5065 Background: Aggressive variant prostate cancer (AVPC) criteria enrich for androgen-indifferent disease that is highly aggressive and has limited therapeutic options. Single-agent immune checkpoint inhibitors have not demonstrated activity in prostate cancer patients. The C-COLA study (NCT03263650) showed that Cabazitaxel/Carboplatin (CabCarb) elicited inflammatory signaling in responsive patients with AVPC. Based on these findings, we hypothesized that chemotherapy-induced inflammation in the tumor microenvironment (TME) may prime tumors for immunotherapeutic approaches. Methods: The C3NIRA trial (NCT04592237) employed one cycle of CabCarb, followed by the addition of the PD-1 blocker Cetrelimab, from cycle 2 for up to 6 cycles of induction treatment in patients with AVPC. Among 120 enrolled men, 20 discontinued due to toxicities and 40 due to progression during induction. Sixty patients completed induction and were subsequently randomized to Niraparib maintenance with or without Cetrelimab. The 30 randomized to Niraparib with Cetrelimab had significantly longer progression free (PFS) and overall survivals than the 30 randomized to Niraparib alone. To investigate whether treatment-induced alterations in the TME could predict immunotherapy response, biopsies were obtained at baseline, and after cycles 1, 3 and 9 for single-cell RNA sequencing (scRNAseq). Results: scRNAseq analysis of tumor biopsies showed a high proportion of GZMK⁺ CD8⁺ T cells and CD8⁺ T effector memory clusters, and a low proportion in FOXP3 + CD4 + T regulatory cells, in the TME of responders (PFS≥10months) after cycle 1 of CabCarb, which was not seen in progressors. Similarly, SPP1⁺ macrophages scoring high immunosuppressive signatures were reduced in responders after CabCarb but not in progressors. Consistent with these suppressive and effector cell changes, LAG3 + TIGIT + CD8⁺ T exhausted cells gradually accumulated in progressors from baseline through CabCarb and Cetrelimab treatment, while remaining unchanged in responders over the same interval. Conclusions: One cycle of CabCarb induction altered the TME in a subset of AVPC tumors, with reduction of suppressive myeloid and T cell populations and expansion of anti-tumor effector cells associated with clinical responses. These early TME changes may predict benefit from the addition of PD-1 blockade and help refine patient selection for immune checkpoint inhibition. Ongoing analyses are identifying additional therapeutic vulnerabilities in the TME that can be targeted to further improve outcomes in men with AVPC. Clinical trial information: NCT04592237 .

Aggressive-variant prostate cancers (AVPCs) respond poorly to anti-androgen therapy but show sensitivity to taxane-platinum chemotherapy, though outcomes remain poor. We conducted a phase 2 trial testing induction cabazitaxel plus carboplatin (CabCarb) followed by olaparib maintenance versus observation in men with AVPC. The primary endpoint of improved progression-free survival (PFS) was not met, likely due to the study being underpowered after 38.5% of patients experienced early progression (ChemoPD) prior to randomization. No genomic alterations predicted ChemoPD; however, transcriptomic analysis revealed the enrichment of metabolic pathways, including arginine metabolism, in ChemoPD tumors. These findings were supported by metabolomics data from preclinical models. In AVPC models, arginine depletion with ADI-PEG20 enhanced CabCarb efficacy in vitro and in vivo. Together, these results provide insight into the heterogeneity of AVPCs and establish a rationale for novel combination treatment strategies to overcome chemotherapy resistance.

Supplementary Table S6. Quantification of each protein-specific band normalized to vinculin (C4-2b-derived models, Fig. 2B).

Supplementary Table S13. Quantification of each protein-specific band normalized to vinculin (macrophage cells (RAW264.7) mono- or cocultured with RM-1-BM EV or SPOPmut models under treatment with OLA, Fig. 5E).

Supplementary Table S8. Quantification of each protein-specific band normalized to vinculin (Dox-inducible RM-1-BM– and C4-2b–derived models, Fig. 3A, 3B).

Supplementary Table S9. Quantification of each protein-specific band normalized to vinculin (Dox-inducible C4-2b–derived models, Fig. 4D, upper panel).

Supplementary Fig. S18. Original unedited IB images for representative Western blots experiments (n-2) used in Figure 5E.

Supplementary Table S5. Complete listing and details of siRNA targeting sequences used for gene knockdown in Fig. 4A and Extended Data Fig. 3.

Supplementary Table S3. Complete listing of antibodies used for immunoblotting (IB) and immunostaining analysis (Fig. 6C, 6D and Extended Data Fig. 5, 6). Vendor and catalog numbers are included in the list.

Selective autophagy of mitochondria is known to promote cancer cell survival and progression, including in triple-negative breast cancer (TNBC). Here, we apply an integrated multi-omics approach together with functional experimental analyses to investigate metabolic adaptations that support mitochondrial quality control in TNBC. We detail a mitochondrial quality control mechanism, complementary to mitophagy, that is enabled by a program of heightened extracellular sphingomyelin salvaging in TNBC coupled with extracellular vesicle-mediated intracellular clearance of mitochondrial damage. Targeting of this onco-metabolic pathway via repurposing of eliglustat, a selective small molecule inhibitor of glucosylceramide synthase, results in ceramide-mediated compensatory mitophagy and cancer cell death in vitro and attenuates tumor growth and prolongs overall survival at clinically achievable doses in orthotopic syngeneic mouse models of TNBC as well as in human cell line-derived xenograft models. Our study defines an unexplored mechanism of aberrant sphingolipid metabolism that underlies an actionable metabolic vulnerability for anti-cancer treatment.

Supplementary Fig. S4. Original unedited IB images for representative Western blots experiments (n-2) used in Figure 3B.

Supplemental Table S11. Quantification of each protein-specific band normalized to vinculin (C4-2b–derived models, Fig. 4E).

Supplementary Fig. S5. Original unedited IB images for representative Western blots experiments (n-2) used in Figure 3C.

Supplementary Fig. S3. Original unedited IB images for representative Western blots experiments (n-3) used in Figure 3A.

Supplementary Fig. S12. Suppression of STING1 protein levels following siSTING transfection compared to siNC.

Supplementary Fig. S11. Original unedited IB images for representative Western blots experiments (n-2) used in Figure 3F.

Supplementary Fig. S6. Far western blotting assay to detect direct protein-protein interaction of SPOP and STING protein in vitro

Supplementary Fig. S21. Double immunofluorescence staining analysis of cleaved caspase 3 and PARP1 in C4-2b-empty vector and C4-2b-SPOPF133V xenograft tumors treated with talazoparib (TALA) or control vehicle

Supplementary Fig. S1. “Immunome” analysis of SPOPmut and SPOPwt from Beltran CRPC cohort using a compendium of publicly available data from purified immune subsets