Exploring the potential of BH3 mimetic therapy in squamous cell carcinoma of the head and neck

Squamous cell carcinoma of the head and neck (SCCHN) is the sixth most common cancer worldwide, with overall survival of less than 50%. Current therapeutic strategies involving a combination of surgery, radiation, and/or chemotherapy are associated with debilitating side effects, highlighting the need for more specific and efficacious therapies. Inhibitors of BCL-2 family proteins (BH3 mimetics) are under investigation or in clinical practice for several hematological malignancies and show promise in solid tumors. In order to explore the therapeutic potential of BH3 mimetics in the treatment of SCCHN, we assessed the expression levels of BCL-2, BCL-XL, and MCL-1 via Western blots and immunohistochemistry, in cell lines, primary cells derived from SCCHN patients and in tissue microarrays containing tumor tissue from a cohort of 191 SCCHN patients. All preclinical models exhibited moderate to high levels of BCL-XL and MCL-1, with little or no BCL-2. Although expression levels of BCL-XL and MCL-1 did not correlate with patient outcome, a combination of BH3 mimetics to target these proteins resulted in decreased clonogenic potential and enhanced apoptosis in all preclinical models, including tumor tissue resected from patients, as well as a reduction of tumor volume in a zebrafish xenograft model of SCCHN. Our results show that SCCHN is dependent on both BCL-XL and MCL-1 for apoptosis evasion and combination therapy targeting both proteins may offer significant therapeutic benefits in this disease.


Introduction
Head and neck cancer, of which squamous cell carcinomas account for over 90% of diagnoses, is the sixth most common cancer worldwide 1 . Although etiological factors, such as tobacco use, alcohol consumption, and human papillomavirus (HPV) infection, as well as the mutation of key genes associated with SCCHN have been identified, this has not translated into better therapies and the prognosis for many HPV negative (HPV − ) patients remains significantly poorer than for HPV positive patients [2][3][4] . With the currently preferred treatment modalities of surgery, radiotherapy and/or chemotherapy, local recurrence or metastasis is observed in >50% of HPV − patients and is often associated with significant morbidity and mortality 5 , thus emphasizing the urgent need for improved therapeutic strategies. Until recently, the only targeted therapy approved for SCCHN was the monoclonal antibody, cetuximab, which inhibits the epidermal growth factor receptor (EGFR). Although EGFR is frequently overexpressed in SCCHN 4 , responses to cetuximab have been varied, and generally disappointing 4,6 . However, there is new promise for immunotherapy, with the recent FDA approval of the PD-1 specific antibody Pembrolizumab for first-line treatment of SCCHN 7 .
Evasion of apoptosis is one of the cardinal features of cancer. Most anticancer therapies facilitate cancer cell death by inducing the intrinsic pathway of apoptosis, which is regulated by the balance of antiapoptotic and proapoptotic BCL-2 family proteins 8 . High expression levels of antiapoptotic BCL-2 family members (BCL-2, BCL-X L , and MCL-1) allow apoptotic evasion in multiple malignancies, prompting the development of a novel class of inhibitors (BH3 mimetics) for these proteins 9 . The first BH3 mimetics, ABT-737 and ABT-263 (Navitoclax), inhibit BCL-2, BCL-X L , and BCL-w 10,11 , and were followed by more selective inhibitors, such as the BCL-2specific Venetoclax (ABT-199), recently approved for treatment of chemorefractory chronic lymphocytic leukemia 12,13 , in addition to newly diagnosed acute myeloid leukemia, in combination with hypermethylating agents, such as azacytidine and decytabine 14 . Similarly, BH3 mimetics that specifically target either BCL-X L or MCL-1 show promise in several hematological malignancies [15][16][17][18][19] . Studies utilizing BH3 mimetics in SCCHN have largely been limited to ABT-737, which reportedly synergizes with cisplatin, etoposide and radiation [20][21][22] .
Our aim was to assess the potential of BH3 mimetic therapy in SCCHN utilizing several preclinical models including explant cultures of surgically resected human tumors and a zebrafish xenograft model.
Cloning and generation of H2B-mRFP-positive cells for zebrafish xenografts Histone H2B-mRFP fusion protein cDNA was amplified from pHIV-H2B-mRFP (AddGene #18982) by polymerase chain reaction and transferred to a modified form of the lentiviral plasmid pLJM1 (AddGene #19319) and sequenceverified. To make H2B-mRFP-expressing lentivirus, 4 × 10 6 HEK293T cells were transfected with the lentiviral packaging plasmids; pMD2.G (AddGene #12259) and psPAX2 (AddGene #12260) together with pLJM1 H2B-mRFP using polyethylenimine (1 mg/ml), and lentivirus harvested 72 h later. This was used for transducing UM-SCC-81B cells, in the presence of Polybrene (4 µg/ml) by spinoculation (2000 rpm, 1 h). Upon validation of nuclear expression of H2B, cells stably expressing H2B were selected by exposure to puromycin (1 µg/ml) for 2 weeks. BH3 profiling, apoptosis, and colony formation assays BH3 profiling and apoptosis (assessed by measuring the extent of phosphatidylserine (PS) externalization) were performed as described previously 16 . For colony formation assays, a defined number of cells were exposed to different drugs and/or X-ray irradiation using a CellRad X-ray irradiator (Faxitron Bioptics) and incubated for 7-14 days. Cell colonies were fixed using 1:7 vol/vol acetic acid/methanol for 5 min and stained with 0.5% crystal violet (in 20% methanol) for 2 h, then counted on a Gel-Count colony analyzer (Oxford Optronix). Plating efficiency (PE) for the untreated controls and surviving fraction (SF) were determined using the formulae:

Western blotting
Whole cell extracts were prepared in RIPA buffer containing cOmplete Mini protease inhibitor cocktail (Sigma, 11836170001) and MG-132 (20 µM) followed by centrifugation at 14,000 rpm at 4°C for 10 min. Proteins were resolved on 10 or 15% sodium dodecyl sulfate polyacrylamide gel electrophoresis, transferred to nitrocellulose membranes (VWR, 10600002.P), incubated with the indicated primary and secondary antibodies and visualized using ECL reagents (GE Healthcare) and ChemiDoc (BioRad).

Immunohistochemistry (IHC), image acquisition, and analysis
Explants, surgically resected from patients, were acclimatized overnight to the culture conditions before exposure to BH3 mimetics for 48 h. Explants were fixed in 4% paraformaldehyde at 4°C for 24 h and infiltrated with paraffin using a HistoCore PEARL Tissue Processor (Leica). Following embedding in paraffin blocks, tissues were cut into 5 μm sections. Sections from explants and TMAs (generated in-house-clinical and pathological patient details in Table 1) were processed and stained using the Bond RX m autostainer (Leica). Briefly, TMA sections were deparaffinized, subjected to heat induced epitope retrieval, endogenous peroxidase activity blocked, and then incubated with the relevant primary antibodies and a non-species-specific linker and HRP, which were detected using DAB + chromogen. Sections were counterstained with hematoxylin, dehydrated, cleared and mounted using EcoMount (Biocare Medical). Images were acquired at 20× magnification using a Nikon Eclipse E800 microscope for explants, or by scanning at 40× magnification using an Aperio slide scanner (Leica) for TMAs. Explant staining was quantified by counting the SCCHN nuclei in each image and determining the percentage of positively stained cells. TMA antibody staining intensity was analyzed using QuPath software 23 . The cytoplasmic DAB optical density was measured for each tumor cell within a tissue sample, and a histo-score (H-score; a quantitative measurement of the intensity of staining for a particular antibody) generated for that sample. The Hscore was averaged between two or three tissue samples available per patient. Kaplan-Meier survival curves were generated using GraphPad Prism version 8.0.1 (GraphPad Software) and overall survival data from patients in the lower-and upper-most quartiles for expression (H-score) of each antiapoptotic protein.

Zebrafish xenografts
Zebrafish studies were completed under the approval of the University of Liverpool Animal Welfare and Ethical Review Body. Nacre ubiq:secAnnexinV-mVenus embryos 24 , obtained from The University of Manchester Biological Services Facility, were incubated at 28°C in egg water (60 μg/ml Tropic Marin salt in distilled water) until 48 h postfertilization (hpf). Following cell implantation, fish were maintained in a humidified light-cycling incubator (14 h on, 10 h off) at 34°C. For toxicity studies, embryos at 72 hpf were exposed to increasing concentrations of BH3 mimetics (at 34°C) and mortality assessed up to 120 hpf. For xenograft studies, UM-SCC-81B cells stably expressing H2B-mRFP were injected into the pericardial cavity at 48 hpf, as Includes those patients who are N0 (no nodal involvement). b Recurrence data is to 10 years post-diagnosis for oral cavity, and 5 years for hypopharynx and larynx.
previously described 25 . At 72 hpf, fish were screened for the presence of red fluorescent cell masses, following anesthetization using MS-222 (160 µg/ml for maximum 15 min) and images acquired using a Leica MZ16F microscope. Successfully xenografted fish were randomly placed individually in wells of a 24-well plate and either DMSO or BH3 mimetics added to 1 mM Tris-buffered egg water. Fish were euthanized by 120 hpf using MS-222 (250 µg/ml). Images of tumor masses were acquired and 2D tumor areas calculated using ImageJ software.

SCCHN cell lines express varying levels of anti-apoptotic proteins and are primed to undergo apoptosis
Given the poorer clinical outcome associated with HPV − SCCHN, we wished to explore the potential of BH3 mimetic therapy in this disease. We first determined the expression profile of the major antiapoptotic BCL-2 family proteins in six HPV − cell lines, derived from different head and neck subsites. MCL-1 and BCL-X L were expressed in all the cell lines, whereas BCL-2 was expressed in only 3 of the 6 cell lines tested (Fig. 1a). To ascertain the responsiveness of SCCHN cell lines to apoptotic stimuli, BH3 profiling was utilized by exposing cells to increasing concentrations of BIM peptide, which binds promiscuously to all anti-apoptotic BCL-2 family members 26 . All cell lines exhibited a concentration-dependent increase in mitochondrial membrane depolarization (Fig. 1b), suggesting they were all primed to undergo apoptosis.
BH3 mimetics, as single agents, do not induce apoptosis in SCCHN cell lines, nor do they sensitize cells to cisplatin or irradiation None of the specific BH3 mimetics, ABT-199 (BCL-2 specific 12 ), A-1331852 (BCL-X L specific 15 ), or S63845 (MCL-1 specific 17 ), induced pronounced apoptosis as single agents (Fig. 1c). Moreover, none of the BH3 mimetics synergized with cisplatin or irradiation in significantly reducing the clonogenic potential of the SCCHN cell lines tested, except UM-SCC-1 cells, which showed synergism between A-1331852 and either cisplatin or radiotherapy (Supplementary Figs. S1 and S2).
IHC of SCCHN tumor tissue reveals high expression of BCL-X L and MCL-1 but little/no BCL-2 Next, we examined the expression of BCL-2, BCL-X L , and MCL-1 in TMAs comprising tissue obtained from 191 SCCHN patients at the time of surgery, from oral cavity, hypopharyngeal and laryngeal tumors. The patient population was representative of the typical spectrum of HPV − SCCHN ( Table 1). The TMAs included tissues classified as core (center of the primary tumor), advancing front (invasive edge of the tumor) and normal (non-malignant adjacent tissue). IHC was performed using antibodies validated by Western blotting and IHC in SCCHN cell lines in which the anti-apoptotic proteins were downregulated with RNAi ( Supplementary Fig. S3). A growing literature surrounds the use of antibodies that are not carefully validated for IHC analysis, which result in inconsistent, and often misleading, correlations with patient outcome 27,28 . While the antibodies against BCL-2 and BCL-X L exhibited specificity in both Western blots (characterized by a specific band at the predicted molecular weight) and IHC (characterized by specific mitochondrial staining, which decreased following siRNA transfections), problems were evident with the different MCL-1 antibodies used ( Supplementary Fig. S3). Some antibodies detected multiple isoforms of MCL-1, whereas others were specific to the predominant, antiapoptotic isoform 29 . IHC revealed differing staining patterns of MCL-1, ranging from nuclear localization to cytosolic/mitochondrial staining (Supplementary Fig. S3). Using RNA interference and specific mitochondrial staining as primary indicators of specificity, we deemed clone D5V5L the most specific MCL-1 antibody, which was used in all subsequent experiments. TMAs were stained with the appropriate antibodies and protein expression levels scored according to the staining intensity, denoted by the histoscore (H-score) (Fig. 1d-f).
Expression of BCL-2 was low in all normal and tumor tissue from the oral cavity and hypopharynx (Figs. 1d and 2a), whereas it appeared to be more highly expressed in the tumor core, advancing front and in the adjacent normal tissues of the larynx (Fig. 2b). BCL-X L expression was significantly elevated in the tumor core and advancing front of oral, hypopharyngeal and laryngeal SCCHN, compared to adjacent normal tissues (Figs. 1e and 2a, b). MCL-1 was highly expressed in the tumor core from the oral cavity compared to the normal tissue, whereas this was not evident in the advancing front (Fig. 1f). MCL-1 expression was generally higher in the tumor core and advancing front of hypopharynx, compared to the normal tissues, whereas in the larynx, MCL-1 was relatively highly expressed both in normal and tumor tissues (Fig. 2a, b). Taken together, this data suggested that BCL-X L is the most consistently upregulated BCL-2 family member in SCCHN, although targeting BCL-X L alone did not result in enhanced apoptosis in SCCHN cell lines (Fig. 1c).

High expression levels of BCL-X L and MCL-1 do not correlate with patient outcome
No correlation was observed between expression of BCL-2, BCL-X L , or MCL-1 and overall survival of patients with oral cavity tumors (Fig. 3a-c). This was surprising as numerous studies have reported positive correlations between these proteins and patient outcome in different malignancies, often using antibodies that had not been validated for IHC analysis. The importance of using carefully validated antibodies was further emphasized by the stark contrast in survival plots generated using different MCL-1 antibodies (Supplementary Fig. S4). Since individual expression levels of MCL-1 and BCL-X L did not correlate with patient survival, we compared the quartiles with the highest and lowest expression of both MCL-1 and BCL-X L together and found no statistically significant differences in overall survival (Fig. 3d).
Combination therapy targeting BCL-X L and MCL-1 induces apoptosis in preclinical models of SCCHN Although we did not find a correlation between BCL-X L and MCL-1 and patient outcome, these antiapoptotic proteins are highly expressed in the majority of SCCHN patients, suggesting their involvement in carcinogenesis. As many solid tumors depend on both BCL-X L and MCL-1 for survival 30 , it raised the possibility that targeting these two proteins together may be a beneficial treatment strategy. In support of this, a combination of specific BCL-X L and MCL-1 inhibitors, A-1331852 and S63845, induced marked apoptosis in all SCCHN cell lines (Fig.  4a). In contrast, exposure to the BCL-2-specific inhibitor, ABT-199, either alone or in combination with A-1331852 or S63845 did not enhance apoptosis (Supplementary Fig.  S5). These results strongly suggested that these cell lines are dependent on both BCL-X L and MCL-1 for survival, with little or no contribution from BCL-2. Furthermore, a combination of A-1331852 and S63845 markedly inhibited the clonogenic potential of all six SCCHN cell lines, whereas either inhibitor alone exerted little or no inhibition (Fig. 4b). Together, A-1331852 and S63845 markedly OC oral cavity, LX larynx, OP oropharynx. b BH3 profiling with increasing concentrations of BIM peptide revealed that the indicated SCCHN cell lines were primed to undergo apoptosis. c SCCHN cell lines exposed to the indicated BH3 mimetics (100 nM) for 24 h failed to undergo apoptosis, as assessed by phosphatidylserine (PS) externalization. d-f Representative IHC images of oral cavity (OC) normal tissue, primary tumor core and advancing front (AF), stained using antibodies against d BCL-2 e BCL-X L , and f MCL-1 (clone D5V5L) and counterstained with hematoxylin. Scale bars 50 μm. Dot plots depict the relative staining intensities (H-scores) of the specified antibody. Each dot represents the data from an individual patient. The IHC images shown correspond to the yellow circles in the graph. b, c One-way ANOVAs with Dunnett's multiple comparisons tests, with a single pooled variance. Error bars = mean ± SEM of at least three independent experiments. d-f Normality tests were performed, followed by Kruskal Wallis tests with Dunn's multiple comparisons tests. Error bars = mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001.
inhibited the growth of 3D tumor spheroids, which mimic the in vivo behavior of tumor cells more closely than 2D cultures 31 , whereas S63845 alone caused a modest inhibition and A-1331852 had no effect (Fig. 4c, d). Consistent with these findings, primary cell lines derived from oral cavity tumors also expressed both BCL-X L and MCL-1 with little detectable BCL-2 (Fig. 4e). Furthermore, the combination of A-1331852 and S63845 induced apoptosis and inhibited the clonogenic potential of all SCCHN patient-derived cell lines (Fig. 4f, g).

Targeting BCL-X L and MCL-1 induces apoptosis in tumor tissues resected from SCCHN patients
To increase the translational relevance and more closely mimic the in vivo setting, tumor tissues that maintain tissue architecture as well as several aspects of the tumor microenvironment were surgically resected from SCCHN patients and exposed to A-1331852 and/or S63845, and the extent of apoptosis assessed by the appearance of cleaved PARP in apoptotic nuclei. Exposure of tumor explants to either A-1331852 or S63845 alone did not induce apoptosis, whereas the combination resulted in a significant induction of apoptosis (Fig. 5a, b).

Targeting BCL-X L and MCL-1 reduces tumor burden in zebrafish xenografts
To ascertain whether this combination was effective in vivo, a NC3Rs (National Center for Replacement, Refinement and Reduction of animals in research)-compliant zebrafish SCCHN xenograft model, which allows moderate to high throughput analyses for cancer discovery studies, was employed. BH3 mimetics have previously been administered to zebrafish embryos, with minimal toxicity 32 . Initial toxicity studies with our BH3 Fig. 2 BCL-X L and MCL-1 but not BCL-2 are highly expressed in hypopharynx and larynx tissue microarrays from SCCHN patients. Representative IHC images of normal tissue, primary tumor core and advancing front (AF) from a hypopharynx (HP) and b larynx (LX) stained against the indicated antibodies and counterstained with hematoxylin. Scale bars 50 μm. The plots depict the relative staining intensities (H-score) of the specified antibody. Each dot represents the data from an individual patient. The IHC images shown correspond to the yellow dots in the graph. Normality tests were performed, followed by Kruskal Wallis tests with Dunn's multiple comparisons tests. Error bars = mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001. mimetics were performed using 3-5 day zebrafish embryos, and sublethal concentrations (1-2 μM) chosen for subsequent experiments (Supplementary Fig. S6). To establish tumors in vivo, fluorescently labeled SCCHN cells were microinjected into the pericardial cavity at 48 hpf. Following verification of the presence of fluorescent cell masses 24 h postinjection, zebrafish embryos were randomly allocated either to a DMSO only (control) group or exposed to a combination of A-1331852 and S63845 for 48 h (72-120 hpf) and final tumor area assessed (Fig. 5c, d). Dual inhibition of BCL-X L and MCL-1 resulted in a significant dose-dependent reduction in tumor size compared to controls (Fig. 5c, d). These results were in agreement with the data generated using cell lines, primary cells and patient explants.

Discussion
A major mechanism by which cancer cells evade apoptosis is overexpression of antiapoptotic BCL-2 family proteins 8 . Hematological malignancies largely depend on a single antiapoptotic member for survival, and as a result, monotherapy with BH3 mimetics has been successful 12,16,17 . However, as our study and others show, solid tumors depend on more than one BCL-2 family member for survival, and in most instances, neutralization of both BCL-X L and MCL-1 is required to induce efficient apoptosis 30,33,34 .
Just a handful of papers exist correlating BCL-X L and MCL-1 expression with patient outcome in SCCHN 36,41,43,47,48 . Many of these studies have relied on mRNA levels rather than IHC and report either a positive or negative correlation to clinical outcome. In our study, expression of BCL-X L and MCL-1 was generally higher in tumors than surrounding normal tissues, suggesting that some selectivity could be gained by targeting these proteins ( Figs. 1 and 2). MCL-1 expression appeared to decrease in the oral cavity advancing front (Fig. 1), possibly due to the functional redundancy between MCL-1 and BCL-X L . BCL-X L may play an additional role in the invasive advancing front, as it has recently been proposed to promote epithelial-mesenchymal transition and tumor metastasis 49 .
Despite high expression in tumor compared to normal tissues, expression levels of BCL-X L and MCL-1 did not correlate with patient outcome in SCCHN (Figs. 1-3). This could have been confounded by heterogeneity in the adjuvant treatment (none or radiotherapy±chemotherapy) post-surgery and warrants a prospective cohort analysis. Although BCL-X L and MCL-1 were not prognostic indicators in SCCHN, dual inhibition resulted in extensive apoptosis in all preclinical models including tumor explants, as well as a decreased tumor burden in zebrafish xenografts (Figs. 4 and 5). Since BCL-X L and MCL-1 share a high degree of functional redundancy both in tumor survival and in normal cellular homeostasis, a combination of inhibitors targeting these proteins may result in some degree of toxicity to normal tissues. Inhibition of BCL-X L results in thrombocytopenia due to platelets being dependent on BCL-X L for survival 50 , whereas inhibition of MCL-1 will affect survival of neutrophils 51 , thus mandating the establishment of a therapeutic index and optimal scheduling for the drug combination, as has been achieved to manage Navitoclaxinduced thrombocytopenia in patients 52 . Furthermore, it is important to note the preferential expression of both BCL-X L and MCL-1 in tumor compared to the normal tissues ( Figs. 1 and 2), which provides a basis for BH3 mimetics to selectively target the tumor tissue.
First-line therapy of cisplatin only modestly improves overall survival in SCCHN patients, at the cost of significant toxicity. Furthermore, many HPV − SCCHN patients with locally advanced disease are simply not fit Fig. 4 Inhibition of both MCL-1 and BCL-X L induces apoptosis in SCCHN cell lines and primary cells. SCCHN cell lines exposed to a combination of A-1331852 and S63845 (100 nM each) for 24 h exhibited a enhanced apoptosis and b decreased clonogenicity, compared to the individual BH3 mimetics. c Phase contrast images (scale bars 100 μm), and d line graph to show fold change in spheroid volume over 12 days, following exposure to the specified drug(s). Dotted lines in c demarcate the outline of each intact spheroid. e Western blot analysis of antiapoptotic protein levels in SCCHN primary cells. These cells, exposed to a combination of A-1331852 and S63845 (100 nM) for 4 h exhibited f enhanced apoptosis and g decreased clonogenicity, compared to the indicated treatments. a, b, f One-way ANOVAs with Dunnett's multiple comparisons tests, with a single pooled variance; d two-way repeated measures ANOVA with Dunnett's multiple comparisons test; g unpaired, two-tailed t tests. Error bars = mean ± SEM of at least three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001. enough for cisplatin therapy, thus emphasizing the need to identify and introduce better, more targeted therapies for this malignancy. It is possible that dual inhibition of BCL-X L and MCL-1 could potentially replace cisplatin in patients unable to tolerate this, or reduce the doses of radiotherapy and/or cisplatin administered to patients.
In conclusion, we convincingly demonstrate that BCL-X L and MCL-1 but not BCL-2 are highly expressed in tumor tissues of SCCHN patients. The low expression or absence of BCL-2 suggests that there may be little benefit in using venetoclax to treat SCCHN. We also show that SCCHN is dependent on both BCL-X L and MCL-1 for survival, which can be efficiently targeted to induce marked apoptosis and tumor shrinkage, suggesting that a combination BH3 mimetic therapy may offer significant benefits in SCCHN.