Targeting BCL2 Overcomes Resistance and Augments Response to Aurora Kinase B Inhibition by AZD2811 in Small Cell Lung Cancer

Small cell lung cancer (SCLC) is an aggressive lung tumor of neuroendocrine origin with poor prognosis (1) and limited treatment options (2). Therapeutic resistance to the standard-of-care treatment (a combination of chemotherapy and immunotherapy) also emerges rapidly, necessitating the need for new and effective treatment approaches, particularly in the relapsed setting (2). SCLC is characterized by a near universal loss of TP53 and RB1 tumor suppressor genes (3). About 20% of SCLC tumors also exhibit genomic amplification of the MYC oncogene (3–5). While the paucity of targetable oncogenic driver mutations in SCLC has precluded the development of effective targeted therapies, the genomic and proteomic alterations prevalent in SCLC have been shown to induce targetable synthetic lethal dependencies (6). Aurora kinases (AURK) are key regulators of mitosis entry and progression. Aurora kinase A (AURKA) regulates mitotic entry, spindle assembly, and centrosome function (7). Aurora kinase B (AURKB), however, forms an integral part of the chromosome passenger complex and regulates chromosome condensation, kinetochore attachment, and spindle assembly checkpoint, ensuring proper chromosomal segregation and cytokinesis (8, 9). Both AURKs are upregulated by MYC and have been shown to be essential for the maintenance of MYC-driven cancers (10, 11). Given their roles in tumorigenesis and overexpression in many cancers, AURK inhibitors have been investigated as potential therapeutic agents (12–14). Small-molecule inhibitors of AURKA and AURKB, such as alisertib and barasertib, have shown promising activity in preclinical SCLC models, particularly in those with high MYC (cMYC) and NEUROD1 expression (15–18). Retrospective analyses have also shown that patients with relapsed SCLC, particularly those with high cMYC expression, had a relatively better treatment outcome with alisertib and/or paclitaxel compared with placebo, underscoring the need for testing AURK inhibitors in biomarker-defined patient cohorts (19).

In this study, we evaluated the therapeutic potential of AURKB inhibition by a selective inhibitor, AZD2811 (also known as AZD1152 hydroxyquinazoline pyrazol anilide; AZD1152-HQPA; ref. 20), using in vitro and in vivo models that recapitulate the genomic and proteomic heterogeneity in SCLC. AZD2811 has also been recently evaluated in phase I clinical trials in SCLC (NCT02579226, NCT04745689; ref. 21). We report here biomarkers of response and rational combinations to address resistance to single-agent AZD2811. Overall, the findings of this study provide promising preclinical proof of concept that the rational and biomarker-driven combinations of AZD2811 could be an effective treatment strategy in SCLC.

Antibodies for Western blotting purchased from Cell Signaling Technology: BCL2 (#15071, 1:1,000, RRID:AB_2744528), BCL-xL (#2764, 1:1,000, RRID:AB_2228008), cleaved PARP (#5625, 1:1,000, RRID:AB_10699459), γH2AX (#9718, 1:1,000, RRID:AB_2118009), phospho-Histone H3 (pHH3; S10; #53348, 1:1,000, RRID:AB_2799431), and Santa Cruz Biotechnology: β-actin (sc-47778, 1:2,000, RRID:AB_626632). AZD2811 was provided by AstraZeneca. Venetoclax was purchased from Selleck Chemicals. Propidium iodide was purchased from Sigma-Aldrich. RNase A (EN0531) and FITC Annexin-V apoptosis detection kit (#556547) were purchased from Thermo Fisher Scientific and BD Pharmingen, respectively.

Human SCLC cell lines were purchased from ATCC or Sigma-Aldrich. The human patient-derived xenograft (PDX) cell line NJH29 was generously provided by Dr. Julien Sage (Stanford University, Stanford, CA). Cell lines were cultured in RPMI, supplemented with 10% FBS and 100 IU/mL penicillin and 100 μg/mL streptomycin, unless otherwise specified by ATCC, at 37°C in a humidified chamber with 5% CO₂. All cell lines were authenticated by short tandem repeat profiling, maintained in culture for less than 2 months and tested regularly for Mycoplasma contamination using MycoAlert Plus (Lonza).

SCLC cells (2 × 10⁵) were transfected with either pLOC lentiviral vectors expressing wild-type human BCL2 or empty plasmid using polybrene (4 μg/mL). The lentivirus-containing medium was removed 12 hours later following transfection. The cells infected with the lentivirus vectors were continuously selected using blasticidin (Wako). BCL2 overexpression was confirmed by Western blotting.

SCLC cells (2 × 10⁵), cultured in the medium containing 10% FBS for 24 hours, were transfected with SilencerSelect BCL2 siRNA (#4392420), BCL-xL siRNA (#4390824), or negative control #1 siRNA (#4390843; Thermo Fisher Scientific) using Lipofectamine RNAiMAX for 48 hours.

SCLC cell lines and BCL2-overexpressing isogenic cell lines were seeded at a density of 2,000 cells/well in 96-well white bottom microtiter plates and treated with DMSO control or AZD2811 at different concentrations in triplicate for 96 hours. Cell viability was measured using CellTiter-Glo (Promega) and luminescence was read on a Synergy HT microplate reader (BioTek). For single-drug treatments, dose–response curves were modeled by nonlinear curve fitting and the IC₅₀ values were estimated using our previously published drexplorer software (22). Replicate reproducibility was determined by concordance correlation coefficient and goodness of fit by residual standard error. For drug combination experiments, cells were treated with AZD2811 and/or venetoclax (or BCL-xL inhibitor) or with DMSO control. The AUC for the observed effect of the combination was compared with the AUC for the additive effect predicted by the BLISS model. The difference between the two AUCs, denoted by ΔAUC, was computed. ΔAUC values <−0.1 was considered to be a greater than additive, based on an estimated 10% margin of experimental variability (22).

Proteomic expression of SCLC cell lines was performed as described previously (23).

Cells (1 × 10⁶) were seeded in 10 cm dishes and treated as indicated for 48 hours. Cells were collected by centrifugation at 1,500 rpm for 5 minutes and washed with ice-cold PBS. The cell pellet was then lysed with reverse phase protein array (RPPA) lysis buffer supplemented with protease and phosphatase inhibitor cocktail. The lysate was centrifuged at 14,000 rpm for 10 minutes to remove cell debris. Total protein concentration of the supernatant was measured using DC protein assay reagent (Bio-Rad). A total of 30 μg of cell lysate was boiled for 5 minutes at 95°C with 4X laemmli buffer, resolved on a 10% or 15% polyacrylamide gel and electroblotted onto a nitrocellulose membrane. Membranes were blocked in 1X Caesin blocking solution (Bio-Rad) for 1 hour at room temperature and incubated overnight with primary antibodies at specified dilutions at 4°C. The membranes were then washed with TBS with 0.1% Tween-20 and incubated with appropriate horseradish peroxidase–linked secondary antibodies [Santa Cruz Biotechnology anti-mouse (sc-516102, 1:2000, RRID:AB_2687626), Cell Signaling Technology anti-rabbit (#7074, 1:2000, RRID:AB_2099233)] for 1 hour at room temperature. The immunoblots were visualized using the SuperSignal West Pico Plus chemiluminescent substrate (Thermo Fisher Scientific) on a Bio-Rad ChemiDoc Touch imaging system. Tumor fragments were homogenized and subjected to Western blotting similarly.

A total of 0.5 × 10⁶ cells were plated in a 60 mm dish and treated with DMSO or AZD2811 (30 nmol/L). Cells were harvested at 48 hours, washed with ice-cold PBS, and fixed in 70% cold ethanol overnight at 4°C. Cells were collected by centrifugation, washed twice with cold PBS, and stained with 50 μg/mL propidium iodide and 250 μg/mL RNAase A for 1 hour at 37°C. Cells were analyzed on a Gallios cell analyzer (Beckman Coulter) and data were analyzed using Flow Jo software (Treestar; RRID:SCR_008520).

SCLC cells (0.5 × 10⁶) were treated with indicated concentrations of AZD2811, venetoclax, or their combination for 72 hours and harvested. Live cells were then stained with Annexin-V FITC antibody and propidium iodide, using the Annexin-V FITC Apoptosis Detection Kit (BD Biosciences), per the manufacturer's instructions. Cell staining was analyzed on Gallios cell analyzer (Beckman Coulter).

SCLC cell lines (25,000 cells/well) were treated with indicated concentrations of AZD2811, venetoclax, or their combination in triplicate for 72 hours in a 96-well white bottom microtiter plate. Caspase activity was measured using Caspase-Glo 9 assay and Caspase-Glo 3/7 assay kits (Promega).

Six-week-old athymic nude mice from Envigo were used for the animal experiments and maintained in strict accordance with protocols approved by the Institutional Animal Care and Use Committee of the University of Texas MD Anderson Cancer Center and the NIH Guide for the Care and Use of Laboratory Animals. All efforts were made to minimize animal suffering. Circulating tumor cell–derived xenograft (CDX) model (SC16), was generated as described previously and maintained in vivo by serial transplantation (24). SC49PDX, established from the core needle biopsy as described previously (24), was provided by Dr. Jack Roth. Low passage viable tumor fragments were implanted subcutaneously into the flank of nude mice. When tumors reached an average volume of about 100 mm³, the mice were randomized into two groups and treated with vehicle or AZD2811NP (10 mg/kg, diluted in saline, intravenous injection once a week). Tumor volumes and weights were measured every 2–3 days. Cell line xenograft experiments (H1048, H69, and H211) were done in accordance with all relevant ethical regulations for animal testing and research following AstraZeneca's global bioethics policy and received ethical approvals from the AstraZeneca ethical committee. Left flank of nude female mice were injected subcutaneously with cells resuspended in PBS. When the tumors reached nearly 200 mm³, the mice were randomized into four groups and treated with vehicle, AZD2811NP (25 mg/kg diluted in saline, intravenous injection once a week), venetoclax (100 mg/kg, dosed orally once a day), or their combination. Mice weights were measured daily and tumor volumes were measured twice weekly.

Studies using SCLC PDX models were carried out at XenTech, in accordance with the French regulatory legislation concerning the protection of laboratory animals and in accordance with a currently valid license for experiments on vertebrate animals, issued by the French Ministry of Higher Education, Research and Innovation. Female athymic nude mice ages 6 to 9 weeks were anesthetized with 100 mg/kg ketamine hydrochloride and 10 mg/kg xylazine. Skin of anesthetized mice was asepticized with a chlorhexidine solution, incised on the interscapular region, and one 20 mm³ tumor fragment was placed in the subcutaneous tissue. Skin was closed with clips. All mice from the same experiment were implanted on the same day. Mice were randomized to the different treatment groups. For each group, 3 mice with established tumors, with tumor volume ranging 60 to 200 mm³, were included in the study. Mice were treated with AZD2811NP (25 mg/kg diluted in saline, intravenous injection once a week), venetoclax (100 mg/kg, dosed orally once a day), AZD0466 (34 mg/kg, intravenous injection once a week), and the combination of AZD2811NP and venetoclax/AZD0466 using above mentioned doses and schedules as indicated.

Data statistics and bioinformatics analyses were performed using R (version 3.3.0, https://www.r-project.org/; RRID:SCR_001905) and Bioconductor packages (https://www.bioconductor.org/). Statistical comparisons were performed using unpaired t test for two-tailed P value, unless specified otherwise. For RPPA expression analyses, Benjamini–Hochberg method was used to control FDR (25).

Any materials are available from the corresponding author upon reasonable request.

A previous study showed that loss of RB1, a hallmark of SCLC, resulted in a hyperdependency on AURKB and enhanced sensitivity to AURKB inhibition (26). On the basis of this, we evaluated the effects of AZD2811, a selective AUKRB inhibitor, on cell viability in a panel of 57 SCLC cell lines, where most of them exhibited RB1 deficiency at the genetic and/or proteomic level. Interestingly, only 15 of 57 SCLC cell lines (26%) were highly sensitive to AZD2811 (IC₅₀ < 30 nmol/L) while 9 of 57 (16%) showed intermediate sensitivity (IC₅₀ = 30–100 nmol/L; Fig. 1A). A majority of SCLC cell lines (58%), despite RB1 deficiency, were highly resistant to AZD2811 (IC₅₀ > 100 nmol/L, the highest concentration tested; Fig. 1A). Our data therefore suggest that, while RB1 loss is near universal in SCLC, it is not sufficient for sensitivity to AURKB inhibition by AZD2811. Therefore, to identify potential predictors of drug response, we applied an unbiased approach by comparing the proteomic expression profiles, generated by RPPA (23), of SCLC cell lines that were sensitive to AZD2811 (IC₅₀ < 100 nmol/L) to those that were resistant (IC₅₀ > 100 nmol/L).

A synthetic lethal interaction between MYC overexpression and AURKB inhibition has been previously reported in other cancers (27). MYC/cMYC has also been previously shown to be a biomarker of sensitivity to both AURKA and AURKB inhibitors in SCLC (6, 18, 19, 28). Consistent with these, we found that the mean expression of cMYC protein was higher in AZD2811-sensitive cell lines (fold change = 1.95, P < 0.05; Supplementary Fig. S1A). However, of the 24 sensitive SCLC cell lines, only 11 exhibited high cMYC levels, based on the bimodal distribution of cMYC protein expression by RPPA, while eight cell lines with high cMYC levels were resistant to AZD2811 (Fig. 1A; Supplementary Table S1). Furthermore, several SCLC cell lines, even in the absence of MYC amplification/cMYC overexpression, were found to be highly sensitive to AZD2811, highlighting that sensitivity to AURKB inhibition was not limited to MYC-driven SCLC.

Beyond cMYC, we found that SCLC cell lines that were resistant to AZD2811 expressed high levels of the antiapoptotic protein BCL2 (fold change = −1.6, P = 0.06; Fig. 1B). No differences in the other BCL2 family members were seen. Similarly, we also found BCL2 to be inversely associated with in vivo AZD2811 sensitivity in PDX models established from either the circulating tumor cells (CDX) or a core needle biopsy (PDX) from patients with SCLC, following relapse to frontline platinum therapy (24). In two such models—MDA-SC16 (CDX) and MDA-SC49 (PDX)—where we previously reported AURKA and AURKB expression in specific cell clusters (24), we tested efficacy of the slow release nanoparticle formulation of AZD2811 (ref. 29; AZD2811NP) as a single agent at a low dose (10 mg/kg). AZD2811NP induced significant growth inhibition of the MDA-SC49 tumors, which express high levels of MYC and low levels of BCL2, while it had little effect on MDA-SC16 tumors, which have high BCL2 and low MYC expression (Fig. 1C).

BCL2 is frequently overexpressed in SCLC cell lines and tumors (30, 31). When we analyzed the transcriptomic profiles of 57 SCLC cell lines and 81 treatment-naïve SCLC tumors (3), BCL2 was found to be overexpressed in both cohorts (Fig. 1D). As reported previously (17), BCL2 mRNA expression was frequently high in many ASCL1 (SCLC-A) and POU2F3 (SCLC-P) expressing SCLC cell lines and tumors. At the proteomic level, BCL2 levels were highly correlated with BIM expression (rho = 0.68, P < 0.001) and modestly correlated with BCL-xL and phosphoBAD (rho > 0.3, P < 0.05; Fig. 1D). BCL2 also showed a bimodal distribution in the SCLC cell lines (bimodality index = 1.73; Supplementary Fig. S1B). In line with the role cMYC plays in repressing BCL2 expression (10), several SCLC cell lines in the BCL2-low group, which were also predominantly sensitive to AZD2811, expressed high cMYC protein levels (fold change = 2.8, P value by t test = 0.001; Fig. 1E). However, overall, we observed only a modest inverse correlation between cMYC and BCL2 expression, at both mRNA (rho = −0.32, P < 0.05) and protein levels (rho = −0.25, P < 0.05), among the SCLC cell lines, suggesting that high cMYC expression is not always associated with low BCL2 expression. In fact, a subset of SCLC cell lines as well as tumors expressed high levels of both cMYC/MYC and BCL2/BCL2 (Fig. 1D). Furthermore, these cMYC-high cell lines with high BCL2 levels were predominantly resistant to AZD2811, highlighting BCL2 as a strong predictor of resistance to AZD2811 in SCLC cell lines (P = 0.011 by Pearson χ² test; Fig. 1E; Supplementary Fig. S1C).

To determine whether BCL2 was indeed a causal mediator of resistance to AZD2811, we generated BCL2-overexpressing isogenic cell line pairs from AZD2811-sensitive SCLC cell lines, expressing low endogenous BCL2 expression and varying cMYC levels. In 96-hour cell viability assays, SCLC cell lines stably overexpressing BCL2 were consistently more resistant to AZD2811, compared with the isogenic vector-control cells, with a significant increase in IC₅₀ (P < 0.004 by paired t test; Fig 1F). Interestingly, a sensitive cell line (H1048), which endogenously expresses high levels of cMYC as well as BCL2, could be made more resistant to AZD2811 by further increasing BCL2 levels. Conversely, to determine whether AZD2811 resistance could be reversed by reducing BCL2 expression, we transiently silenced BCL2 in AZD2811-resistant SCLC cell lines that endogenously expressed high levels of BCL2. Compared with the non-targeting control siRNA, silencing of BCL2 was sufficient to significantly enhance sensitivity to AZD2811 in these cell lines (Supplementary Fig. S1D). Thus, these results show that high BCL2 expression predicts and mediates resistance to the AURKB inhibitor, AZD2811. We also found that high BCL2 levels in SCLC cell lines were associated with resistance to the AURKA inhibitor, alisertib (Supplementary Fig. S1E and S1F).

AURKB inhibition causes failed cytokinesis, endoreduplication, and polyploidy (32, 33). Consistent with this, flow cytometry analysis showed an increase in the 8N cell population following 48 hours treatment with AZD2811 (30 nmol/L), indicating severe mitotic defects and polyploidy, in the BCL2-low sensitive cell lines. On the other hand, AZD2811 induced polyploidy in some BCL2-high resistant cell lines while it had no effect on cell-cycle progression in others (Supplementary Fig. S2A). Because AURKB inhibition–induced polyploidy triggers apoptotic cell death (34, 35), we next assessed apoptosis by Annexin-V and propidium iodide staining in SCLC cell lines treated with AZD2811 (30 nmol/L) for 72 hours. As expected, there was a significant increase in apoptotic cell death following AZD2811-induced polyploidy in the sensitive cell lines. In contrast, apoptosis was consistently suppressed in the AZD2811-resistant BCL2-high SCLC cell lines (Fig. 2A; Supplementary Fig. S2B).

Overexpression of BCL2 similarly mitigated induction of apoptosis by AZD2811. In the sensitive isogenic vector-control cell lines, treatment with even a low dose of AZD2811 (30 nmol/L) strongly induced the activity of apoptosis initiator, caspase-9, albeit in a cell line–dependent manner. Apoptosis executor, caspase-3/7, activity was also similarly induced. AZD2811 treatment also significantly increased cleavage of PARP, a substrate of caspase-3 and marker of late apoptosis, in the vector-control cells (Fig. 2B and C; Supplementary Fig. S2C). Together, these data indicate activation of the apoptotic cascade following AZD2811 treatment in BCL2-low cell lines (Fig. 2B and C). In contrast, overexpression of BCL2 suppressed caspase activation and PARP cleavage in response to treatment with AZD2811 (Fig. 2B and C). Because polyploidy results in genetic instability, we examined whether AZD2811 treatment also increased DNA damage. In the sensitive vector-control cell lines, treatment with AZD2811 strongly induced accumulation of γH2AX, a marker of DNA damage (Fig. 2C). In contrast, endogenous as well as AZD2811-induced DNA damage was suppressed following BCL2 overexpression. Similarly, knockdown of BCL2 restored γH2AX accumulation and PARP cleavage in response to AZD2811 treatment in the resistant BCL2-high SCLC cell lines (Fig. 2D). An increase in cleaved caspase-3 levels in response to AZD2811 treatment was also observed following BCL2 knockdown. Previous studies in other cancer types have shown that BCL-xL inactivation had a sensitizing effect on AURKB inhibitors and other antimitotic agents (34, 36). However, in the resistant SCLC cell lines, BCL-xL knockdown did not significantly enhance the sensitivity of the cells to AZD2811 and had no effect on the apoptotic and DNA damage responses (Supplementary Fig. S2D and S2E). Together, these results demonstrate that intrinsic resistance to AURKB inhibition in SCLC is predominantly driven by high BCL2 levels, which inhibit the induction of apoptosis and DNA damage.

Next, we examined whether pharmacologic modulation of BCL2 could restore AZD2811 sensitivity. As shown above, BCL2-overexpressing isogenic cell lines were resistant to AZD2811. The single-agent activity of a selective BCL2 inhibitor, venetoclax, in these cells was also modest. However, the addition of venetoclax markedly sensitized these BCL2-overexpressing SCLC cells to AZD2811 (ΔAUC_BLISS score < −0.3; Fig. 3A; Supplementary Fig. S3A). Treatment of BCL2-overexpressing cell lines with venetoclax also restored AZD2811-induced apoptosis, indicated by increased PARP cleavage, and augmented DNA damage (γH2AX accumulation; Fig. 3B). On the other hand, the addition of venetoclax had no effect on the sensitivity of the isogenic vector-control cell lines (H446^Vec, H1876^Vec), expressing low BCL2 levels, to AZD2811 or on apoptosis. A similar effect was also observed in inherently AZD2811-resistant SCLC cell lines. The combination of AZD2811 with venetoclax significantly reduced cell viability, as compared with either drug as single agents (ΔAUC_BLISS score < −0.2; Fig. 3C; Supplementary Fig. S3A). Interestingly, in H1048, a BCL2-high cell line that also expresses high cMYC and sensitivity to AZD2811, the addition of venetoclax further improved the effect of AZD2811, highlighting that this drug combination could be effective in a BCL2-high background, irrespective of the single-agent sensitivity. In comparison with BCL2 inhibition, combination with a selective BCL-xL inhibitor (A-1331852) only modestly sensitized the resistant SCLC cell lines to AZD2811 (Supplementary Fig. S3B). Treatment with the combination of AZD2811 (30 nmol/L) and venetoclax (100 nmol/L) also significantly increased caspase-3/7 activity, cleavage of the caspase substrate PARP as well as DNA damage (Fig. 3D and E). Consistent with these observations, the fraction of cells positive for Annexin-V and propidium iodide, signaling the late apoptotic events of phosphatidylserine externalization and loss of membrane integrity, was dramatically increased at 72 hours after treatment with the drug combination (Fig. 3F; Supplementary Fig. S3C). Together, these findings demonstrate that the combination of AZD2811 with a selective BCL2 inhibitor, such as venetoclax, can overcome the intrinsic resistance and enhance apoptotic response to AURKB inhibition in SCLC cell lines.

We next assessed the in vivo efficacy of AZD2811NP in combination with venetoclax in a panel of 12 SCLC cell line–derived and PDX models that are representative of the different SCLC subtypes and express different levels of BCL2 (Fig 4A). These models were treated for a duration of 4 weeks with AZD2811NP monotherapy, venetoclax monotherapy, or their combination. As summarized in Fig. 4A, AZD2811NP induced stasis and tumor regression in 7 out of 12 models, whereas venetoclax monotherapy showed tumor regression in only one BCL2-high model (LC-F-22). In comparison, the combination of the two drugs demonstrated greatly improved efficacy over either single agent in 8 out of 12 models (Fig. 4A). In concordance with the in vitro data, H1048 xenograft model, with high cMYC and BCL2 expression, was sensitive to AZD2811NP in vivo and addition of venetoclax further improved the response (Fig. 4B). Both single agents AZD2811NP and venetoclax, and the combination thereof were well tolerated with no significant body weight loss in mice (Supplementary Fig. S4A). Similar results were also observed in the H69 xenograft model (Fig. 4C; Supplementary Fig. S4B). In four SCLC PDX models, with high BCL2 expression by RNA-seq, (SC101, SC96, LC-F-22, and SC61; Fig. 4A), the combination of AZD2811NP and venetoclax showed improved antitumor efficacy compared with either monotherapy (Fig. 4D–G). In addition, sustained tumor growth inhibition was observed even after cessation of treatment, most pronounced in the SC61 model where no tumor reemergence was seen up to 30 days after treatment in the combination group (Fig. 4E–G). In contrast, xenograft and PDX models with low BCL2 expression either did not benefit from the combination therapy (Supplementary Fig. S4C and S4D) or did not exhibit sustained tumor inhibition upon treatment cessation (Supplementary Fig. S4E). Next, to confirm target engagement in vivo as well as to investigate the effect on the apoptotic pathway, tumor samples from SC61 PDX models were analyzed 72 hours after treatment start. As expected, a reduction of pHH3 levels, indicative of effective AURKB inhibition, was observed in tumors treated with AZD2811NP and the combination therapy. Consistent with the in vitro data, AZD2811NP monotherapy induced apoptosis, indicated by increased PARP1 cleavage, which was further increased in the tumors treated with AZD2811NP and venetoclax combination (Fig. 4H).

Next, we also evaluated the optimal dosing and scheduling of an AZD2811NP and venetoclax combination to minimize potential toxicities while achieving sustained tumor inhibition. In the H211 xenograft model, venetoclax was administered on a 7, 5, 3, 2, or 1 day per week schedule in combination with AZD2811NP dosed once weekly (Fig. 5A). Of note, the nanoparticle formulation provides slow release of AZD2811 and sustained drug exposure (29). A continuous daily dosing of venetcolax (7 days per week) in combination with AZD2811NP induced tumor regression of 87% (Fig. 5A). In comparison, a reduced venetoclax dosing frequency of 5 days per week or only 3 days per week was still sufficient to achieve significant tumor regression (65% and 62%, respectively; Fig. 5A and B). Reducing concurrent venetoclax dosing further to only 1 or 2 days per week did not increase the efficacy of the combination beyond that of AZD2811 monotherapy (Fig. 5A). Interestingly, if the sequence of dosing was staggered and venetoclax was administered 2 days after AZD2811NP, then only two doses of venetoclax were sufficient to achieve the same level of tumor regression as that obtained with 5 days per week concurrent dosing of venetoclax with AZD2811NP, suggesting that the dose and schedule of this combination could be optimized to increase the therapeutic index (Fig. 5A and B). To further test this hypothesis, three SCLC PDX models were treated with AZD2811NP and venetoclax on a staggered 3 days a week dosing schedule (Fig. 5C). In addition, we also tested the combination of AZD2811NP with AZD0466 (a drug–dendrimer conjugate of the BCL2/xL dual inhibitor AZD4320; ref. 37). Administration of both venetoclax and AZD0466 was done one day after AZD2811NP dosing, as indicated (Fig. 5C). The combination of AZD2811NP with either venetoclax or AZD0466 induced similar tumor growth inhibition in all tested models and these values were comparable with those obtained with continuous dosing of venetoclax (Fig. 5C). Altogether, these data demonstrate that venetoclax dosing frequency or schedule allow for modifications, which could reduce potential clinical toxicities but without compromising the efficacy of the combination.

RB1 deficiency has been shown to disrupt microtubule dynamics and exacerbate mitotic abnormalities, resulting in an increased sensitivity to AURKB inhibition (26, 38). Therefore, AURKB has emerged as an attractive therapeutic target in SCLC which exhibits a near universal loss of function of RB1 as well as TP53 (3). The results of our study however show that despite near ubiquitous RB1 inactivation, additional factors influence response to AURKB inhibitors in SCLC. Using unbiased proteomic approaches, we identified BCL2 as a key driver of intrinsic resistance to AURKB inhibitors in SCLC cells.

The antiapoptotic BCL2 family proteins, including BCL2 and BCL-xL, have been shown to be critical for survival from mitotic catastrophe and apoptotic cell death (34). Polyploidization, such as induced by AURKB inhibitors, has also been shown to result in MCL1 reduction and to shift the antiapoptotic burden onto BCL-xL in colon carcinoma cells (34). Accordingly, synergistic interactions between AURKB inhibitors and BCL-xL inhibition have been demonstrated in colon carcinoma and other cancers (34, 36). Polyploidy induced by AURKB inhibition also triggers cell death through lethal autophagy, which is regulated by both BCL2 and BCL-xL (26, 39). In SCLC, BCL2 overexpression as well as BCL-xL amplification have been reported (4, 31), and BCL-xL levels are also modestly correlated with BCL2 expression. However, our results show that BCL2, rather than BCL-xL, plays a predominant role in protecting against AURKB-induced cell death in SCLC.

cMYC has previously been identified as a biomarker of sensitivity to AURK inhibitors in SCLC (6, 10, 15, 16, 18). A retrospective study showed improved clinical outcomes in a subset of patients with cMYC-high relapsed SCLC receiving the AURKA inhibitor alisertib and paclitaxel, as compared with chemotherapy alone (19). cMYC has also been shown to epigenetically repress BCL2, shifting the apoptotic dependency to MCL1 in SCLC (10). Given that AURKB inhibition also reduces MCL-1 stability (34), we found several cMYC-overexpressing SCLC cells were sensitive to AURKB inhibitors. Mechanistically, a synthetic lethal interaction between cMYC and AUKRB inhibition, mediated by apoptosis as well as lethal autophagy, has been reported previously (27). Consistent with this, we found that cMYC-high SCLC cell lines, with low BCL2 levels, that were sensitive to the AURKB inhibitor AZD2811 underwent robust apoptotic cell death. cMYC overexpression is also frequently seen in the NEUROD1-driven SCLC cell lines, which were previously reported to be sensitive to AZD2811 (17). However, amplification of MYC family genes including MYC is observed in only 20% of patients with SCLC (3, 5). In contrast, there is high prevalence of BCL2 expression in SCLC (30, 31). We show here that high BCL2 expression strongly predicts resistance to AURKA and AURKB inhibitors. Our analyses also revealed an appreciable number of SCLC cell lines and tumors with concurrent overexpression of both cMYC and BCL2. Cooperative interaction between the two, wherein high BCL2 expression mitigates the mitotic stress and apoptosis induced due to cMYC overexpression, has been described previously (40). Importantly, we found that a majority of these SCLC cell lines overexpressing both cMYC and BCL2 were resistant to AZD2811, thus suggesting that some of the cMYC-high SCLC could also benefit from the combined inhibition of AURKB and BCL2. Overall, we show that while cMYC-high/BCL2-low status predicted sensitivity, BCL2-high levels were associated with resistance to AURKB inhibition in SCLC, independent of cMYC.

AURKB inhibitors are in clinical trials in SCLC, other advanced solid tumors, and hematologic malignancies as monotherapy and in combination with other drugs (NCT03216343, NCT04830813, NCT01118611, NCT05505825). However, the limited activity in unselected patient populations and dose-limiting toxicity issues has hampered the clinical success of AURKB inhibitors (12). There is also the added concern that cancer cells surviving AURKB-induced polyploidy could result in increased tumor heterogeneity and therapeutic resistance (41). BCL2 paralogs have also been explored as potential targets in SCLC (42–44). A BCL-2/BCL-xL dual inhibitor (navitoclax) showed limited response as monotherapy in refractory metastatic SCLC with dose-dependent thrombocytopenia (45). On the other hand, the selective BCL2 inhibitor, venetoclax, FDA approved for treatment of relapsed chronic lymphocytic leukemia, is well tolerated and has minimal dose-limiting toxicities, often seen with BCL-xL inhibitors (46). Our data demonstrate that sustained SCLC tumor regression can be achieved even with intermittent dosing of AZD2811 and venetoclax, while offsetting potential dose-limiting toxicities. In conclusion, this study provides compelling evidence for the combination of AURKB and BCL2 inhibitors as a useful therapeutic strategy to overcome inherent resistance to AURKB inhibition in SCLC, especially in the BCL2-high patient population. As such, this combination could help in broadening the response to AURKB inhibitors to a larger subset of SCLC.

C.M. Della Corte reports other support from Amgen, AstraZeneca, Roche, MSD, and Merck outside the submitted work. U.M. Polanska reports other support from AstraZeneca outside the submitted work. C. Andersen is currently an employee of AstraZeneca. J. Saeh is an employee of AstraZeneca. J.E. Pease reports other support from AstraZeneca during the conduct of the study, as well as other support from AstraZeneca outside the submitted work. J. Travers reports other support from AstraZeneca during the conduct of the study. G. Fabbri reports other support from AstraZeneca outside the submitted work. C.M. Gay reports grants and personal fees from AstraZeneca, as well as personal fees from BeiGene, Bristol Myers Squibb, Daiichi Sankyo, G1 Therapeutics, Jazz Pharmaceuticals, Monte Rosa, and OncLive during the conduct of the study; in addition, C.M. Gay has a patent for Methods and Systems for Classification and Treatment of Small Cell Lung Cancer pending. J. Urosevic reports other support from AstraZeneca outside the submitted work. L.A. Byers reports other support from Merck Sharp & Dohme Corp., Arrowhead Pharmaceuticals, Chugai Pharmaceutical Co., AstraZeneca Pharmaceuticals, Genetech Inc., BeiGene, AbbVie, Jazz Pharmaceuticals, Puma Biotechnology, Amgen, and Daiichi Sankyo outside the submitted work. No disclosures were reported by the other authors.

K. Ramkumar: Conceptualization, investigation, visualization, writing–original draft, writing–review and editing. A. Tanimoto: Investigation. C.M. Della Corte: Investigation. C.A. Stewart: Investigation. Q. Wang: Formal analysis. L. Shen: Formal analysis. R.J. Cardnell: Writing–review and editing. J. Wang: Formal analysis. U.M. Polanska: Investigation. C. Andersen: Investigation. J. Saeh: Investigation. J.E. Pease: Investigation. J. Travers: Investigation. G. Fabbri: Resources, supervision, investigation. C.M. Gay: Supervision, writing–review and editing. J. Urosevic: Conceptualization, investigation, visualization, writing–original draft. L.A. Byers: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

This work was funded by grants from NCI UT Lung SPORE grant CA070907 (L.A. Byers, J. Wang), and also supported by grants from NIH R01 CA207295 (L.A. Byers), U01 CA213273 (L.A. Byers), U01 CA256780 (L.A. Byers), NIH CCSG P30-CA016672 (L.A. Byers, J. Wang), R50 CA243698 (C.A. Stewart), and the Lung Cancer Moonshot Program (L.A. Byers, J. Wang, C.M. Gay). The authors also wish to thank the U.T. MD Anderson Flow Cytometry and cellular imaging facility, which is supported in part by NIH through the MD Anderson Cancer Center support grant CA016672, for providing support with FACS analysis. AZD2811 and AZD2811NP was generously provided by AstraZeneca.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.