/P-gp was seen in R-Cal33 cells than in R-UMSCC-1 cells, R-Cal33 cells were less resistant to carfilzomib (64-fold) and oprozomib (32-fold) than R-UMSCC-1 cells (205-fold and 46-fold resistant to carfilzomib and oprozomib, respectively). It is interesting to note that R-UMSCC-1 cells, in contrast to R-Cal33 cells, also exhibited overexpression of Bcl-2 relative to parental cells. Thus, in the more resistant R-UMSCC-1 cells, the acquired resistance may reflect contributions from both MDR-1/P-gp and Bcl-2. This also suggests that inclusion of a pan-Bcl-2 inhibitor targeting both Bcl-2 and Mcl-1 may be particularly beneficial in some cases of acquired resistance. Future studies will be needed to determine an optimal co-targeting regimen for overcoming acquired resistance to carfilzomib/oprozomib, but several options, including combination(s) of these agents with an HDACi, an MDR-1/P-gp inhibitor, and/or a pan-Bcl-2 inhibitor should be considered.

Material and Methods

Establishment of carfilzomib-resistant HNSCC cell lines

Resistant cell lines were established through exposure to gradually increasing concentrations of carfilzomib (Onyx Pharmaceuticals, Inc.) over a 12-mo time period. Cells were initially exposed to 30 nM carfilzomib and, finally, to 1 µM carfilzomib. The parental cell lines (P-UMSCC-1 and P-Cal33) were maintained in DMEM containing 10% fetal bovine serum (HyClone Laboratories), and 100 units/mL penicillin and 100 μg/mL streptomycin (Invitrogen). Resistant cell lines (R-UMSCC-1 and R-Cal33) were maintained in the identical medium supplemented with 1 μM carfilzomib. Prior to the initiation of experiments involving treatment with different agents, the carfilzomib-resistant cells were cultured for 72 h in the absence of carfilzomib to prevent interference.

Cell survival and apoptosis assays

Cell survival was examined by MTT (thiazolyl blue tetrazolium bromide) assays (Sigma). Briefly, cells were seeded in triplicate at 3–5 × 10³ cells/well in 96-well plates, allowed to grow overnight, then treated for 48 h with either 0.1% DMSO, as drug diluent control, or various doses of carfilzomib or oprozomib (Onyx Pharmaceuticals, Inc.), cisplatin (University of Pittsburgh Cancer Institute Pharmacy), paclitaxel (Sigma), or bortezomib, vorinostat, or entinostat (LC Laboratories). Following treatment, cells were incubated in the dark with 20 μL/well MTT solution (5 mg/mL in PBS) for 3 h at 37 °C. The supernatants were then replaced with DMSO (100 μL/well) and the plates were agitated to dissolve the formazan crystal product. Absorbance was measured at 570 nm using a multi-well plate reader (Molecular Devices). Data were normalized to DMSO control, and the half-inhibitory concentrations (IC₅₀) of the drugs were calculated using GraphPad Prism software (GraphPad Software, Inc., version 4.03). Apoptosis was quantified by performing Annexin V/propidium iodide staining (BD BioScience, Inc.), as previously described.³⁴

Determination of synergy

Cells were seeded in triplicate (5 × 10³/well) in 96-well plates and allowed to recover overnight. The cells were then treated for 48 h with varying doses of individual drugs alone or varying doses of a constant ratio of two drugs together. Treatment with DMSO (0.1%) was used as a control. Following performance of MTT assays, data were analyzed and graphed using GraphPad PRISM software. Combination indexes (CIs) were determined according to the method of Chou and Talalay⁴⁷ using CalcuSyn V2 software (BIOSOFT). CI values lower than 1.0 were considered evidence of synergism.

Immunoblotting

Following treatment, floating and attached cells were collected and lysed in lysis buffer (10 mM TRIS-HCl, pH 7.6, 5 mM EDTA, 50 mM NaCl, 1% Triton X-100) containing protease inhibitor cocktail (1 tablet/10 mL; Roche Diagnostics Co.). Lysates were subjected to sonication and centrifugation, and the supernatants were transferred to new tubes. Bio-Rad Protein Assay dye concentrate (Bio-Rad) was used to determine protein concentrations in the lysates. Equivalent quantities of protein were electrophoresed on 10% or 12.5% SDS-PAGE gels and transferred to nitrocellulose membranes. Membranes were blocked at room temperature for 1 h in TBST buffer (150 mM NaCl, 50 mM TRIS-HCl, pH 8.0, and 0.1% Tween 20) containing 5% nonfat milk, then probed overnight at 4 °C with primary antibodies directed against caspase-3 (AAP-113, Enzo Life Sciences, Inc.), PARP (9542S, Cell Signaling Technology Inc.), Bik (SC-10770) or Mcl-1 (SC-12756, Santa Cruz Biotechnology, Inc.), MDR-1/P-gp (GTX108370, Gene Tex, Inc.), Bcl-2 (M0887, Dako, Inc.), or β-actin (A5441, Sigma). After washing, the membranes were probed with secondary antibodies for 1 h at room temperature, washed again, then developed using enhanced chemiluminescence reagent (PerkinElmer Life and Analytical Science, Inc.).

Clonogenic survival assays

Cells were seeded at 2 × 10⁶/dish in 100 mm dishes, grown overnight, then treated for 12 h with carfilzomib or vorinostat, alone or in combination. After treatment, cells were washed twice with PBS and detached from plates using trypsin. Single-cell suspensions were diluted in DMEM containing 10% FBS and an equal number of cells (200/well for R-UMSCC-1; 300/well for R-Cal33) were replated into 6-well plates. The cells were then grown for 10–15 d in the absence of drugs, followed by staining of colonies for 30 min in a solution of 6% glutaraldehyde and 0.5% crystal violet in water. The plates were washed in water until no more dye was detected in the rinse. After air-drying, colonies composed of 50 or more cells were counted in each well.

Transfection of siRNA

Cells were seeded at 3.5 × 10⁴cells in 24-well plates and transfection of siRNAs was achieved using Lipofectamine 2000 (Invitrogen), as previously described.³⁴ Nonspecific siRNA and siRNA for Bik (5′-GGGAUGUUCU UAGAAGUUUT T-3′) were obtained from Ambion.

Statistical analysis

Statistical analyses were performed using GraphPad Prism software. Comparisons between two groups applied one-way ANOVA followed by Tukey multiple comparison tests. Statistical significance was defined as P less than 0.05.

Abbreviations:
HNSCC	=	head and neck squamous cell carcinoma
CFZ	=	carfilzomib
OPZ	=	oprozomib
BTZ	=	bortezomib
CT-L	=	chymotrypsin-like
HDACi	=	histone deacetylase inhibitor
VOR	=	vorinostat
ENT	=	entinostat

Disclosure of Potential Conflicts of Interest

C.J. Kirk is employed by Onyx Pharmaceuticals, Inc. as a Vice President of Research. C.J. Kirk has ownership interest (including patents) in Onyx Pharmaceuticals, Inc. No potential conflicts of interest were disclosed by the other authors.

Acknowledgments

The authors thank Changyou Li and Tonia Buchholz for helpful discussions. These studies were supported by NIH grants R01 CA137260 and P50 CA097190. This project also used UPCI share resources supported in part by P30 CA047904.