Elucidating molecularly stratified single agent, and combination, therapeutic strategies targeting MCL1 for lethal prostate cancer

Jiménez-Vacas, Juan M.; Westaby, Daniel; Figueiredo, Ines; De Haven Brandon, Alexis; Padilha, Ana; Yuan, Wei; Seed, George; Bogdan, Denisa; Gurel, Bora; Bertan, Claudia; Miranda, Susana; Lambros, Maryou; Montero-Hidalgo, Antonio J.; Coleman, Ilsa; Yu, Ivan Pak Lok; Buroni, Lorenzo; Zeng, Wanting; Neeb, Antje J.; Welti, Jon; Rekowski, Jan; Paravati, Roberta; Gabel, Florian; Pandell, Nicole; Ferreira, Ana; Crespo, Mateus; Riisnaes, Ruth; Das, Souvik; Taylor, Joe; Waldron, Nick; Hobern, Emily; Valenti, Melanie; Ning, Jian; Bernett, Ilona; Liodaki, Kate; Persse, Thomas; Galipeau, Patricia; Wilkinson, Scott; Trostel, Shana Y.; Karzai, Fatima; Chau, Cindy H.; Beatson, Erica L.; Zhang, Xiaohu; Klumpp-Thomas, Carleen; Varkaris, Andreas; Luque, Raul M.; Swain, Amanda; Raynaud, Florence; Lack, Nathan A.; Thomas, Craig J.; Ha, Gavin; Figg, William D.; Bezzi, Marco; Sowalsky, Adam G.; Nelson, Peter S.; Carreira, Suzanne; Balk, Steven P.; de Bono, Johann S.; Sharp, Adam

doi:10.1038/s41467-025-64042-5

Download PDF

Article
Open access
Published: 08 October 2025

Elucidating molecularly stratified single agent, and combination, therapeutic strategies targeting MCL1 for lethal prostate cancer

Nature Communications volume 16, Article number: 8806 (2025) Cite this article

14 Altmetric
Metrics details

Subjects

Abstract

Metastatic castration-resistant prostate cancer (mCRPC) is a lethal disease requiring additional therapeutic strategies. MCL1, an anti-apoptotic BCL2 family member, promotes cancer-cell survival, but its role in mCRPC remains poorly understood. Here, we characterise MCL1 in multiple mCRPC biopsy cohorts and patient-derived models, assessing responses to MCL1 inhibition. MCL1 copy number gain (14%–34%) correlates with increased MCL1 expression and worse outcomes. MCL1 inhibition exhibits anti-tumour effects in MCL1-gained mCRPC models. Co-inhibition of MCL1 and AKT induces cancer-specific cell death in PTEN-loss/PI3K-activated models in vitro and in vivo, modulating BAD-BCLXL and BIM-MCL1 interactions, with durable anti-tumour activity in models with AKT inhibitor acquired resistance. Finally, CDK9-mediated MCL1 downregulation combined with AKT inhibition recapitulates these findings, providing further opportunities for clinical translation. These data support early phase clinical trials targeting MCL1, both as monotherapy for MCL1-gained mCRPC, and in combination with AKT inhibition for PTEN-loss/PI3K-activated mCRPC.

Introduction

Despite significant progress in the treatment of metastatic castration-resistant prostate cancer (mCRPC), the disease remains lethal, with a median overall survival (OS) ranging from 2 to 3 years^1,2,3. Additional therapeutic strategies are urgently required to improve the outcome for patients with advanced disease. Cell death escape and circumvention of apoptosis is a hallmark of cancer, driving tumourigenesis and therapeutic resistance⁴. Intrinsic apoptosis is tightly regulated by a multitude of pro- and anti-apoptotic B-cell lymphoma 2 (BCL2) proteins, which interact on the mitochondrial outer membrane. Among the anti-apoptotic members, Myeloid Cell Leukaemia 1 (MCL1) has been shown to promote cancer cell survival in various malignancies, including prostate cancer (PC)^5,6. Moreover, MCL1 is one of the most frequently amplified genes in human malignancies, with amplification found in 10.9% of cancers across multiple tumour types⁷.

Beyond its anti-apoptotic function, it is postulated that MCL1 is involved in other key cancer cell processes such as cell cycle regulation, DNA repair, metabolism and calcium homoeostasis⁸. In addition, MCL1 may play a role in the development of resistance to androgen deprivation therapy (ADT) and DNA damage-inducing chemotherapies in PC^9,10, although further studies are required in this space. Moreover, MCL1 is thought to promote cell survival in therapy-induced senescent PC cells, a phenotype that contributes to tumour proliferation and metastasis via the Senescence-Associated Secretory Phenotype and has recently been reported to play a key role in endocrine treatment resistance^11,12. Taken together, MCL1 holds promise as a therapeutic target to induce apoptotic cell death, rather than growth inhibition in mCRPC, potentially mitigating the risk of acquired therapeutic resistance.

Importantly, small molecule BH3 mimetics have been developed to inhibit the anti-apoptotic BCL2 proteins, by binding to their hydrophobic groove⁵. Although developing high-affinity MCL1-specific BH3 mimetics has been challenging, there are now multiple agents under investigation in first-in-human clinical trials, with a focus on hematopoietic cancers. Despite these advances, preclinical studies have shown limited anti-tumour activity for single protein targeting, potentially due to functional redundancies among the anti-apoptotic BCL2 family members, including BCLXL^13,14,15. Indeed, it has been shown that PC is co-dependent on MCL1 and BCLXL and that both proteins must be inhibited to induce apoptosis^16,17, which suggests that targeting pathways regulating the expression or activity of BCLXL, in combination with an MCL1 inhibitor, may drive apoptotic cell death. Taken together, the identification of predictive biomarkers that enrich for response to single agent MCL1 inhibition in PC, and the discovery of rational MCL1 inhibitor combination strategies that drive apoptosis in PC, would address an urgent unmet clinical need.

Here, we show that MCL1 copy number gain and amplification occur early in lethal PC evolution and associate with worse clinical outcomes. We demonstrate that single agent MCL1 inhibition has marked anti-tumour activity in PC models with MCL1 copy number gain, and that MCL1 targeting (directly through BH3 mimetics or indirectly through CDK9 inhibition) synergises with AKT inhibition to deliver cancer-specific killing in PTEN-loss/PI3K-activated PC models, with ongoing anti-cancer activity in PC models with acquired AKT resistance. These data support the evaluation of therapeutic strategies targeting MCL1, as a single agent for PC with MCL1 copy number gains, and in combination with AKT inhibition in PTEN loss/PI3K-activated PC, within ‘proof of concept, proof of mechanism’ early phase clinical trials.

Results

MCL1 copy number gains are common in lethal prostate cancer, occur early in tumour evolution, and increase with castration-resistance

To investigate the genomic status of MCL1 in PC, we examined gene copy number alteration data from multiple PC biopsy cohorts. MCL1 is located on chromosome 1q21 and falls within a commonly amplified multi-gene cassette (Fig. 1A, B). MCL1 amplification (high level copy number gain) was rarely seen in primary PC from the The Cancer Genome Atlas (TCGA) cohort (1%), the majority of which are cured after radical treatment (Fig. 1A, B). In advanced CRPC biopsies from the Stand Up To Cancer/Prostate Cancer Foundation (SU2C/PCF) cohort, MCL1 amplification was more common (14%) (Fig. 1A, B). Analysis of matched same-patient CSPC and CRPC biopsies from the Royal Marsden Hospital (RMH) cohort showed a similar frequency of MCL1 amplification (CSPC 11.36%, CRPC 13.64%), suggesting this event occurs early in tumour evolution and associates with CRPC development (Fig. 1A, B). Despite this, a higher proportion of CRPC biopsies (34.09%) exhibited MCL1 copy number gain when compared to matched, same patient, CSPC biopsies (15.91%) (Fig. 1A, B), suggesting this might be part of an adaptive response emerging with treatment resistance. In addition, consistent with treatment resistance emergence, androgen receptor (AR) locus (Xq12) amplification was almost exclusively observed in CRPC biopsies (36%) when compared to CSPC biopsies from the same patient (Fig. 1A, B).

**Fig. 1: MCL1 (1q21) copy number gain/amplification is common in CRPC, occurs early in tumour evolution, and associates with poor prognosis.**

MCL1 copy number gains associate with worse clinical outcome

Having shown that genomic gains of MCL1 are common in mCRPC, we evaluated their association with prognostic markers and clinical outcome. In primary PC (TCGA cohort), MCL1 copy number gain/amplification was associated with significantly shorter progression-free interval (PFI) (HR 1.93, CI 1.05-3.53, p = 0.03) (Fig. 1C), a higher Gleason score and pathological tumour stage, and an increased chance of lymph node involvement at diagnosis (Fig. 1D). Furthermore, in a phase II trial evaluating 6 months of neoadjuvant ADT and enzalutamide prior to radical prostatectomy in men with localised intermediate- or high-risk PC¹⁸, MCL1 somatic copy number gains at baseline were associated with an inferior pathologic response to therapy. Four out of 37 tumours (11%) had MCL1 copy number gain/amplification, all of which were classified as pathologic incomplete or non-responders, providing evidence that gains in MCL1 may be associated with worse clinical outcome (Fig. 1E, F). In mCRPC, analysis of plasma cell-free DNA from patients enrolled in the FIRSTANA and PROSELICA clinical trials showed that subjects whose tumours have MCL1 copy number gain/amplification have shorter overall survival (HR 1.35, CI 0.95–1.90, p = 0.09) (Fig. 1G). MCL1 copy number gain/amplification was also associated with prior exposure to second-generation hormonal therapies (abiraterone and/or enzalutamide), whereas no significant differences were observed in terms of prior radical prostatectomy or radiotherapy (Supplementary Fig. 4A). MCL1 copy number gain/amplification did not, however, associate with taxane response highlighting its potential role as a prognostic, rather than predictive biomarker, in this specific context (Supplementary Fig. 4B). When considering prognostic markers in CRPC, MCL1 copy number gain/amplification associated with higher PSA and lower haemoglobin (Fig. 1H), but not with the other putative prognostic biomarkers analysed herein (Supplementary Fig. 4C–E). Overall, these data suggest that MCL1 copy number changes as PC progresses to mCRPC and is associated with worse clinical outcome.

MCL1 RNA and protein are highly expressed in mCRPC

Transcriptomic analysis revealed an association between MCL1 copy number gain/amplification and increased RNA expression in the TCGA and SU2C/PCF cohorts, as well as in the rapid autopsy University of Washington/Fred Hutchinson (UW/FH) CRPC cohort which also demonstrated highly concordant intra-patient, inter-tumour, MCL1 expression¹⁹ (Fig. 1I, J). The same trend was observed in the RMH cohort although not statistically significant (p = 0.301) (Fig. 1I). These data underline the functional significance of MCL1 copy number gain/amplification. However, a substantial number of tumours with unaltered MCL1 copy number exhibit high RNA expression, suggesting that additional mechanisms can also drive MCL1 overexpression (Fig. 1I, J). To investigate the correlation between MCL1 RNA and protein levels, Immunohistochemistry (IHC) was performed on 30 mCRPC biopsies from the RMH mCRPC cohort (Fig. 1K). MCL1 staining exhibited a cytoplasmic and granular pattern, which is consistent with its mitochondrial localisation (Fig. 1L). MCL1 RNA expression was found to be positively correlated with MCL1 protein cytoplasmic optical density values (n = 30, Spearman r = 0.482, p = 0.007) (Fig. 1K, L), suggesting that elevated RNA levels led to functionally relevant protein expression. While a range of expression levels was observed, none of the tumours exhibited loss of MCL1 protein or RNA (Supplementary Fig. 5A). Cytoplasmic MCL1 H-score was found to be correlated with MCL1 cytoplasmic optical density levels (Supplementary Fig. 5B). In addition, higher tumour MCL1 expressions (H-Score ≥100) was associated with shorter overall survival in CRPC patients (HR 2.09, 95% CI 0.85–5.12, p = 0.11) (Supplementary Fig. 6). These data confirm that MCL1 copy number changes associate with MCL1 RNA and protein, which are broadly expressed in advanced PC, further underlying its potential importance in PC biology.

Response to MCL1 inhibition is enriched in PC models with MCL1 copy number gain

Following the demonstration that MCL1 copy number was gained in a subset of mCRPC patients and linked to poor prognosis, we explored MCL1 targeting as a therapeutic strategy, hypothesising that MCL1 copy number gain would indicate dependency and enrich response to MCL1 inhibition, as suggested in other tumour types^7,20,21. We first characterised seven prostate-derived cell lines and found that 22Rv1 and DU145 harbour MCL1 copy number gain; however, only 22Rv1 exhibited markedly elevated MCL1 protein levels, suggesting additional mechanisms regulate MCL1 protein levels in DU145 (Fig. 2A–C). Among the tested cell lines, 22Rv1 showed the highest sensitivity to MCL1 inhibition, as evidenced by increased caspase 3/7 activity and reduced cell viability at 24 h and 6 days (Fig. 2D–F). LNCaP95 displayed a comparable response, indicating that sensitivity to MCL1 inhibition can be influenced by factors beyond basal MCL1 copy number or protein levels (Fig. 2D-F). The lack of response in DU145 may stem from its reported apoptosis resistance due to BAX deficiency²² (Fig. 2D–F). Next, we evaluated the apoptotic response to MCL1 inhibition of five CRPC PDX-O models with varying MCL1 copy numbers and protein levels (Fig. 2G–K). Both AZD5991 and S63845 increased caspase activity (6 h) and reduced organoid viability (at 96 h) in the MCL1 copy number gain models CP50c and CP267c, which exhibited high MCL1 protein expression (Fig. 2G–K and Supplementary Fig. 7). In contrast, models with unaltered MCL1 copy number and lower protein levels (CP142c, CP253c and CP336c) were minimally impacted (Fig. 2G–K and Supplementary Fig. 7). The effects of MCL1 inhibition demonstrated herein reflect short-term apoptosis; thus, the potential impact on PDX-O proliferation may require extended incubation times. These data suggest that, at least a subset of CRPC cells with high MCL1 levels may be sensitive to single agent MCL1 inhibition.

**Fig. 2: Prostate cancer models with MCL1 copy number gain are sensitive to MCL1 inhibition.**

MCL1 inhibition synergises with PI3K inhibition to trigger apoptosis in CRPC cells

In order to enhance and/or broaden the effect of MCL1 inhibition, we conducted a screen testing 166 FDA-approved drugs (at 5 µM), which included AR inhibitors (apalutamide, darolutamide and enzalutamide), as single agents and in combination with AZD5991 (1 µM). We evaluated the impact on apoptosis induction (6 h) and cell viability (24 h) in the C4-2 and LNCaP95 CRPC models, as well as in the immortalised prostate epithelium cell line PNT2. Specifically, PI3K pathway inhibitors (copanlisib, idelalisib, duvelisib and alpelisib) showed synergistic effects, decreasing cell viability, and inducing apoptosis in C4-2 and LNCaP95, but not in PNT2, suggesting that this combination selectively kills tumour cells (Fig. 3A, B and Supplementary Fig. 8). These results were validated using copanlisib and buparlisib at four different concentrations, revealing a dose-dependent induction of caspase 3/7 activity (6 h), accompanied by a dose-dependent decrease in cell viability (24 h) when combined with AZD5991 (Supplementary Fig. 9A, B). Taken together, these data demonstrate that MCL1 and PI3K pathway inhibitors synergise to induce cancer cell specific death in PC models.

**Fig. 3: PI3K/AKT pathway inhibition drives apoptosis when combined with MCL1 inhibition in LNCaP95 and C4-2 cell lines.**

MCL1 inhibition synergises with AKT inhibition to trigger apoptosis in CRPC cells

Given that AKT inhibition has shown anticancer activity in mCRPC clinical trials, we next tested whether direct inhibition of AKT could also elicit a robust synergistic effect when combined with MCL1 inhibition^23,24. The ZIP synergy scores obtained in response to the co-inhibition of AKT (capivasertib and ipatasertib) surpassed those observed with PI3K inhibitors (copanlisib and buparlisib) when combined with the MCL1 inhibitor AZD5991 in LNCaP95 and C4-2 cells (Fig. 3C, D and Supplementary Fig. 10). These results were further validated independently in an all vs all screen of 42 agents with 30 distinct mechanisms of action (861 different combinations), carried out at the National Cancer Institute (NCI). Specifically, the MCL1 inhibitor S63845 synergised with the PI3K inhibitors copanlisib and paxalisib, as well as with the AKT inhibitors capivasertib and ipatasertib to decrease cell viability at 48 h in the LNCaP95 cells (Supplementary Fig. 11A, B). The induction of apoptosis in response to the combination of AZD5991 with capivasertib and ipatasertib was further validated by protein analyses at 6 h, with cleavage of PARP, caspase 7 (determined by western blot) and caspase 3 (determined by both western blot and IHC) in LNCaP95 and C4-2 cells (Fig. 3E, F and Supplementary Fig. 12). These results were orthogonally validated using siRNAs to target AKT1/2/3 and MCL1, confirming “on-target” pharmacological mechanism of action (Supplementary Fig. 13). Overall, these findings demonstrate that, beyond the synergy observed with MCL1 and PI3K inhibition, the combined impact of MCL1 and AKT inhibition is more pronounced, leading to a heightened induction of apoptosis in PC models.

The synergistic anti-tumour activity of AKT and MCL1 inhibition is mediated through alterations in BAD-BCLXL and BIM-MCL1 interactions

Given the essential role of BH3-only proteins in the control of the intrinsic apoptosis pathway, through the regulation of the activity of anti-apoptotic BCL2 family members²⁵, we conducted a pro-apoptotic BCL2 family siRNA screen in C4-2 and LNCaP95 cells in response to capivasertib, ipatasertib, AZD5991, and their combinations. Specifically, individual silencing of the BH3-only proteins BIM and BAD, and the mitochondrial pore forming effector BAK partially prevented the induction of apoptosis (assessed by caspase 3/7 assay, and cleaved caspase 3 and cleaved PARP protein levels) and the reduction in cell viability (assessed by CellTiter-Glo) induced by the combined treatments in LNCaP95 and C4-2 cells (Supplementary Fig. 14 and Supplementary Fig. 15). When silencing LNCaP95 and C4-2 cells with a combination of BIM, BAD and BAK siRNAs, a near-complete rescue was achieved (Fig. 4A, B), suggesting that BIM, BAD and BAK are co-operative in the molecular mechanism underlying the synergistic effects driven by AKT and MCL1 co-inhibition. Furthermore, western blot analyses showed no significant changes in BAK, BAD, or BIM protein levels in response to capivasertib, ipatasertib, AZD5991, or their combinations (Fig. 4C). However, phosphorylation of BAD at Ser136 decreased with AKT inhibitors (capivasertib and ipatasertib) and combined treatments (Fig. 4C). Co-immunoprecipitation experiments revealed that ipatasertib increased BAD-BCLXL and BIM-MCL1 binding, with the latter being more evident when BIM was pulled down, in both LNCaP95 and C4-2 lines (Fig. 4D). Additionally, AZD5991 (as a single treatment or combined with ipatasertib) disrupted BIM-MCL1 and BAK-MCL1 binding (Fig. 4D), in keeping with its mechanism of action²⁶. These results were orthogonally validated by PLA in C4-2 cells (Supplementary Fig. 16). Taken together, these data suggest that the induction of apoptosis by MCL1 and AKT inhibition is mediated by BAD-driven sequestration of BCLXL and the displacement of BIM from MCL1.

**Fig. 4: AKT and MCL1 co-inhibition triggers apoptosis through dysregulation of BAD, BIM and BAK interactions in LNCaP95 and C4-2 cell lines.**

AKT and MCL1 co-inhibition triggers BAK-dependent apoptosis in PTEN-loss/PI3K-activated CRPC cells

To identify potential predictive biomarkers, we assessed phospho-BAD, total BAD, BIM and BAK levels and examined responses to AKT and MCL1 co-inhibition in seven cell lines with different molecular backgrounds (Fig. 5A, B and Supplementary Fig. 17A, B). Synergistic caspase 3/7 activation and reduced viability were observed in PTEN loss lines (LNCaP, C4-2, LNCaP95) and the PIK3CA-mutated line 22Rv1, which exhibited higher levels of phospho-BAD, total BAD, BIM, and BAK (Fig. 5A, B and Supplementary Fig. 17A, B). In contrast, no synergy was observed in PNT2, DU145 (both PTEN WT) nor in PC-3 despite harbouring PTEN loss (Fig. 5B and Supplementary Fig. 17A, B). Specifically, PC-3 cells have low levels of phospho-BAD, BIM and BAK, and additionally we showed that they rely on BAX rather than BAK (as opposed to the responder LNCaP95) to undergo apoptosis (Fig. 5B and Supplementary Fig. 17C, D). This suggests that undergoing apoptosis in a BAK-dependent manner is required for PC cells to be sensitive to AKT and MCL1 co-inhibition.

**Fig. 5: AKT and MCL1 co-inhibition triggers apoptosis in prostate cancer patient-derived xenograft organoids harbouring PI3K/AKT pathway hyperactivating aberrations.**

We next analysed CRPC PDX-Os with PI3K/AKT pathway aberrations, including CP50c (AKT1 amplification), which shows the highest PI3K-AKT pathway score among the SU2C/PCF samples²⁷, and CP253c, CP267c, and CP336c (PTEN loss by IHC), as well as the PTEN WT model CP142c (Supplementary Fig. 18 and 19). CP142c has significantly lower PI3K/AKT pathway activity than CP50c, CP253c, CP267c and CP336c, which was inferred by the Reactome PI3K/AKT signalling in cancer score and AKT downstream markers including phospho-GSK3B (Ser9) and phospho-PRAS40 (Thr246) (Fig. 5C and Supplementary Fig. 19). Among the five models, CP50c, CP253c, and CP142c were AR-FL and AR-V7 positive, with CP50c and CP253c also ERG positive, indicating TMPRSS2:ERG fusion²⁸ (Supplementary Fig. 19). Synergistic caspase 3/7 activation and reduced viability were observed in response to ipatasertib combined with MCL1 inhibitors (AZD5991 and S63845) in CP50c, CP253c, CP267c, and CP336c, all of which exhibited higher basal phospho-BAD levels than CP142, where no synergy was observed (Fig. 5D–G). In contrast, BAD, BIM, and BAK levels varied across models (Fig. 5D and Supplementary Fig. 19).

To confirm that PTEN loss is a key determinant of sensitivity to AKT and MCL1 co-inhibition, we treated PC Preclinical Mouse Modelling Platform organoids (ProMPt-Os) with different genomic backgrounds. Only PTEN loss models (PTEN^KO/TP53^WT/MYC^OE and PTEN^KO/TP53^KO/MYC^OE) responded synergistically to the combination of AKT and MCL1 inhibitors, as evidenced by a more pronounced increase in caspase 3/7 activity and reduced viability compared to single-agent treatments, while no synergy was observed in the PTEN WT model (PTEN^WT/TP53^KO/MYC^OE) (Fig. 6A–D). Notably, both PTEN loss models exhibited higher levels of phospho-BAD (Ser136), BAK, and BIM than the PTEN WT model (Fig. 6E). Basal phosphorylated BAD levels strongly correlated with the degree of response to AKT and MCL1 co-inhibition across all models, suggesting its potential as a predictive biomarker for sensitivity to this combination therapy (Fig. 6F, G). These data demonstrate that co-inhibition of AKT and MCL1 is synergistic and induces BAK-dependent cancer-specific killing in PTEN-loss/PI3K-activated PC models, supporting the clinical evaluation of this therapeutic strategy in a molecularly defined patient population.

**Fig. 6: AKT and MCL1 co-inhibition triggers apoptosis in PTEN-deficient mouse prostate cancer organoids.**

Synergistic inhibition of tumour growth by AKT and MCL1 co-inhibition in vivo

To evaluate AKT and MCL1 co-inhibition in vivo, we assessed CP253c, a PTEN loss, AR-FL/AR-V7 positive and ERG rearranged PDX model derived from a lymph node biopsy of an mCRPC patient previously treated with abiraterone (Supplementary Fig. 3 and Supplementary Fig. 19), in response to ipatasertib [50 mg/kg, p.o., 5 days/week (Monday–Friday)], S63845 [25 mg/kg, i.v., 3 days/week (Monday, Wednesday, and Friday)], combined treatment, or vehicle. The combined treatment led to a significant suppression of tumour growth, while single-agent treatments showed no statistically significant effect, indicating a synergistic interaction between the two drugs in vivo (Fig. 7A–C). No mice exhibited weight loss exceeding 20% of their baseline or showed any signs of discomfort or distress Supplementary Fig. 20). However, this should be taken with caution as S63845 (and other MCL1 inhibitors) show less affinity for mouse MCL1 than human MCL1²⁹. Histopathological examination of terminal tumours revealed a significant increase in cleaved caspase 3 in response to the combined treatment, and a decrease in KI67 with ipatasertib treatment, suggesting that the combined treatment inhibited the growth of CP253c by triggering apoptosis (Fig. 7D). Furthermore, downstream markers of the PI3K/AKT pathway (i.e., pGSK3B and pPRAS40) were downregulated in response to both ipatasertib single treatment and the combined treatment with S63845 (Fig. 7D). These data orthogonally validate our in vitro studies, confirming that AKT and MCL1 co-inhibition demonstrates marked anti-tumour activity in a PTEN loss patient-derived model in vivo, with associated pharmacodynamic biomarker modulation.

**Fig. 7: The combination of ipatasertib and S63845 has anti-tumour activity in the CRPC PDX CP253c in vivo.**

CRPC cells with acquired resistance to capivasertib remain sensitive to AKT and MCL1 co-inhibition, unlike those with acquired resistance to MCL1 inhibitors

To mimic acquired resistance to AKT inhibition seen in mCRPC patients during clinical trials^23,24, we generated capivasertib-resistant ‘Capi-R’ C4-2 and LNCaP95 cell lines. We first validated the models, demonstrating a >10-fold higher capivasertib IC₅₀ in Capi-R compared with parental cells (Fig. 8A, B). Both C4-2 and LNCaP95 Capi-R cells exhibited a slower growth rate compared to their parental counterparts (Fig. 8A, B). The IC₅₀ of capivasertib significantly decreased in the presence of AZD5991 in both LNCaP95 and C4-2 Capi-R cells (Fig. 8C, D), suggesting sensitivity to this combined treatment. Moreover, a synergistic increase in caspase 3/7 activity was observed in response to the combination of capivasertib and AZD5991 in LNCaP95 and C4-2 Capi-R cells (Fig. 8E, F). Furthermore, western blot analysis of cleaved PARP and caspase 3 orthogonally validated apoptosis induction (Fig. 8G, H). Apoptosis induction was more pronounced in LNCaP95 Capi-R compared to C4-2 Capi-R cells (Fig. 8E, F, G, H). BAK, BAD, and BIM were downregulated in Capi-R cells, with a more pronounced decrease in C4-2 compared to LNCaP95 Capi-R cells (Fig. 8I, J and Supplementary Fig. 21A, B), which may contribute to their reduced sensitivity to the combined treatment. Noteworthy, phosphorylation levels of BAD (Ser136) decreased with capivasertib in both C4-2 and LNCaP95 Capi-R cells (Fig. 8K, L and Supplementary Fig. 21C, D).

**Fig. 8: CRPC cells with acquired resistance to capivasertib remain sensitive to AKT and MCL1 co-inhibition.**

We next generated AZD5991- and S63845-resistant LNCaP95 cells and assessed their response (Supplementary Fig. 22A–C). While AKT inhibition alone continued to reduce cell viability, the combination of AKT and MCL1 inhibitors no longer increased caspase 3/7 activity or synergistically decreased viability (Supplementary Fig. 22D–F). The MCL1 inhibitor-resistant cell lines exhibited lower levels of phospho-BAD, BIM, and BAK compared to parental LNCaP95 cells (Supplementary Fig. 22G).

These results suggest that co-inhibition of AKT and MCL1 induces apoptosis, retaining efficacy in PC cells with acquired resistance to AKT inhibitors (as long as AKT inhibition continues to dephosphorylate BAD), while this synergy is lost in cells with acquired resistance to MCL1 inhibitors.

Targeting MCL1 through CDK9 inhibition recapitulates the phenotype of MCL1 inhibition in combination with AKT blockade in CRPC cells

CDK9 inhibition rapidly downregulates MCL1 protein levels in several tumour-types^30,31,32. We thus hypothesised that CDK9 inhibitors could also trigger acute apoptosis when combined with AKT inhibitors, in CRPC cells with PI3K/AKT hyperactivation. First, fadraciclib significantly decreased MCL1 protein levels after 6-h of incubation in LNCaP95 and C4-2 cells (Fig. 9A, B). The co-inhibition of AKT (using capivasertib and ipatasertib) and CDK9 (using fadraciclib and AZD4573), synergistically induced apoptosis and reduced cell viability (24 h) (Fig. 9A, B and Supplementary Fig. 23–24). Consistent with the mechanism underlying the synergistic effects of AKT and MCL1 co-inhibition, the caspase 3/7 activity induction and cell viability reduction observed in response to the combination of fadraciclib with AKT inhibitors, were prevented when cells were pre-treated with a combination of BIM, BAD, and BAK siRNAs (Fig. 9C). Furthermore, the response profile to AKT inhibitors and fadraciclib combinations mirrored that of AKT and MCL1 co-inhibition, with the same cell lines, PDX-Os and ProMPt-Os exhibiting sensitivity to both strategies (Fig. 9D–F and Supplementary Fig. 25). In vivo analysis of CP253c PDX showed that the combination of ipatasertib with fadraciclib significantly reduced tumour growth, whereas single treatments showed no significant impact (Fig. 10A). No mice exhibited a weight loss exceeding 20% of their baseline or showed any signs of discomfort or distress (Supplementary Fig. 26). Subsequent examination of terminal tumours by IHC revealed that cleaved caspase 3 induction was observed only in response to the combined treatment, while both single treatments decreased KI67, suggesting that single treatments impacted cell growth, whereas combined treatment triggered apoptosis in vivo (Fig. 10D). Additionally, downstream markers of AKT activity (pGSK3B and pPRAS40) were decreased in response to both ipatasertib alone and when combined with fadraciclib, and MCL1 protein levels were decreased by fadraciclib (Fig. 10D). Finally, while AKT inhibitors combined with fadraciclib continued to synergistically induce apoptosis and cell death in C4-2 and LNCaP95 Capi-R cells, this effect was no longer observed in LNCaP95 cells with acquired resistance to MCL1 inhibitors (Supplementary Fig. 27–28).

**Fig. 9: AKT and CDK9 co-inhibition recapitulates the effects of AKT and MCL1 co-inhibition in vitro.**

**Fig. 10: AKT and CDK9 co-inhibition recapitulates the effects of AKT and MCL1 co-inhibition in vivo.**

These data demonstrate that CDK9-mediated MCL1 downregulation, combined with AKT inhibition, replicated the in vitro and in vivo findings of direct MCL1 inhibition, providing further opportunities to clinically translate this therapeutic strategy.

Discussion

MCL1 supports cell survival by sequestering pro-apoptotic BCL2 proteins, preventing mitochondrial outer membrane permeabilization and the initiation of the intrinsic apoptosis pathway. Herein, we characterise MCL1 status at genomic, RNA and protein level in multiple independent mCRPC cohorts, including in matched same-patient CSPC and CRPC biopsies. We show that MCL1 amplification is a truncal event occurring early in lethal PC evolution, whilst MCL1 copy number gain occurs at diagnosis but also emerges with treatment resistance, probably representing an adaptive response. Interestingly, these genomic changes associate with worse clinical outcomes and inferior responses to neoadjuvant ADT and enzalutamide. Indeed, previous studies have shown MCL1 upregulation confers resistance to ADT in PC cells⁹, therefore we suggest that the selection of subclones harbouring MCL1 copy number gains or amplifications is a potential mechanism contributing to tumour progression and therapy resistance. However, no association was observed between MCL1 gain/amplification and taxane response, which could potentially be explained by the ability of taxanes to induce MCL1-independent mitotic catastrophe in these tumours^33,34. These data support the urgent development of molecularly stratified therapeutic strategies in MCL1 overexpressing mCRPC to improve the outcome for patients with this lethal disease.

We show that CRPC models with MCL1 copy number gain respond to MCL1-targeting BH3 mimetics, likely due to their increase reliance on MCL1 for survival, with inhibition inducing apoptosis, as previously observed in other tumour-types^7,20,21. A limitation of our study is the lack of in vivo validation testing the anti-tumour effects of MCL1 inhibition in models with MCL1 copy number gain. Despite this, we believe that our data support single-agent MCL1 inhibition as an effective anti-cancer strategy for CRPC with MCL1 copy number gains in a molecularly stratified approach. Previous studies have reported limited anti-tumour activity for single agent MCL1 inhibitors in PC models, but this is likely due to the selection of models with unaltered MCL1 copy number status as well as redundancy with other BCL2 family members including BCLXL^15,16. Therefore, the major challenge for drug development in this space, is to identify therapeutic strategies that can inhibit both MCL1 and BCLXL to drive cancer cell specific cell death, whilst maintaining tolerability.

Taking this important point into consideration, and to broaden the effect of MCL1 inhibition in CRPC without MCL1 copy number gains, we explored the efficacy of MCL1 inhibition in combination with FDA-approved oncological drugs with a view to discover synergistic combinations that could be rapidly translated to the clinic. This led to the discovery that MCL1 inhibition synergises with AKT blockade to induce BAK-dependent apoptosis by modulating BAD-BCLXL and BIM-MCL1 interactions. This therapeutic combination has been investigated in other tumour types, including myeloma³⁵, leukaemia³⁶, and breast cancer³⁷. However, our study highlights its therapeutic potential in CRPC and offers significant advancements in understanding the underlying molecular mechanisms. Specifically, our study demonstrates that AKT inhibition de-phosphorylates, and subsequently activates the BH3-only protein BAD, as previously reported³⁸. This activation leads to increased binding of BAD to BCLXL, inhibiting its anti-apoptotic function, and therefore triggering apoptosis when combined with MCL1 inhibition³⁹. We showed that AKT inhibition also increases BIM-MCL1 binding, likely in response to the displacement of BIM from BCLXL due to the competitive binding of BAD. This might contribute to increasing dependency on MCL1 in the context of AKT inhibition (and subsequent BCLXL blockade), following the direct activation model⁴⁰. Altogether, these mechanistic data are consistent with previous findings showing that PC models are co-dependent on MCL1 and BCLXL, with dual blockade needed to induce apoptosis¹⁶.

One potential challenge linked to MCL1 and BCLXL dual inhibition is the anticipated clinical toxicity^41,42. Critically, this is less likely to be the case for co-inhibition of MCL1 and AKT, as we observed cancer cell killing only in models that harbour PI3K/AKT pathway activating genomic aberrations (such as PTEN loss and AKT1 amplification), which is likely attributed to their significant dependency on phosphorylated BAD for survival^38,39,43. Indeed, our data suggest that high basal levels of phosphorylated BAD (Ser136) may be a more sensitive predictive biomarker for this combination therapy than genomic PI3K/AKT pathway aberrations (e.g., PTEN loss). Although single agent MCL1 demonstrated promising activity in PC with MCL1 copy number gains, this remains a small subset when compared with activation of the PI3K/AKT pathway, which accounts for around 50% of mCRPC, and is associated with worse clinical outcome^44,45. Taken together, these data demonstrate that co-inhibition of MCL1 and AKT deliver tumour specific killing in PTEN-loss/PI3K-activated lethal PC with high levels of phospho-BAD, providing a molecularly driven therapeutic strategy for a common subtype of lethal PC with poor prognosis⁴⁵.

AKT inhibitors have shown promise in clinical trials for patients with mCRPC, but have not yet met the threshold for regulatory approval^23,24. As such, the development of drug combination strategies to enhance anti-tumour activity or reverse resistance to AKT inhibition is an unmet clinical need. In this space, it should be noted that the co-inhibition of MCL1 and AKT not only drove apoptotic cell death in PC cells sensitive to AKT inhibition but also exerted anti-cancer activity in PC models that had acquired resistance to single agent AKT blockade. This is consistent with the reported sensitivity of AKT inhibition resistant breast cancer cell lines (developed by CRISPR-Cas9 KO techniques, as opposed to acquired resistance) to MCL1 blockade³⁷. Furthermore, we show that dual blockade of AKT and MCL1 is effective in mCRPC models characterised by biomarkers associated with resistance to AR signalling inhibitors and PC progression, including AR negativity, AR-V7 expression and TMPRSS2:ERG fusion^{46,47,48,49,50,51,52}. In conclusion, AKT and MCL1 co-inhibition demonstrates efficacy in overcoming resistance to AKT blockade and provides a targeted therapeutic strategy for common mCRPC lethal subtypes with potential for significant clinical impact. Importantly, we observed that cells with acquired resistance to MCL1 inhibitors no longer responded to dual AKT/MCL1 inhibition. This suggests that combining AKT and MCL1 blockade upfront, rather than after progression on MCL1 inhibitors, may be a more effective treatment strategy.

CDK9 inhibitors rapidly downregulate MCL1 protein levels and have shown anti-tumour activity with manageable toxicity in early-phase clinical trials for haematological malignancies^{30,31,32,53,54,55}. Here, we demonstrate that co-inhibition of CDK9 and AKT mirrors the effects of MCL1 and AKT co-inhibition, inducing cell death in PTEN-loss/PI3K-activated PC cells with high phospho-BAD levels. While CDK9 is a broad transcriptional regulator affecting multiple genes beyond MCL1, some studies suggest that MCL1 depletion is a key driver of the apoptosis-related anti-tumour effects of CDK9 inhibitors^30,56,57. Our data support that CDK9-dependent acute MCL1 downregulation contributes, at least in part, to the synergy observed with CDK9 and AKT co-inhibition. Specifically, the synergistic effects observed with both CDK9/AKT and MCL1/AKT co-inhibition strategies require the same BH3-only proteins (BIM, BAD, and BAK), and cells with acquired resistance to MCL1 inhibitors exhibit resistance to both combinations. However, we acknowledge that additional factors acutely regulated by CDK9 inhibition may also contribute to this synergy. Despite this, our findings support CDK9 inhibition combined with AKT blockade as an alternative, promising therapeutic strategy for PTEN-loss/PI3K-activated PC. It is important to acknowledge the challenges associated with direct MCL1 inhibition, particularly due to on-target cardiotoxicity^58,59. In this context, future strategies, including antibody-drug conjugates and proteolysis-targeting chimeras (PROTACs), similar to those developed for other BCL2 family members, could enable PC-specific MCL1 targeting while minimising the risk of cardiotoxicity^60,61.

Critically, our data support the urgent evaluation of therapeutic strategies targeting MCL1, directly or indirectly, as a single agent for PC harbouring MCL1 copy number gains, and in combination with AKT blockade in PTEN-loss/PI3K-activated lethal PC, within ‘proof of concept, proof of mechanism’ early phase clinical trials.

Methods

Patients and sample collection

Royal Marsden Hospital cohorts

All patients had mCRPC, were treated at the RMH, provided written informed consent, and were enroled in protocols approved by the RMH ethics review committee. For the matched same-patient cohort, treatment-naïve diagnostic samples were retrieved from referring hospitals. CRPC tissue was collected from mCRPC patients. Clinical data and demographics were collected retrospectively from the RMH electronic patient record system. Clinical characteristics and sample disposition are as previously described [RMH matched same-patient cohort⁶², RMH CRPC cohort⁶³]. For the RMH matched same-patient cohort⁶², 44 patients were included with copy number analysis data by low pass whole genome sequencing of tumour biopsy DNA. For the RMH CRPC cohort, 86/100 patients had RNA-sequencing data available and were included for this study.

FIRSTANA and PROSELICA cohort

All patients had mCRPC and participated in the FIRSTANA or PROSELICA prospective, randomised, open-label, international phase III trials. In FIRSTANA, patients were randomised to receive cabazitaxel (20 mg or 25 mg/m²) or docetaxel (75 mg/m²), all with prednisone (10 mg/day) as first-line chemotherapy for mCRPC⁶⁴. In PROSELICA, patients were randomised to receive either cabazitaxel 20 mg/m² or cabazitaxel 25 mg/m², each with prednisone 10 mg/day as second-line chemotherapy after docetaxel. All patients included in our study consented to allow for the collection of plasma. Plasma cell-free DNA low-pass whole-genome sequencing data were obtained from baseline samples. Samples with a tumour fraction >5% were assessed for genomic copy number burden (FIRSTANA n = 79, PROSELICA n = 73). OS was defined as the time from randomisation to death of any cause. Clinical characteristics are presented in Supplementary Table 1.

Stand Up To Cancer/Prostate Cancer Foundation cohort

Copy number variation (n = 336) and transcriptome (n = 159) data from mCRPC biopsies, generated by the International SU2C/PCF, were downloaded and re-analysed^45,65.

The Cancer Genome Atlas cohort

Clinical data and associated genomic data (n = 494) were downloaded from the TCGA-Clinical Data Resource (CDR) and Progression Free Interval (PFI) was used to evaluate outcome⁶⁶. PFI was defined as the time from the date of diagnosis until the date of the first occurrence of a new tumour event, which includes progression of the disease, locoregional recurrence, distant metastasis, new primary tumour, or death with tumour.

National Cancer Institute neoadjuvant cohort

For the NCI neoadjuvant cohort, tissue collection and analysis from patients with intermediate- and high-risk locally advanced PC treated with neoadjuvant ADT and enzalutamide prior to radical prostatectomy. This trial was approved by the Institutional Review Board of the Center for Cancer Research, NCI, Bethesda, MD (ClinicalTrials.gov identifier: NCT02430480).

University of Washington/Fred Hutchinson (UW/FH) cohort

For UW/FH cohort, tissue was obtained from patients who had died of mCRPC and had signed written informed consent for a rapid autopsy performed within 6 h of death, under the aegis of the PC Donor Programme at the University of Washington. All procedures involving human subjects in the rapid autopsy programme were approved by the Institutional Review Board of the University of Washington and of the Fred Hutchinson Cancer Center.

Bioinformatic analyses

Copy number analysis

Copy number variations data were processed and analysed as previously described in biopsies from RMH⁶², SU2C/PCF^45,65. In short, exome-sequencing reads were aligned to GRCh37 using BWA MEM. GATK realignment and recalibrate base scores were applied. Germline variance and somatic variance were called using GATK and Mutect2 by comparing tumour-normal paired comparison. ASCAT were used for copy number analysis. TCGA CNA segment data was download from cBioportal (n = 489). Copy number analysis for FIRSTANA/PROSELICA⁶⁴ cohorts were performed using low pass whole genome sequencing from plasma DNA. Sequencing reads were aligned to GRCh37 using BWA MEM. Aligned reads were quantified by HMMcopy readCounter, and copy number and the tumour fraction were calculated using ichorCNA. For the National Cancer Institute neoadjuvant cohort, exome sequencing of tumour foci and processing of data to derive gene-level somatic copy number estimates was described previously⁶⁷.

Transcriptome analysis

TCGA transcriptomic data were downloaded from cBioPortal^66,68. RMH and SU2C/PCF RNA sequencing dataset analysis (TopHat2 pipeline): CRPC transcriptomes generated by the SU2C/PCF Dream Team, were downloaded and reanalysed⁴⁵. Paired-end transcriptome sequencing reads for each of the SU2C/PCF (n = 106) and RMH (n = 86) cohorts, comprising samples with matched RNA and copy number data, were aligned to the human reference genome (GRCh37/hg19) using Tophat2 (v2.0.7)⁶³. Gene expression, Fragments Per Kilobase of transcript per Million mapped reads (FPKM), was calculated using Cufflinks. The top expressed genes (n = 15000) were analysed for each cohort respectively. The Spearman’s rank correlation coefficient (Spearman’s rho) between each gene’s expression (FPKM) and MCL1 expression (FPKM) was calculated, and subsequently used for pathway analysis. Pathway analysis was performed using the Gene Set Enrichment Analysis (GSEA) Pre-Ranked algorithm from GSEA software (v4.1.0). GSEA Pre-Ranked results were obtained using the H collection of Hallmark gene sets (MSigDB v7.0), with default parameters. UW/FH mCRPC RNA sequencing: RNA isolation and sequencing of mCRPC tumours were performed as described previously. Sequencing reads were mapped to the hg38 human genome using STAR.v2.7.3a. All subsequent analyses were performed in R. Gene level abundance was quantitated using GenomicAlignments and transformed to log2 FPKM using edgeR. Boxplots by patient were filtered to include only patients with at least two tumours profiled and matched copy number data (n = 115 tumours from 50 patients) and are ordered by per-patient median log2 FPKM gene expression. PI3K/AKT pathway score was calculated by accumulated z-score using genes of the REACTOME_PI3K_AKT_SIGNALING_IN_CANCER pathway obtained from the molecular signature database (MSigDB)⁶⁹.

Immunohistochemistry

IHC was performed on freshly sectioned tissue blocks where adequate material was present (≥100 tumour cells, reviewed by histopathologist B.G). BIM, BAD, and BAK antibodies were validated for specificity and optimised for IHC (Supplementary Figs. 1 and 2), while MCL1 antibody validation was reported previously⁷⁰. Details on the IHC assays are available in Supplementary Table 2. Protein quantification was determined for by a histopathologist (B.G) blinded to clinical data using H score method; [(% of absent staining x 0) + (% of weak staining x 1) + (% of moderate staining x 2) + (% of strong staining x 3)], resulting in a score between 0 and 300⁷¹. In addition, a pathologist-supervised machine learning algorithm (HALO AI, Indica Labs) was trained to calculate cytoplasmic MCL1 protein levels in patient samples. First, PC cells were differentiated from benign stroma. Colour deconvolution for DAB and haematoxylin staining was performed, and cell recognition and nuclear segmentation optimised. The analysis algorithm was then adjusted to provide continuous data on the intensity of MCL1 cytoplasmic signal intensity normalised per cell (accounting for cell size), averaged across the number of tumour cells identified per sample, providing a value between 0 (no staining) and 1 (black) called optical density (OD).

Fluorescence in situ hybridisation

Fluorescence In Situ Hybridisation (FISH) was performed in cell lines and patient-derived xenograft-organoids (PDX-Os) as previously described⁷². Briefly, fresh cell lines and organoids were resuspended in 10 mL of phosphate-buffered saline and treated with a fixative solution (FS) containing a 3:1 mixture of methanol and acetic acid (600 μL). The suspension was vigorously vortexed and subsequently centrifuged at 500 × g for 5-min. The supernatant was then carefully decanted, and the cell pellet was resuspended in 15 mL of FS. Following another round of centrifugation and removal of the supernatant, this process was repeated to ensure thorough fixation. A final 2 mL of FS was added to resuspend the cells. To facilitate optimal cell spreading for subsequent analysis, cells were released from a height of 20 cm onto positively charged slides. A dual-colour FISH assay was optimised using a cocktail of two FISH probes commercially available (ZytoLight SPEC MCL1/1p12 Dual Colour Probe, Z-2173-200, Bio SB). The MCL1/1p12 ratio ≥2.0 as previously reported²¹.

Drug compounds

AZD4573 and CYC065 (fadraciclib) were kind gifts from AstraZeneca and Cyclacel respectively. AZD5363 (capivasertib) and AZD5991 were (in part) a kind gift from AstraZeneca, with further stocks purchased from MedChemExpress. GDC-0068 (ipatasertib), ABT-263 (navitoclax), S63845, copanlisib and buparlisib were purchased from MedChemExpress. AZD5991 and S63845 specificity and selectivity are reported elsewhere^26,29. All drugs were prepared as 10 mM stock solution in DMSO and subsequently diluted in the appropriate media to achieve a final DMSO assay concentration of ≤0.1%.

Cell lines

PC cell lines (LNCaP, C4-2, 22Rv1, PC-3, and DU145) were purchased from American Type Culture Collection (ATCC) and normal-like prostate cell line PNT2 from Sigma-Aldrich. LNCaP95 cells were generously provided by Drs Alan K. Meeker and Jun Luo (Johns Hopkins University, Baltimore, MD). The cell culture conditions are depicted in Supplementary Table 3. Cell lines underwent regular Short Tandem Repeat (STR) profiling and mycoplasma testing.

Patient-derived xenograft (PDX) organoids (PDX-Os)

PDX-Os were generated from PDXs derived from human CRPC biopsies using methods described previously^27,73,74. The patient treatment histories are detailed in Supplementary Fig. 3. Organoids were formed in Matrigel® matrix (356231, Corning) domes in 24-well plates and then re-seeded into 96-well plates for drug experiments.

Prostate Cancer Preclinical Mouse Modeling Platform (ProMPt) organoids (ProMPt-Os)

The mouse organoids utilised in this study are part of the ProMPt. The mouse strains include PtenLoxP⁷⁵; p53LoxP (JAX stock: #008462)⁷⁶; MYCStopFL (JAX stock #020458)⁷⁷; and the UBC-Cre-ERT2 (JAX stock #007001)⁷⁸. Organoids were derived from mouse prostate tissue as previously described⁷⁹. Briefly, mouse prostate organoids were established from prostates isolated from 3-month-old mice. Luminal and basal cells from mouse prostate tissue were separated by double antibody staining for CD24 and CD49f markers. Sorted luminal cells were plated in 30 µl Matrigel drops on 6-well plates and covered with complete mouse prostate organoid medium (1x B27, 1.25 mM N-acetyl-l-cysteine, 50 ng/ml EGF, 200 nM A83-01, 100 ng/ml Noggin, 500 ng/ml R-spondin, 1 nM dihydrotestosterone, 10 µM Y-27632 dihydrochloride in adDMEM/F12 containing penicillin/streptomycin, 10 mM HEPES and 2 mM GlutaMAX). To induce Cre-ERT2-mediated excision of all LoxP exons and STOP cassettes, organoids were treated for two passages with 4-hydroxytamoxifen. Successful CRE-mediated excision was confirmed using quantitative PCR and western blot analysis.

Viability and apoptosis assays

PC cells, PDX-Os and ProMPt-Os were treated two days after seeding, or three days after small interfering RNA (siRNA) transfection, with BH3 mimetics, CDK9 inhibitors, or vehicle. PDX-Os and ProMPt-Os were cultured in full media without Y-27632 (ROCK inhibitor) for 24 h prior to treatment and maintained in the same media throughout the experiment. Cell line viability was evaluated 24 h post-treatment, and PDX-Os/ProMPt-Os viability at both 24 and 96 h after treatment using CellTiter-Glo 2D (G9241; Promega) and CellTiter-Glo3D (G9681; Promega), respectively. The in vitro caspase activity was evaluated 6 h post-treatment using the 2D Caspase-Glo 3/7 Luminescent assay (G8093; Promega) for cell lines and the 3D Caspase-Glo 3/7 Luminescent assay (G8983; Promega) for PDX-Os/ProMPt-Os, in accordance with the manufacturer’s instructions.

Food and drug administration-approved drug screen

The National Cancer Institute/National Institutes of Health (NCI/NIH) Food and Drug Administration (FDA)-approved anti-cancer Oncology Set was obtained from the NCI/NIH DTP Open Chemical Repository. The screen included 166 different compounds, delivered in DMSO at 10 mmol/L (Supplementary Table 4). C4-2, LNCaP95 and PNT2 cells were seeded in 384 well plates at 2000 cells per well. 48 h after seeding, cells were treated with the FDA approved compounds at 5 µM in absence and presence of AZD5991 at 1 µM with a final DMSO concentration in the wells ≤0.01%. Cell viability and apoptosis induction were assessed 24 h and 6 h post-treatment, using CellTiter-Glo 2D and Caspase-Glo 3/7 2D, respectively. The screen was performed as a single biological experiment (n = 1) without technical replicates.

High-throughput combination matrix drug screening

For pairwise drug-combination assessments in matrix format, compounds were acoustically dispensed into a 1536-well white solid bottom tissue culture-treated plate (EWB041000A, Aurora Microplates, Whitefish, MT, USA) with an Echo 550 acoustic liquid handler (Labcyte, San Jose, CA, USA) using previously described methods^80,81. A 9-point custom concentration range with 1:2 dilution between points was used for each drug-pair tested. Bortezomib (final concentration 20.3 µM) was used as a positive control for cell cytotoxicity. Cells were seeded into compound-containing plates at a density of 500 cells/well, in a final volume 5 µL of growth media by using a Multidrop Combi dispenser (Thermo Fisher Scientific). Plates were covered by a stainless-steel gasketed lid to prevent evaporation and incubated for 48 h in a humidified CO2 incubator. At the 48 h time point, 3 µL of Cell Titer Glo (Promega) was added to each well using a BioRAPTR® (Beckton Coulter, Brea, CA, USA) and plates were incubated at room temperature for 15 min with the stainless-steel lid in place. Luminescence readings were taken using a Viewlux reader (PerkinElmer) with a 2 s exposure time per plate. Viability of compound treated wells was normalised to DMSO and empty well controls present on each plate, and combination-response plotting was automatically performed for each individual drug+drug combination.

siRNA transfection

Cell lines were transfected with ON-TARGETplus siRNA (SMARTpool, mixture of 4 siRNAs) (Dharmacon, Horizon) using 0.2% Lipofectamine RNAiMAX Reagent (Thermo Fisher Scientific) as per manufacturer’s guidelines. For the pro-apoptotic BCL2 family siRNA screen, 3000–5000 cells were seeded in 96-well plates and treated with AZD5991 (1 µM) after 72 h of siRNA exposure. Cell viability (assessed with CellTiter-Glo 2D) and caspase activity (assessed with 2D Caspase-Glo 3/7) were determined 6- and 24 h post-treatment, respectively. Concentration of the siRNAs ranged from 25–75 nM and is specified on figure legends. Detailed information on the siRNAs used in this study is provided in Supplementary Table 5.

Western blot

Cells were plated in 6-well plates at a seeding density of 400,000 cells per well and allowed to attach and grow for 48 h. After a 6-h treatment period, cells were lysed and harvested using Pierce RIPA buffer containing a Pierce Protease and Phosphatase inhibitor tablet (Thermo Fisher Scientific). Following sonication (10 s), the protein lysates were centrifuged for 15-min at 4 °C. Protein concentrations were then determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). A total of 25 μg of protein were loaded on 4–12% NuPAGE Bis-Tris gels (Invitrogen), separated via electrophoresis and subsequently transferred (30 V, overnight) to Immobilon-P PVDF membranes (pore size 0.45 μm). The membranes were incubated with primary and secondary antibodies (Supplementary Table 6) in 5% milk in TBS supplemented with 0.1% Tween20 (Sigma-Aldrich). Chemiluminescence was detected using the ChemiDoc Touch Imaging System (Bio-Rad).

Co-immunoprecipitation

Cells were seeded in 100 mm dishes at a density of 6 million cells, with three dishes per experimental condition. Cells were treated 48 h after seeding. Following a 6-h treatment, cells were washed with cold PBS and lysed in HMKEN buffer (10 mM HEPES, pH 7.2, 5 mM MgCl2, 142 mM KCl, 2 mM EGTA, 0.2% IGEPAL® CA-630, protease, and phosphatase inhibitors). The lysate samples were triturated, and then preclearing was performed by incubating with 1:1 protein A:G Dynabeads (Life Technologies) on a rotator at 4 °C for 1-h. Primary antibodies [BCLXL (#2764, Cell Signaling), MCL1 (16225-1-AP, Proteintech) and BIM (#2933, Cell Signaling Technology)] at a concentration of 0.45 µg were added to the precleared lysates and incubated on a rotator at 4 °C overnight. Subsequently, samples were incubated with 1:1 protein A:G Dynabeads for 2 h at 4 °C. Beads underwent five washes in HMKEN buffer, were resuspended in 50 µL of HMKEN buffer supplemented with reducing and loading buffer. Samples were processed for western blotting.

Proximity ligation assay

Duolink® proximity ligation assay (PLA) was performed on formalin-fixed paraffin-embedded sections from C4-2 cells treated with ipatasertib (1 µM), AZD5991 (1 µM), combined treatment and control (DMSO). Antigen retrieval was performed on a Bond RX autostainer (Leica Biosystems) for 10-min with Bond ER2 solution (AR9640, Leica Biosystems). The following interactions were studied using the red fluorescence assay (DUO92008, Sigma-Aldrich): BAD (1:100, ab32445, abcam)-BCLXL (1:50, ab77571, abcam) and BIM (1:250, #2933, Cell Signaling Technology)-MCL1 (1:100, sc-69838, Santa Cruz). Antibody cocktails were incubated at room temperature for 1-h and the remaining protocol was performed according to manufacturer’s instructions (Sigma–Aldrich). Duolink® in situ mouse PLUS and rabbit MINUS probes were used. The omission of one antibody from an interaction was used as negative control. At the end of the procedure the slides were mounted using Duolink® In Situ Mounting Medium which contains DAPI for nuclear staining. Imaging was performed and PLA dots were counted using pathologist supervised HALO^TM image analysis software (Indica Labs).

In vivo analyses with the CP253C CRPC PDX model

All procedures involving mice were conducted in compliance with the Institute of Cancer Research guidelines and approved by the ICR Animal Welfare and Ethical Review Body, adhering to the UK Animals (Scientific Procedures) Act 1986. Animals were housed in a pathogen-free facility in ventilated cages, at 20–24 °C and 40–60% relative humidity, with a 12 h light/dark cycle. CP253c PDX tumours were implanted into the left flank of NSG mice (7 weeks old at implantation; n = 31). Sex was considered in the study design and only male mice were used in the experiment given the disease aetiology. Upon reaching a tumour size of 50–300 mm³, mice were randomly assigned to one of six treatment arms: arm 1 received the vehicle, arm 2 was administered ipatasertib (50 mg/kg), arm 3 received S63845 (25 mg/kg), arm 4 was treated with fadraciclib (40 mg/kg), arm 5 received a combination of ipatasertib (50 mg/kg) and S63845 (25 mg/kg), and arm 6 received a combination of ipatasertib (50 mg/kg) and fadraciclib (40 mg/kg). Ipatasertib and fadraciclib were reconstituted in a solution of 0.5% methylcellulose and 0.2% Tween 80 in water, while S63845 was reconstituted in a solution containing 25 mM HCl and 20% hydroxypropyl-beta-cyclodextrin in water. Ipatasertib and fadraciclib were administered orally 5 days/week (Monday–Friday), while S63845 was delivered intravenously 3 days/week (Monday, Wednesday, and Friday). Tumour measurements and body weight were taken every 2–4 days, grouped in 3-day intervals. Mice were monitored for potential signs of discomfort or distress, including changes in physical appearance (facial expression, piloerection, hunched posture), behaviour (reduced activity), and physiological signs (abnormal breathing or altered food/water intake). The experiment concluded either 24 days after the initial treatment or earlier if tumours approached 14 mm diameter. None of the tumours exceeded the approved maximal size/burden allowed by our license (17 mm diameter). Mice were euthanized 6-h after the final dose, and tumour samples were collected for IHC analysis.

Development of drug-resistant cell lines

LNCaP95 and C4-2 cells were gradually exposed to increasing concentrations of capivasertib to develop capivasertib-resistant (Capi-R) cells, while LNCaP95 cells were similarly treated with AZD5991 or S63845 to generate AZD5991- and S63845-resistant (AZD5991-R and S63845-R) cells. Parental cell lines were maintained in parallel with 0.01% DMSO as a vehicle control. Resistance was defined after 6 days of drug exposure: Capi-R cells showed a >10-fold increase in capivasertib IC50, while AZD5991-R and S63845-R cells showed no significant change in growth when treated with 1 µM AZD5991 or S63845, respectively. Mycoplasma testing and STR profiling confirmed cell line identity and integrity.

Statistical analyses

All statistical analyses were performed using R Statistical Software version 4.1.3 (R Core Team, Vienna, Austria) or GraphPad Prism version 10 (GraphPad Software Inc., San Diego, California, USA). OS and PFI analyses were performed using Kaplan-Meier methodology and the log-rank test. The Mann-Whitney U test was used to compare pathologic tumour response in the NCI neoadjuvant cohort and to compare RNA expression, between tumours with MCL1 copy number gain/amplification and those without. Correlation analyses were performed with Spearman’s rank correlation coefficient. The linear mixed effects model was employed to compare the growth rate between the different treatment arms, and the vehicle, in the CP253c in vivo experiment and was performed using the R package nlme. Estimates and p-values refer to the interaction term of treatment arm and time indicating the difference in tumour volume growth rate between each treatment arm and the vehicle arm. Zero Interaction Potency (ZIP) synergy score was calculated, and ZIP surface plots were downloaded, using SynergyFinder+ web application^82,83. Statistical comparisons of cell viability or caspase 3/7 activity between the control (or scramble) and treatments (or siRNAs) were conducted using one-way ANOVA. Post-hoc Tukey tests were employed for comparisons among all groups, whereas post-hoc Dunnett tests were used for comparisons against the control group. Mean and standard deviation were presented unless specified otherwise. A two-sided p value ≤ 0.05 was deemed to be statistically significant. All experiments were performed in biological triplicates (n = 3) unless stated otherwise.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Source Data files are provided with this paper. All raw data are available in the Source Data. Transcriptomic and copy number variation datasets used in this study have been previously published and made available^{19,45,62,65,66,67,68}. The datasets used in this study are publicly available from the following sources: The TCGA data can be accessed in cBioPortal. The SU2C/PCF cohort data are available in cBioPortal and on GitHub under the code prad_su2c_2019. The RMH cohort data are available in the European Nucleotide Archive under accession number PRJEB32038. Finally, the NCI neoadjuvant cohort data have been deposited in the Database of Genotypes and Phenotypes (dbGaP) under accession number phs001938.v2.p1. No new omics datasets were generated. Further reasonable data access requests can be submitted to the corresponding authors. Source data are provided with this paper.

References

Mehtala, J. et al. Overall survival and second primary malignancies in men with metastatic prostate cancer. PLoS One 15, e0227552 (2020).
Article CAS PubMed PubMed Central Google Scholar
Moreira, D. M. et al. Predicting time from metastasis to overall survival in castration-resistant prostate cancer: results from search. Clin. Genitourin. Cancer 15, 60–66.e2 (2017).
Article PubMed Google Scholar
Sumanasuriya S., De Bono J. Treatment of advanced prostate cancer-a review of current therapies and future promise. Cold Spring Harb. Perspect. Med. 8, a030635 (2018).
Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell 144, 646–674 (2011).
Article CAS PubMed Google Scholar
Westaby, D. et al. Targeting the intrinsic apoptosis pathway: a window of opportunity for prostate cancer. Cancers. 14, 51 (2021).
Carneiro, B. A. & El-Deiry, W. S. Targeting apoptosis in cancer therapy. Nat. Rev. Clin. Oncol. 17, 395–417 (2020).
Article PubMed PubMed Central Google Scholar
Beroukhim, R. et al. The landscape of somatic copy-number alteration across human cancers. Nature 463, 899–905 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar
Widden, H. & Placzek, W. J. The multiple mechanisms of MCL1 in the regulation of cell fate. Commun. Biol. 4, 1029 (2021).
Article CAS PubMed PubMed Central Google Scholar
Santer, F. R. et al. Mechanistic rationale for MCL1 inhibition during androgen deprivation therapy. Oncotarget 6, 6105–6122 (2015).
Article PubMed PubMed Central Google Scholar
Reiner, T. et al. Mcl-1 protects prostate cancer cells from cell death mediated by chemotherapy-induced DNA damage. Oncoscience 2, 703–715 (2015).
Article PubMed PubMed Central Google Scholar
Troiani, M. et al. Single-cell transcriptomics identifies Mcl-1 as a target for senolytic therapy in cancer. Nat. Commun. 13, 2177 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Guo, C. et al. Targeting myeloid chemotaxis to reverse prostate cancer therapy resistance. Nature 623, 1053–1061 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Wei, G. et al. Chemical genomics identifies small-molecule MCL1 repressors and BCL-xL as a predictor of MCL1 dependency. Cancer Cell 21, 547–562 (2012).
Article CAS PubMed PubMed Central Google Scholar
Goldsmith, K. C. et al. Mitochondrial Bcl-2 family dynamics define therapy response and resistance in neuroblastoma. Cancer Res. 72, 2565–2577 (2012).
Article CAS PubMed PubMed Central Google Scholar
Pilling, A. B. & Hwang, C. Targeting prosurvival BCL2 signaling through Akt blockade sensitizes castration-resistant prostate cancer cells to enzalutamide. Prostate 79, 1347–1359 (2019).
Article CAS PubMed PubMed Central Google Scholar
Arai, S. et al. Tyrosine Kinase inhibitors increase MCL1 degradation and in combination with BCLXL/BCL2 inhibitors drive prostate cancer apoptosis. Clin. Cancer Res. 24, 5458–5470 (2018).
Article CAS PubMed PubMed Central Google Scholar
Brim, B. C. et al. Direct Co-Targeting of Bcl-xL and Mcl-1 Exhibits Synergistic Effects in AR-V7-Expressing CRPC Models. Cancer Res Commun. 5, 1396–1408 (2025).
Karzai, F. et al. Sequential prostate magnetic resonance imaging in newly diagnosed high-risk prostate cancer treated with neoadjuvant enzalutamide is predictive of therapeutic response. Clin. Cancer Res. 27, 429–437 (2021).
Article CAS PubMed Google Scholar
Kumar, A. et al. Substantial interindividual and limited intraindividual genomic diversity among tumors from men with metastatic prostate cancer. Nat. Med. 22, 369–378 (2016).
Article CAS PubMed PubMed Central Google Scholar
Slomp, A. et al. Multiple myeloma with 1q21 amplification is highly sensitive to MCL-1 targeting. Blood Adv. 3, 4202–4214 (2019).
Article CAS PubMed PubMed Central Google Scholar
Munkhbaatar, E. et al. MCL-1 gains occur with high frequency in lung adenocarcinoma and can be targeted therapeutically. Nat. Commun. 11, 4527 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Chandra, D., Choy, G., Daniel, P. T. & Tang, D. G. Bax-dependent regulation of Bak by voltage-dependent anion channel 2. J. Biol. Chem. 280, 19051–19061 (2005).
Article CAS PubMed Google Scholar
de Bono, J. S. et al. Randomized Phase II study evaluating akt blockade with ipatasertib, in combination with abiraterone, in patients with metastatic prostate cancer with and without PTEN Loss. Clin. Cancer Res. 25, 928–936 (2019).
Article PubMed Google Scholar
Sweeney, C. et al. Ipatasertib plus abiraterone and prednisolone in metastatic castration-resistant prostate cancer (IPATential150): a multicentre, randomised, double-blind, phase 3 trial. Lancet 398, 131–142 (2021).
Article CAS PubMed Google Scholar
Kale, J., Osterlund, E. J. & Andrews, D. W. BCL-2 family proteins: changing partners in the dance towards death. Cell Death Differ. 25, 65–80 (2018).
Article CAS PubMed Google Scholar
Tron, A. E. et al. Discovery of Mcl-1-specific inhibitor AZD5991 and preclinical activity in multiple myeloma and acute myeloid leukemia. Nat. Commun. 9, 5341 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Welti, J. et al. Targeting bromodomain and extra-terminal (BET) family proteins in castration-resistant prostate cancer (CRPC). Clin. Cancer Res. 24, 3149–3162 (2018).
Article CAS PubMed Google Scholar
Park, K. et al. Antibody-based detection of ERG rearrangement-positive prostate cancer. Neoplasia 12, 590–598 (2010).
Article CAS PubMed PubMed Central Google Scholar
Kotschy, A. et al. The MCL1 inhibitor S63845 is tolerable and effective in diverse cancer models. Nature 538, 477–482 (2016).
Article ADS PubMed Google Scholar
Cidado, J. et al. AZD4573 Is a highly selective CDK9 inhibitor that suppresses MCL-1 and induces apoptosis in hematologic cancer cells. Clin. Cancer Res. 26, 922–934 (2020).
Article CAS PubMed Google Scholar
Frame, S. et al. Fadraciclib (CYC065), a novel CDK inhibitor, targets key pro-survival and oncogenic pathways in cancer. PLoS One 15, e0234103 (2020).
Article CAS PubMed PubMed Central Google Scholar
Dey, J. et al. Voruciclib, a clinical stage oral CDK9 inhibitor, represses MCL-1 and sensitizes high-risk diffuse Large B-cell Lymphoma to BCL2 inhibition. Sci. Rep. 7, 18007 (2017).
Article ADS PubMed PubMed Central Google Scholar
Fabbri, F. et al. Mitotic catastrophe and apoptosis induced by docetaxel in hormone-refractory prostate cancer cells. J. Cell. Physiol. 217, 494–501 (2008).
Article CAS PubMed Google Scholar
Wang, Y. & Poon, R. Y. C. MARCH5 regulates mitotic apoptosis through MCL1-dependent and independent mechanisms. Cell. Death. Differ. 30, 753–765 (2023).
Article CAS PubMed Google Scholar
Bloedjes, T. A. et al. AKT signaling restrains tumor suppressive functions of FOXO transcription factors and GSK3 kinase in multiple myeloma. Blood. Adv. 4, 4151–4164 (2020).
Article CAS PubMed PubMed Central Google Scholar
Li, Y. et al. AKT inhibition sensitizes acute leukemia cells to S63845-induced apoptosis. Hematology 28, 2214465 (2023).
Article PubMed Google Scholar
Dunn, S. et al. AKT-mTORC1 reactivation is the dominant resistance driver for PI3Kbeta/AKT inhibitors in PTEN-null breast cancer and can be overcome by combining with Mcl-1 inhibitors. Oncogene 41, 5046–5060 (2022).
Article CAS PubMed PubMed Central Google Scholar
Datta, S. R. et al. Akt phosphorylation of BAD couples survival signals to the cell-intrinsic death machinery. Cell 91, 231–241 (1997).
Article CAS PubMed Google Scholar
Zha, J., Harada, H., Yang, E., Jockel, J. & Korsmeyer, S. J. Serine phosphorylation of death agonist BAD in response to survival factor results in binding to 14-3-3 not BCL-X(L). Cell 87, 619–628 (1996).
Article CAS PubMed Google Scholar
Kim, H. et al. Hierarchical regulation of mitochondrion-dependent apoptosis by BCL-2 subfamilies. Nat. Cell Biol. 8, 1348–1358 (2006).
Article CAS PubMed Google Scholar
Mason, K. D. et al. Programmed anuclear cell death delimits platelet life span. Cell 128, 1173–1186 (2007).
Article CAS PubMed Google Scholar
Weeden, C. E. et al. Dual inhibition of BCL-XL and MCL-1 is required to induce tumour regression in lung squamous cell carcinomas sensitive to FGFR inhibition. Oncogene 37, 4475–4488 (2018).
Article CAS PubMed Google Scholar
Yang, E. et al. Bad, a heterodimeric partner for Bcl-XL and Bcl-2, displaces Bax and promotes cell death. Cell 80, 285–291 (1995).
Article CAS PubMed Google Scholar
Ferraldeschi, R. et al. PTEN protein loss and clinical outcome from castration-resistant prostate cancer treated with abiraterone acetate. Eur. Urol. 67, 795–802 (2015).
Article CAS PubMed PubMed Central Google Scholar
Abida, W. et al. Genomic correlates of clinical outcome in advanced prostate cancer. Proc. Natl. Acad. Sci. USA 116, 11428–11436 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Carver, B. S. et al. Aberrant ERG expression cooperates with loss of PTEN to promote cancer progression in the prostate. Nat. Genet. 41, 619–624 (2009).
Article CAS PubMed PubMed Central Google Scholar
King, J. C. et al. Cooperativity of TMPRSS2-ERG with PI3-kinase pathway activation in prostate oncogenesis. Nat. Genet. 41, 524–526 (2009).
Article CAS PubMed PubMed Central Google Scholar
Sharp, A. et al. Androgen receptor splice variant-7 expression emerges with castration resistance in prostate cancer. J. Clin. Investig. 129, 192–208 (2019).
Article PubMed Google Scholar
Attard, G. et al. Characterization of ERG, AR and PTEN gene status in circulating tumor cells from patients with castration-resistant prostate cancer. Cancer Res. 69, 2912–2918 (2009).
Article CAS PubMed Google Scholar
Yu, J. et al. An integrated network of androgen receptor, polycomb, and TMPRSS2-ERG gene fusions in prostate cancer progression. Cancer Cell 17, 443–454 (2010).
Article CAS PubMed PubMed Central Google Scholar
Wang, H. T. et al. Neuroendocrine Prostate Cancer (NEPC) progressing from conventional prostatic adenocarcinoma: factors associated with time to development of NEPC and survival from NEPC diagnosis-a systematic review and pooled analysis. J. Clin. Oncol. 32, 3383–3390 (2014).
Article PubMed Google Scholar
Vlachostergios, P. J., Puca, L. & Beltran, H. Emerging variants of castration-resistant prostate cancer. Curr. Oncol. Rep. 19, 32 (2017).
Article PubMed PubMed Central Google Scholar
Piha-Paul, S. et al. A phase 1/2, open-label, multicenter study to investigate the safety, pharmacokinetics, and efficacy of fadraciclib (CYC065), an oral CDK2/9 inhibitor, in subjects with advanced solid tumors and lymphoma. Eur. J. Cancer 174, S17–S18 (2022).
Article Google Scholar
Strati, P. et al. P1126: phase 1B/2A study of AZD4573 (CDK9i) and acalabrutinib in patients (PTS) with relapsed/refractory diffuse large B-cell lymphoma (R/R DLBCL). Hemasphere 7, e47925e9 (Suppl) (2023).
Do, K. T. et al. Abstract CT037: phase I safety, pharmacokinetic and pharmacodynamic study of CYC065, a cyclin dependent kinase inhibitor, in patients with advanced cancers (NCT02552953). Cancer Res. 78, CT037–CT037 (2018).
Article Google Scholar
Baker, A. et al. The CDK9 Inhibitor dinaciclib exerts potent apoptotic and antitumor effects in preclinical models of Mll-rearranged acute myeloid leukemia. Cancer Res. 76, 1158–1169 (2016).
Article CAS PubMed Google Scholar
Polier, G. et al. Wogonin and related natural flavones are inhibitors of CDK9 that induce apoptosis in cancer cells by transcriptional suppression of Mcl-1. Cell Death Dis. 2, e182 (2011).
Article CAS PubMed PubMed Central Google Scholar
Wang, X. et al. Deletion of MCL-1 causes lethal cardiac failure and mitochondrial dysfunction. Genes Dev. 27, 1351–1364 (2013).
Article CAS PubMed PubMed Central Google Scholar
Desai, P. et al. A Phase 1 First-in-human study of the Mcl-1 inhibitor Azd5991 in patients with relapsed/refractory hematologic malignancies. Clin. Cancer Res. 30, 4844–4855 (2024).
Tolcher, A. W. et al. A first-in-human study of mirzotamab clezutoclax as monotherapy and in combination with taxane therapy in relapsed/refractory solid tumors: dose escalation results. J. Clin. Oncol. 39, 3015 (2021).
Article Google Scholar
Khan, S. et al. A selective BCL-X(L) PROTAC degrader achieves safe and potent antitumor activity. Nat. Med. 25, 1938–1947 (2019).
Article CAS PubMed PubMed Central Google Scholar
Mateo, J. et al. Genomics of lethal prostate cancer at diagnosis and castration resistance. J. Clin. Investig. 130, 1743–1751 (2020).
Article CAS PubMed PubMed Central Google Scholar
Fenor de la Maza, M. D. et al. Immune biomarkers in metastatic castration-resistant prostate cancer. Eur. Urol. Oncol. 5, 659–667 (2022).
Sumanasuriya, S. et al. Elucidating prostate cancer behaviour during treatment via low-pass whole-genome sequencing of circulating tumour DNA. Eur. Urol. 80, 243–253 (2021).
Article CAS PubMed PubMed Central Google Scholar
Robinson, D. et al. Integrative clinical genomics of advanced prostate cancer. Cell 161, 1215–1228 (2015).
Article CAS PubMed PubMed Central Google Scholar
Liu, J. et al. An Integrated TCGA pan-cancer clinical data resource to drive high-quality survival outcome analytics. Cell 173, 400–416.e11 (2018).
Article CAS PubMed PubMed Central Google Scholar
Wilkinson, S. et al. Nascent prostate cancer heterogeneity drives evolution and resistance to intense hormonal therapy. Eur. Urol. 80, 746–757 (2021).
Article CAS PubMed PubMed Central Google Scholar
Cerami, E. et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2, 401–404 (2012).
Article PubMed Google Scholar
Milacic, M. et al. The reactome pathway knowledgebase 2024. Nucleic Acids Res. 52, D672–D678 (2024).
Article CAS PubMed Google Scholar
Westaby, D. et al. BCL2 expression is enriched in advanced prostate cancer with features of lineage plasticity. J. Clin. Investig. 134, e179998 (2024).
Detre, S., Saclani Jotti, G. & Dowsett, M. A “quickscore” method for immunohistochemical semiquantitation: validation for oestrogen receptor in breast carcinomas. J. Clin. Pathol. 48, 876–878 (1995).
Article CAS PubMed PubMed Central Google Scholar
Reid, A. H. et al. Molecular characterisation of ERG, ETV1 and PTEN gene loci identifies patients at low and high risk of death from prostate cancer. Br. J. Cancer 102, 678–684 (2010).
Article CAS PubMed PubMed Central Google Scholar
Welti, J. et al. Targeting the p300/CBP Axis in lethal prostate cancer. Cancer Discov. 11, 1118–1137 (2021).
Article CAS PubMed PubMed Central Google Scholar
Gil, V. et al. HER3 Is an actionable target in advanced prostate cancer. Cancer Res 81, 6207–6218 (2021).
Article CAS PubMed PubMed Central Google Scholar
Chen, Z. et al. Crucial role of p53-dependent cellular senescence in suppression of Pten-deficient tumorigenesis. Nature 436, 725–730 (2005).
Article ADS CAS PubMed PubMed Central Google Scholar
Marino, S., Vooijs, M., van Der Gulden, H., Jonkers, J. & Berns, A. Induction of medulloblastomas in p53-null mutant mice by somatic inactivation of Rb in the external granular layer cells of the cerebellum. Genes Dev. 14, 994–1004 (2000).
Article CAS PubMed PubMed Central Google Scholar
Calado, D. P. et al. The cell-cycle regulator c-Myc is essential for the formation and maintenance of germinal centers. Nat. Immunol. 13, 1092–1100 (2012).
Article CAS PubMed PubMed Central Google Scholar
Ruzankina, Y. et al. Deletion of the developmentally essential gene ATR in adult mice leads to age-related phenotypes and stem cell loss. Cell Stem Cell 1, 113–126 (2007).
Article CAS PubMed PubMed Central Google Scholar
Drost, J. et al. Organoid culture systems for prostate epithelial and cancer tissue. Nat. Protoc. 11, 347–358 (2016).
Article CAS PubMed PubMed Central Google Scholar
Mathews Griner, L. A. et al. High-throughput combinatorial screening identifies drugs that cooperate with ibrutinib to kill activated B-cell-like diffuse large B-cell lymphoma cells. Proc. Natl. Acad. Sci. USA 111, 2349–2354 (2014).
Article ADS CAS PubMed PubMed Central Google Scholar
Lin, G. L. et al. Therapeutic strategies for diffuse midline glioma from high-throughput combination drug screening. Sci. Transl. Med. 11, eaaw0064 (2019).
Zheng, S. et al. SynergyFinder Plus: toward better interpretation and annotation of drug combination screening datasets. Genom. Proteom. Bioinforma. 20, 587–596 (2022).
Article Google Scholar
Yadav, B., Wennerberg, K., Aittokallio, T. & Tang, J. Searching for drug synergy in complex dose-response landscapes using an interaction potency model. Comput. Struct. Biotechnol. J. 13, 504–513 (2015).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

The authors gratefully acknowledge the patients and the families of patients who contributed to this study. This work was supported by Prostate Cancer UK (TLD-PF19-006 to J.M.J.-V.; Career Acceleration Fellowship to J.T.; Research Funding to J.S.dB.), the Department of Defence Prostate Cancer Research Programme (Early Investigator Research Award to J.M.J-V. and S.W.; Impact Awards to A.G.S. and S.P.B.; Idea Development Award to A.G.S.), Cancer Research UK (Clinical Research Training Fellowship to D.W. and N.W.; Radiation Research Network Seed Funding to A.Sharp; Convergence Centre Grant funding, Centre Programme and Experimental Cancer Medicine Centre grants to J.S.dB.), MINECO (FPU18/02485 to A.J.M-H.; PID2022-138185OB-I00 to R.M.L), National Cancer Institute/National Institutes of Health (R50CA274336, P50CA0907186, P01CA163227, and R21CA277368 to I.C. and P.S.N.; the Intramural Research Programme of the Center for Cancer Research, Research Funding to W.D.), Royal Society of Medicine (International Exchanges Award IES\R3\213131 to A.Sharp and N.A.L.), Terry Fox Research Institute New Frontiers Programme (F22-00589 to N.A.L.), Prostate Cancer Foundation Canada grant-in-aide (to N.A.L.), American Association for Cancer Research (AACR Clinical/Translational Cancer Research Fellowship 18-40-11-BEZZ to M.B.), Medical Research Council (Research Funding to J.S.d.B.), the Prostate Cancer Foundation (Young Investigator Award to S.W.; Challenge Award to S.P.B., J.S.dB. and A.S.), the Movember Foundation through the London Movember Centre of Excellence (CEO13 2-002 to J.S.d.B.), the National Institute for Health and Care Research Biomedical Research Centre (pump priming award to A.Sharp), the Wellcome Trust (Clinical Research Career Development Fellowship to A.Sharp). Portions of this work was supported by the NCI (HHSN261200800001E), the Intramural Research Programme of the NCI, NIH, and the NIH National Center for Advancing Translational Sciences. Portions of this work utilised the computational resources of the NIH HPC Biowulf cluster. The authors thank the patients, families and investigators, and the Institute for Prostate Cancer Research for supporting the University of Washington Tissue Acquisition Programme. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services or the Department of Defence, nor does mention of trade names, commercial products, or organisation imply endorsement by the U.S. Government. J.S.d.B. is a National Institute for Health Research (NIHR) Senior Investigator. The views expressed in this article are those of the author(s) and not necessarily those of the National Health Service, the NIHR, or the Department of Health, nor does mention of trade names, commercial products, or organisation imply endorsement by the UK Government.

Author information

These authors contributed equally: Juan M. Jiménez-Vacas, Daniel Westaby.
These authors jointly supervised this work: Johann S. de Bono, Adam Sharp.

Authors and Affiliations

The Institute of Cancer Research, London, UK
Juan M. Jiménez-Vacas, Daniel Westaby, Ines Figueiredo, Alexis De Haven Brandon, Ana Padilha, Wei Yuan, George Seed, Denisa Bogdan, Bora Gurel, Claudia Bertan, Susana Miranda, Maryou Lambros, Antonio J. Montero-Hidalgo, Lorenzo Buroni, Wanting Zeng, Antje J. Neeb, Jon Welti, Jan Rekowski, Roberta Paravati, Florian Gabel, Nicole Pandell, Ana Ferreira, Mateus Crespo, Ruth Riisnaes, Souvik Das, Joe Taylor, Nick Waldron, Emily Hobern, Melanie Valenti, Jian Ning, Ilona Bernett, Kate Liodaki, Amanda Swain, Florence Raynaud, Marco Bezzi, Suzanne Carreira, Johann S. de Bono & Adam Sharp
The Royal Marsden Hospital, London, UK
Daniel Westaby, Johann S. de Bono & Adam Sharp
Maimonides Institute for Biomedical Research of Cordoba (IMIBIC); Department of Cell Biology, Physiology, and Immunology, University of Cordoba; Reina Sofia University Hospital; CIBER Physiopathology of Obesity and Nutrition (CIBERobn), Cordoba, Spain
Antonio J. Montero-Hidalgo & Raul M. Luque
Divisions of Human Biology and Clinical Research, Fred Hutchinson Cancer Center, Seattle, WA, USA
Ilsa Coleman, Thomas Persse, Patricia Galipeau, Gavin Ha & Peter S. Nelson
Vancouver Prostate Centre, University of British Columbia, Vancouver, BC, Canada
Ivan Pak Lok Yu & Nathan A. Lack
Genitourinary Malignancies Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA
Scott Wilkinson, Shana Y. Trostel, Fatima Karzai, Cindy H. Chau, Erica L. Beatson, William D. Figg & Adam G. Sowalsky
Division of Preclinical Innovation, National Center for Advancing Translational Sciences, National Institute of Health, Rockville, MD, USA
Xiaohu Zhang, Carleen Klumpp-Thomas & Craig J. Thomas
Hematology-Oncology Division, Beth Israel Deaconess Medical Center, Boston, MA, USA
Andreas Varkaris & Steven P. Balk
Department of Medical Pharmacology, School of Medicine, Koc University, Istanbul, Turkey
Nathan A. Lack
Lymphoid Malignancies Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA
Craig J. Thomas

Authors

Juan M. Jiménez-Vacas
View author publications
Search author on:PubMed Google Scholar
Daniel Westaby
View author publications
Search author on:PubMed Google Scholar
Ines Figueiredo
View author publications
Search author on:PubMed Google Scholar
Alexis De Haven Brandon
View author publications
Search author on:PubMed Google Scholar
Ana Padilha
View author publications
Search author on:PubMed Google Scholar
Wei Yuan
View author publications
Search author on:PubMed Google Scholar
George Seed
View author publications
Search author on:PubMed Google Scholar
Denisa Bogdan
View author publications
Search author on:PubMed Google Scholar
Bora Gurel
View author publications
Search author on:PubMed Google Scholar
Claudia Bertan
View author publications
Search author on:PubMed Google Scholar
Susana Miranda
View author publications
Search author on:PubMed Google Scholar
Maryou Lambros
View author publications
Search author on:PubMed Google Scholar
Antonio J. Montero-Hidalgo
View author publications
Search author on:PubMed Google Scholar
Ilsa Coleman
View author publications
Search author on:PubMed Google Scholar
Ivan Pak Lok Yu
View author publications
Search author on:PubMed Google Scholar
Lorenzo Buroni
View author publications
Search author on:PubMed Google Scholar
Wanting Zeng
View author publications
Search author on:PubMed Google Scholar
Antje J. Neeb
View author publications
Search author on:PubMed Google Scholar
Jon Welti
View author publications
Search author on:PubMed Google Scholar
Jan Rekowski
View author publications
Search author on:PubMed Google Scholar
Roberta Paravati
View author publications
Search author on:PubMed Google Scholar
Florian Gabel
View author publications
Search author on:PubMed Google Scholar
Nicole Pandell
View author publications
Search author on:PubMed Google Scholar
Ana Ferreira
View author publications
Search author on:PubMed Google Scholar
Mateus Crespo
View author publications
Search author on:PubMed Google Scholar
Ruth Riisnaes
View author publications
Search author on:PubMed Google Scholar
Souvik Das
View author publications
Search author on:PubMed Google Scholar
Joe Taylor
View author publications
Search author on:PubMed Google Scholar
Nick Waldron
View author publications
Search author on:PubMed Google Scholar
Emily Hobern
View author publications
Search author on:PubMed Google Scholar
Melanie Valenti
View author publications
Search author on:PubMed Google Scholar
Jian Ning
View author publications
Search author on:PubMed Google Scholar
Ilona Bernett
View author publications
Search author on:PubMed Google Scholar
Kate Liodaki
View author publications
Search author on:PubMed Google Scholar
Thomas Persse
View author publications
Search author on:PubMed Google Scholar
Patricia Galipeau
View author publications
Search author on:PubMed Google Scholar
Scott Wilkinson
View author publications
Search author on:PubMed Google Scholar
Shana Y. Trostel
View author publications
Search author on:PubMed Google Scholar
Fatima Karzai
View author publications
Search author on:PubMed Google Scholar
Cindy H. Chau
View author publications
Search author on:PubMed Google Scholar
Erica L. Beatson
View author publications
Search author on:PubMed Google Scholar
Xiaohu Zhang
View author publications
Search author on:PubMed Google Scholar
Carleen Klumpp-Thomas
View author publications
Search author on:PubMed Google Scholar
Andreas Varkaris
View author publications
Search author on:PubMed Google Scholar
Raul M. Luque
View author publications
Search author on:PubMed Google Scholar
Amanda Swain
View author publications
Search author on:PubMed Google Scholar
Florence Raynaud
View author publications
Search author on:PubMed Google Scholar
Nathan A. Lack
View author publications
Search author on:PubMed Google Scholar
Craig J. Thomas
View author publications
Search author on:PubMed Google Scholar
Gavin Ha
View author publications
Search author on:PubMed Google Scholar
William D. Figg
View author publications
Search author on:PubMed Google Scholar
Marco Bezzi
View author publications
Search author on:PubMed Google Scholar
Adam G. Sowalsky
View author publications
Search author on:PubMed Google Scholar
Peter S. Nelson
View author publications
Search author on:PubMed Google Scholar
Suzanne Carreira
View author publications
Search author on:PubMed Google Scholar
Steven P. Balk
View author publications
Search author on:PubMed Google Scholar
Johann S. de Bono
View author publications
Search author on:PubMed Google Scholar
Adam Sharp
View author publications
Search author on:PubMed Google Scholar

Contributions

J.M.J.-V., D.W., J.S.d.B. and A.Sharp. conceived and designed the study. J.M.J.-V., D.W., I.F., A.DH.B., A.P., B.G., C.B., S.M., M.L., A.J.M.-H., I.C., I.P.L.Y., L.B., W.Z., A.J.N., J.W., R.P., F.G., N.P., A.F., M.C., R.R., S.D., J.T., N.W., E.H., M.V., J.N., I.B., K.L., T.P., P.G., S.W., S.Y.T., F.K., C.H.C., E.L.B., X.Z., C.K.-T. and A.V. performed experiments and/or collected data. J.M.J.-V., D.W., W.Y., G.S., D.B. and J.R. analysed the data and/or performed statistical analyses. J.M.J.-V., D.W., R.M.L., A.S., F.R., N.A.L., C.J.T., G.H., W.D.F., M.B., A.G.S., P.S.N., S.C., S.P.B., J.S.d.B. and A.Sharp. interpreted the results and/or contributed to data visualisation. J.M.J.-V., D.W., J.S.d.B. and A.Sharp draughted the manuscript. All authors provided advice on, reviewed, edited and approved the manuscript. J.S.d.B. and A.Sharp served as sponsors and monitored the study.

Corresponding authors

Correspondence to Johann S. de Bono or Adam Sharp.

Ethics declarations

Competing interests

J.M.J.-V., D.W., I.F., A.d.H.V., A.P., W.Y., G.S., D.B., B.G., C.B., S.M., L.B., W.Z., A.J.N., J.W., J.R., R.P., A.F., M.C., R.R., S.D., J.T., E.H., M.V., J.N., K.L., A.S., F.R., S.C., J.S.d.B. and A.Sharp are employees of the ICR, which has a commercial interest in abiraterone, PARP inhibition in DNA repair defective cancers, and PI3K/AKT pathway inhibitors (no personal income). A.G.S. reports that the National Cancer Institute (NCI) has a Cooperative Research and Development Agreement (CRADA) with Astellas. Resources are provided by this CRADA to the NCI. A.G.S. gets no personal funding from this CRADA but is the primary investigator of the CRADA. J.S.d.B. has served on advisory boards and received fees from many companies, including Amgen, Astra Zeneca, Bayer, Bioxcel Therapeutics, Daiichi, Genentech/Roche, GSK, Merck Serono, Merck Sharp & Dohme, Pfizer, and Sanofi Aventis. He is an employee of the ICR, which has received funding or other support for his research work from AZ, Astellas, Bayer, Cellcentric, Daiichi, Genentech, Genmab, GSK, Janssen, Merck Serono, MSD, Menarini/Silicon Biosystems, Orion, Sanofi Aventis, Sierra Oncology, Taiho, Pfizer, Vertex. J.S.d.B. was named as an inventor, with no financial interest, for patent 8,822,438, submitted by Janssen, that covers the use of abiraterone acetate with corticosteroids. J.S.d.B. has been the CI/PI of many industry-sponsored clinical trials. A.Sharp has received travel support from Sanofi, Roche-Genentech and Nurix, and speaker honoraria from Astellas Pharma and Merck Sharp & Dohme. He has served as an advisor to DE Shaw Research, CHARM Therapeutics, Ellipses Pharma and Droia Ventures. A.Sharp has been the CI/PI of industry-sponsored clinical trials. The remaining authors declare no conflicts of interest.

Peer review

Peer review information

Nature Communications thanks Andrea Lunardi, and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Reporting Summary

Transparent Peer Review file

Source data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Jiménez-Vacas, J.M., Westaby, D., Figueiredo, I. et al. Elucidating molecularly stratified single agent, and combination, therapeutic strategies targeting MCL1 for lethal prostate cancer. Nat Commun 16, 8806 (2025). https://doi.org/10.1038/s41467-025-64042-5

Download citation

Received: 07 March 2024
Accepted: 04 September 2025
Published: 08 October 2025
DOI: https://doi.org/10.1038/s41467-025-64042-5

Subjects

Abstract

Introduction

Results

MCL1 copy number gains are common in lethal prostate cancer, occur early in tumour evolution, and increase with castration-resistance

MCL1 copy number gains associate with worse clinical outcome

MCL1 RNA and protein are highly expressed in mCRPC

Response to MCL1 inhibition is enriched in PC models with MCL1 copy number gain

MCL1 inhibition synergises with PI3K inhibition to trigger apoptosis in CRPC cells

MCL1 inhibition synergises with AKT inhibition to trigger apoptosis in CRPC cells

The synergistic anti-tumour activity of AKT and MCL1 inhibition is mediated through alterations in BAD-BCLXL and BIM-MCL1 interactions

AKT and MCL1 co-inhibition triggers BAK-dependent apoptosis in PTEN-loss/PI3K-activated CRPC cells

Synergistic inhibition of tumour growth by AKT and MCL1 co-inhibition in vivo

CRPC cells with acquired resistance to capivasertib remain sensitive to AKT and MCL1 co-inhibition, unlike those with acquired resistance to MCL1 inhibitors

Targeting MCL1 through CDK9 inhibition recapitulates the phenotype of MCL1 inhibition in combination with AKT blockade in CRPC cells

Discussion

Methods

Patients and sample collection

Royal Marsden Hospital cohorts

FIRSTANA and PROSELICA cohort

Stand Up To Cancer/Prostate Cancer Foundation cohort

The Cancer Genome Atlas cohort

National Cancer Institute neoadjuvant cohort

University of Washington/Fred Hutchinson (UW/FH) cohort

Bioinformatic analyses

Copy number analysis

Transcriptome analysis

Immunohistochemistry

Fluorescence in situ hybridisation

Drug compounds

Cell lines

Patient-derived xenograft (PDX) organoids (PDX-Os)

Prostate Cancer Preclinical Mouse Modeling Platform (ProMPt) organoids (ProMPt-Os)

Viability and apoptosis assays

Food and drug administration-approved drug screen

High-throughput combination matrix drug screening

siRNA transfection

Western blot

Co-immunoprecipitation

Proximity ligation assay

In vivo analyses with the CP253C CRPC PDX model

Development of drug-resistant cell lines

Statistical analyses

Reporting summary

Data availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding authors

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Supplementary information

Supplementary Information

Reporting Summary

Transparent Peer Review file

Source data

Source data

Rights and permissions

About this article

Cite this article

Share this article

Search

Quick links