DLL3-Targeted Strategies in Advanced Prostate Cancer: Current Evidence and Future Perspectives Authors: Giovanna Pecoraro, Alberto De Giorgi, Giulia Montelatici, Giuseppe Salfi, Hui-Ming Lin, Fabio Turco, Tarek Taha, Himisha Beltran, Johann de Bono, Rahul Raj Aggarwal, Gunhild von Amsberg, Markus Joerger, Martina Imbimbo, Ilaria Colombo, Silke Gillessen, Martino Pedrani Published: 2026 ABSTRACT AbstractInhibition of androgen receptor (AR) signaling remains the cornerstone of systemic therapy for advanced prostate cancer (PC). However, a subset of aggressive tumors either arises de novo with neuroendocrine features or emerges under treatment pressure through lineage plasticity and AR independence. These lethal states are encompassed within the spectrum of aggressive-variant prostate cancer (AVPC), an umbrella term that includes both histologically confirmed neuroendocrine prostate cancer (NEPC)—comprising de novo NEPC and treatment-emergent NEPC (t-NEPC)—and clinically or molecularly defined AVPC lacking histologic confirmation but sharing neuroendocrine-like, AR-indifferent, or small-cell features. These phenotypes are characterized by rapid progression, visceral dissemination, low or discordant prostate-specific antigen (PSA) levels relative to tumor burden, and poor prognosis. Treatment options for NEPC/AVPC remain limited and largely rely on platinum-based chemotherapy, which usually provides only modest and transient benefit. This unmet need has intensified interest in lineage-associated vulnerabilities. Delta-like ligand 3 (DLL3), an inhibitory Notch ligand with restricted expression in normal adult tissues, is aberrantly upregulated in several neuroendocrine malignancies and has emerged as a clinically actionable target. In prostate cancer, DLL3 expression is enriched in neuroendocrine tumor cells, being detected in approximately 76.6% of castration-resistant NEPC compared with only 12.5% of castration-resistant adenocarcinoma, supporting its development as both a biomarker and therapeutic vulnerability. Clinical success of DLL3-targeted therapies in small-cell lung cancer further supports evaluation of DLL3-directed strategies in NEPC and related AVPC states. This review summarizes the biological rationale, translational evidence, and emerging clinical data supporting DLL3-targeted therapies in prostate cancer. Investigational platforms include antibody-drug conjugates, bispecific and trispecific T-cell engagers, and DLL3-directed radiopharmaceuticals. Early clinical studies suggest that activity is largely confined to DLL3-expressing neuroendocrine tumors, highlighting the importance of biomarker-guided patient selection. Delta-like ligand 3-directed therapies may reshape the management of DLL3-expressing prostate cancer if ongoing efforts to refine biomarkers improve patient enrichment and optimize trial design are successfully translated into clinical practice.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s40265-026-02355-5. Key Points Neuroendocrine and aggressive-variant prostate cancer represent treatment-resistant states driven by lineage plasticity and androgen receptor independence, for which effective therapies remain limited.DLL3 is selectively expressed in neuroendocrine prostate cancer and related phenotypes, supporting its role as both a biomarker and a therapeutic target.Emerging DLL3-targeted strategies—including antibody-drug conjugates, T-cell engagers, and radiopharmaceuticals—show early signals of activity, but their clinical utility will depend on improved patient selection and prostate-specific trial design. Introduction Prostate cancer (PC) is the second most commonly diagnosed malignancy among men worldwide and a leading cause of cancer-related mortality, with more than 1.5 million new cases and over 396,000 deaths annually [1, 2]. Although outcomes for localized disease are generally favorable, progression to metastatic PC and castration-resistant disease markedly worsens prognosis. The median overall survival (OS) of metastatic castration-resistant PC (mCRPC) remains approximately two years despite major therapeutic advances [3]. Androgen receptor (AR) signaling is the dominant oncogenic driver in conventional prostate adenocarcinoma and cornerstone of systemic therapy. Treatment intensification with androgen-deprivation therapy (ADT) plus androgen-receptor pathway inhibitors (ARPIs), with or without docetaxel, has substantially improved survival in metastatic hormone-sensitive disease. Cabazitaxel, PARP inhibitors, and radioligand therapies such as radium-223 and lutetium-177-prostate-specific membrane antigen (PSMA) have further expanded the therapeutic armamentarium for mCRPC [4, 5]. However, these advances largely benefit tumors that retain AR dependence and PSMA expression, features that may be lost during neuroendocrine trans-differentiation. Therapeutic resistance in advanced PC arises not only through AR pathway reactivation but also through lineage plasticity, enabling tumor cells to bypass AR dependence and adopt an AR-independent neuroendocrine phenotype [6]. Within this context, a broad spectrum of highly aggressive AR-independent disease is encompassed by the term aggressive-variant prostate cancer (AVPC). Aggressive-variant prostate cancer can be subdivided into two categories: clinically defined AVPC (c-AVPC), referring to metastatic prostate cancer with adverse clinical or pathological features—such as neuroendocrine histology, visceral or lytic metastases, bulky nodal disease, low prostate-specific antigen relative to tumor burden, or rapid progression to castration resistance—and molecular AVPC (m-AVPC), defined by combined loss-of-function alterations in TP53, RB1, and PTEN [7]. These classifications are inherently imperfect, with substantial overlap between c-AVPC, m-AVPC, and histologically defined neuroendocrine prostate cancer (NEPC) [7–9]. Aggressive-variant prostate cancer encompasses a biological continuum ranging from conventional prostate adenocarcinoma with no or limited neuroendocrine (NE) marker expression, to tumors with increasingly pronounced NE features, consistent with a dynamic process of trans-differentiation. Clinically, these AR-indifferent states are typically associated with low or discordant PSA levels relative to tumor burden and rapid disease progression [10]. Dynamic model of lineage plasticity helps explain the coexistence of heterogeneous phenotypic states within advanced PC, spanning adenocarcinoma with minimal neuroendocrine differentiation, mixed or hybrid phenotypes, and fully developed NEPC [11]. Neuroendocrine prostate cancer represents the histologically defined end of this spectrum and arises either de novo or, more commonly, as treatment-emergent NEPC (t-NEPC) following exposure to ADT and/or AR pathway inhibitors. Importantly, advances in the biological and molecular understanding of NE differentiation in PC have informed the evolution of classification systems. The Prostate Cancer Foundation (PCF) 2013 framework and the WHO 2022 classification are broadly concordant in defining core NE neoplasms (e.g., small-cell and large-cell NE carcinomas), but diverge in the categorization of intermediate phenotypes [12, 13]. In particular, the PCF 2013 framework adopts a broader definition of the NEPC spectrum, including conventional adenocarcinomas with NE differentiation identified by immunohistochemistry (IHC), even in the absence of overt NE morphology. In contrast, the WHO 2022 classification applies a more restrictive, morphology-driven approach, classifying such tumors within usual adenocarcinoma and reserving the diagnosis of NEPC primarily for high-grade NE carcinomas. This distinction is clinically relevant, as scattered or heterogeneous expression of NE markers (e.g., chromogranin A, synaptophysin) is frequently observed in conventional prostate adenocarcinoma, yet its biological and prognostic significance remains uncertain and technically challenging to assess, particularly in limited biopsy samples [14]. A meta-analysis estimated that approximately 16% of patients treated with potent AR-directed therapies develop t-NEPC, typically within two years, with median survival after diagnosis often below 12 months [10]. Both de novo and t-NEPC are largely insensitive to AR-targeted therapies and derive only limited, transient benefit from platinum-based chemotherapy [15]. More broadly, effective therapies specifically targeting NEPC and AVPC remain lacking. Platinum-based chemotherapy is the standard of care for de novo NEPC and for tumors that evolve toward t-NEPC or AVPC phenotypes, with immune checkpoint inhibitors occasionally used in selected patients but without robust prospective evidence [4, 16]. Responses are typically short-lived, and no validated targeted therapies exist for NEPC or AVPC. However, emerging data suggest potential benefit from chemo-immunotherapy combinations, as preliminary results from the Phase 1b/2 KEYNOTE-365 study indicate improved response rates and survival outcomes with the addition of pembrolizumab to platinum-based chemotherapy in NEPC [17]. The identification of lineage-associated targets that are selectively expressed in AR-indifferent or neuroendocrine PC has become a major focus of translational research. Delta-like ligand 3 (DLL3), an inhibitory ligand of the Notch signaling pathway, has emerged as a particularly attractive candidate, as it is minimally expressed in most non-malignant adult tissues but aberrantly expressed on the cell surface of NE malignancies, thus enabling antibody-based targeting. Importantly, DLL3 has already been clinically validated as a therapeutic target in small-cell lung cancer (SCLC), a prototypical high-grade NE carcinoma [18, 19]. Informed by a systematic review of the literature and subsequent expert critical appraisal, as detailed in the Appendix S1 and Fig. S1, the present review aims to identify and critically examine the available clinical and translational evidence on DLL3 expression and DLL3-directed therapies in advanced PC, integrating biological rationale with emerging clinical data and ongoing trials to define the current therapeutic landscape and the potential future role of DLL3 targeting in AR-independent PC. Biological Rationale for Targeting DLL3 in Prostate Cancer DLL3 Structure, Function, and Role in Notch Signaling Delta-like ligand 3 is a single-pass transmembrane protein of 619 amino acids, comprising an extracellular Delta/Serrate/Lag2 (DSL) domain and six epidermal growth factor (EGF)-like repeats, plus a short intracellular tail [20]. It is a ligand of the Notch receptor, which is a key regulator of cell-fate decisions, including the balance between NE and non-NE lineages. Classical ligands of Notch (e.g. DLL1, DLL4, Jagged1/2) activate Notch in trans at the cell surface, i.e., by engaging Notch receptors on neighboring cells and triggering their proteolytic activation. Despite the structural similarity of DLL3 to canonical Notch ligands, DLL3 behaves as a non-canonical, predominantly inhibitory ligand of Notch. Delta-like ligand 3 binding to Notch is predominantly in cis, i.e., on the same cell, preventing proper trafficking of Notch receptors to the plasma membrane, and redirecting them to endosomal/lysosomal compartments for degradation [21]. This intracellular “checkpoint” effectively reduces the pool of functional Notch receptors available at the cell surface, thereby attenuating downstream Notch signaling and influencing NE lineage specification and maintenance [22] (Fig. 1). Delta-like ligand 3 (DLL3) biology, lineage plasticity, and therapeutic targeting in prostate cancer. In normal adult prostate epithelium, DLL3 expression is low and predominantly intracellular, with minimal surface accessibility. In contrast, neuroendocrine PC and other lineage-plastic, AR-indifferent prostate cancer states can display aberrant surface expression of DLL3, creating a lineage-associated therapeutic vulnerability. Delta-like ligand 3 functions as a predominantly cis-inhibitory, non-canonical Notch ligand that impairs Notch receptor trafficking and attenuates downstream signaling, thereby supporting neuroendocrine lineage states. This biologic rationale underpins multiple DLL3-targeted therapeutic modalities, including antibody-drug conjugates, DLL3 × CD3 bispecific T-cell engagers, trispecific DLL3 × CD3 × albumin-binding T-cell engagers designed for half-life extension, and DLL3-directed radiopharmaceuticals using either beta-emitting lutetium-177 or alpha-emitting actinium-225 payloads. Optional DLL3-directed PET/SPECT imaging may support in vivo target confirmation, biomarker-guided patient selection, and theranostic development. The figure also highlights the biologic heterogeneity of advanced prostate cancer, in which mixed target states may coexist and influence treatment selection During embryogenesis, DLL3 is expressed in neural and neuroendocrine tissues, particularly in the developing spinal cord and nervous system. In adult non-malignant tissues, DLL3 expression is low or absent and largely confined to intracellular locations with minimal detectable protein at the cell surface [23]. In contrast, high-grade neuroendocrine neoplasms (NENs), such as SCLC and NEPC, frequently display high expression of DLL3, with a substantial fraction of the protein aberrantly trafficked to and retained at the plasma membrane [24–26]. This expression pattern reflects the transcriptional reprogramming that drives NE differentiation in which sustained DLL3 expression helps maintain a Notch-low, NE state, while the sheer abundance of DLL3 saturates the secretory pathway and results in increased surface localization [24, 25, 27]. The sharp contrast between low, predominantly intracellular expression in normal tissues and aberrant surface overexpression in NENs creates a broad therapeutic window for DLL3-directed antibodies, antibody-drug conjugates (ADCs), T-cell engagers (TCEs), and radiopharmaceuticals, while minimizing off-target toxicity. Delta-like ligand 3 expression could be exploited in two ways—as a selectively expressed surface antigen to deliver cytotoxic/radioactive payloads to NE tumor cells, and as a functional target for inhibiting Notch-regulation of NE differentiation. Functionally, preclinical PC models suggest that DLL3 primarily marks a NE lineage state that can be therapeutically exploited, rather than acting as a classical oncogenic driver. Although DLL3 gene silencing does not consistently reverse NE differentiation or exert major antiproliferative effects, DLL3-positive models are selectively sensitive to multiple DLL3-directed platforms, including ADCs, radiopharmaceuticals, and TCEs, whereas DLL3-negative tumors are largely unresponsive [19, 28–31]. In DLL3-expressing PC xenografts, the ADC rovalpituzumab tesirine (Rova-T) can induce complete and durable regressions after a single dose [28], and DLL3-directed radiopharmaceutical approaches similarly show target-dependent antitumor activity [28, 29]. Likewise, DLL3-directed TCEs, including tarlatamab and HPN328, trigger robust DLL3-dependent T-cell activation and tumor regression in NEPC models; notably, activity can extend to mixed DLL3-positive/DLL3-negative tumors through bystander killing of adjacent DLL3-negative cells [30, 31]. Together, these findings support DLL3-directed targeting as a biologically rational strategy in NEPC. DLL3 Expression and Its Assessment in Prostate Cancer Trials Delta-like ligand 3 is highly expressed in NEPC and in AR-low/RB1-altered lineage-plastic states that overlap with AVPC, while remaining largely absent in conventional adenocarcinoma [6, 26, 28]. Several molecular factors regulate or alter DLL3 expression in PC tissues. Delta-like ligand 3 transcription is closely linked to the activity of NE lineage transcription factors, particularly ASCL1 (achaete-scute homolog 1), which directly drives DLL3 gene expression and whose activity is itself promoted by loss of RB1 and TP53—the two most commonly altered tumor suppressors in NEPC [28, 32]. Accordingly, combined RB1/TP53 loss, which occurs in the majority of NEPC cases, is strongly associated with high DLL3 expression, both at the transcript and protein levels. Enhancer of Zeste 2 Polycomb Repressive Complex 2 Subunit (EZH2), a histone methyltransferase and key epigenetic enforcer of the NE lineage program through PRC2-mediated chromatin silencing of AR and luminal differentiation genes, also promotes DLL3 upregulation indirectly by sustaining an ASCL1-active, Notch-low transcriptional state [33, 34]. Exposure to ADT and AR pathway inhibitors (ARPIs) is a major upstream driver of NE trans-differentiation and the associated increase in DLL3 expression: in treatment-emergent NEPC. Delta-like ligand 3 upregulation typically occurs in the context of therapy-induced lineage plasticity and reflects the progressive silencing of AR-dependent programs [15]. Brain class POU box-containing protein 2 (BRN2 [POU3F2]) and SOX2, additional transcription factors activated during NE reprogramming, further reinforce this state [35]. Conversely, active Notch signaling—such as that mediated by canonical ligands DLL1, DLL4, or Jagged1—suppresses NE differentiation and is associated with lower DLL3 expression, consistent with the cis-inhibitory role of DLL3 in attenuating Notch activity [27]. Together, these data indicate that DLL3 expression is not a static feature but a dynamic, context-dependent property shaped by the interplay of genomic alterations, epigenetic regulators, lineage transcription factor networks, and therapeutic selective pressure. Delta-like ligand 3 expression in PC was evaluated predominantly in mCRPC specimens using IHC on formalin-fixed paraffin-embedded tumor tissue, with complementary assessment by RNA sequencing, single-cell transcriptomics, and circulating tumor cells (CTCs) analyses in selected cohorts (Tables 1 and 2). The most comprehensive tissue analysis by IHC to date, encompassing benign prostate (n = 103), localized PC (n = 194), and mCRPC/NEPC tumors (n = 103), demonstrated DLL3 protein expression in 76.6% of castration-resistant NEPC and 12.5% of CRPC adenocarcinoma, with minimal or absent expression in localized PC and benign tissues [28]. Subsequent metastatic cohort studies integrating IHC and RNA profiling confirmed high expression of DLL3 in AR-negative/NE tumors and its association with RB1-altered, AR-low states in mCRPC specimens and patient-derived PC xenografts [26]. Single-cell and spatial analyses showed that DLL3 expression was largely restricted to NE tumor cell states, with occasional DLL3-positive subpopulations within CRPC adenocarcinoma, consistent with lineage plasticity and mixed phenotypes [36, 37]. Table 1Translational studies evaluating DLL3 expressionStudyYearStudy typeSamples/modelDLL3 assay & specimen typeKey findings (as reported)Puca et al. [28]2019Translational (tissue + PDX)Benign prostate (n = 103), localized PC (n = 194), metastatic CRPC (n = 56), high-grade prostate cancer with NE differentiation (n = 19), small cell carcinoma (n = 28); NEPC xenograftsIHC on FFPE tumor; correlative RNA-seq (antibody clone not reported)DLL3 expressed in 76.6% CRPC-NE (36/47) and in 12.5% mCRPC adenocarcinoma (7/56); minimal in localized PC (1/194) and absent in benign (0/103). DLL3 in CTCs concordant with matched metastatic biopsy. Correlates with NE markers and RB1 loss; associated with worse OS from diagnosis and from metastasisChou et al. [37]2023Translational (biopsies + PDX + imaging)mCRPC biopsies with de novo or treatment-emergent SCNC; NEPC PDX modelsIHC on tumor biopsies; DLL3-targeted PET tracer ([⁸⁹Zr]-DFO-DLL3-scFv) in PDXDLL3 positivity reported by IHC in 9/12 (75%) mCRPC biopsies with de novo/t-SCNC; DLL3 absent in adenocarcinoma controls. PET tracer distinguished DLL3+ vs DLL3− xenografts; tarlatamab showed antitumor activity in DLL3-high PDX; antigen loss/downregulation suggested as resistance mechanism in late relapses after AMG 757 treatmentAjkunić et al. [26]2024Translational (metastatic cohort + PDX)753 autoptic metastatic tissue samples from 52 mCRPC patients; mCRPC PDXsIHC and molecular correlates (IHC and molecular correlates in metastatic tissue samples)No expression of DLL3 in 39/52 (75%) of cases, heterogenous expression in in 9/52 (17%) and homogeneous high expression in 4/52 (8%). DLL3 enriched in AR-negative/NEPC; often co-expressed with CEACAM5 (38/71 [54%]), DLL3 expression alone in 21/71 (30%). Higher DLL3 expression in liver and lung metastases compared to bone metastases. DLL3 associated with RB1 alterations; lower expression in CRPC with AR alterationsZaidi et al. [36]2024Translational (single-cell)Murine and human prostate cancer samples (treatment-resistant states)Single-cell profiling with cell-surface antigen mapping (platform details not reported)Surface DLL3 expression largely restricted to NEPC tumor cell states and concordant with NEPC transcription factor programs; subset of AR+ CRPC adenocarcinoma cells show DLL3 expression, consistent with mixed lineage statesBonstingl et al. [92]2024Translational (CTC/spatial)In vitro PC lines; patient tumor tissue; circulating tumor cells (CTCs)Multiplex in situ CTC/tissue assay (“CoDuCo in situ”) + blood NE markersDLL3 positivity in 3–16% of cells across in vitro models. DLL3 increased in CTCs of two patients (one t-NEPC); in t-NEPC, DLL3 in CTCs co-occurred with elevated NSE and CgA in bloodTendler et al. [44]2024Translational/clinical imaging biomarker (FIH immunoPET)High-grade NE tumors of lung and prostate; NEPC n=4 in total cohort (n = 18)[⁸⁹Zr]Zr-DFO-SC16.56 immunoPET/CT + paired tumor IHC where availableDLL3-avid tumors identified in 12/15 (80%) evaluable patients; PET uptake congruent with DLL3 IHC in 15/16 (94%) with evaluable tissue—supports feasibility of non-invasive DLL3 assessment in advanced disease, including prostate cancerChen et al. [93]2024Translational (transcriptomic signatures; conference)WES + transcriptome signatures performed on 4476 prostate tumor samples (3623 prostate adenocarcinoma, 56 NEPC, 25 mixed adeno-NE)Transcriptomic AR/NE signatures; DLL3 expression measured at RNA levelDLL3 expression higher among NE+ prostate cancer subtypes defined by transcriptomic signatures of AR and NE signalingDudipala et al. [94]2026Translational (pathology cohort)5334 prostate cancer surgical specimens including 175 intraductal carcinoma (IDC-P)Tissue profiling (DLL3 expression; assay type NR)Intraductal carcinoma specimens had significantly higher DLL3 expression compared with non-intraductal prostate cancerMizuno et al. [95]2025Translational (epigenetics)98 tumor tissue samples from 35 patients with metastatic CRPC (9 NEPC and 26 adeno)DNA methylation profiling + expression correlationDNA methylation of two DLL3 gene regions correlated with increased DLL3 expression; highlights intraindividual epigenetic heterogeneity linked to phenotypic subtypes of advanced PCLozada et al. [32]2025Translational (pathology cohort; pan-NEN)Human NE neoplasm specimens; includes prostate NENRNA expression-based DLL3 assessment (plus histopathology context)Prostate NE neoplasms show higher DLL3 expression compared with matched adenocarcinoma from same site; 76% of prostate NEN had high DLL3 RNA expression statusLai et al. [96]2025Translational (real-world cohort)6 metastatic prostate NE carcinoma patientsDLL3 positivity assessed in tumor tissue (IHC implied; assay details not reported)DLL3-positive cancer in 3/6 (50%) NEPC patients (one pure, two mixed adenocarcinoma); reports real-world treatment outcomes and genomic insightsVlachostergios et al. [97]2021Translational (localized PC transcriptomic)122 prostatectomy samples from treatment-naïve localized prostate cancerRNA expression analysisDLL3 RNA expression enriched in tumors with alterations in endothelin axis and NF-κB subunit genes (NE differentiation hallmarks), suggesting early NE lineage signals.Korsen et al. [83]2022Translational imaging (preclinical bridge)Prostate cancer cell lines + NEPC xenografts[⁸⁹Zr]-DFO-SC16 immunoPET; correlates with tumor phenotypeH660 NEPC cells positive for DLL3 and negative for AR/PSA/PSMA; PET uptake restricted to H660 xenografts, supporting DLL3 imaging specificity for NEPC-like diseaseKorsen et al. [98]2022Translational theranostic (preclinical bridge)Prostate cancer cell lines + NEPC xenografts[1⁷⁷Lu]-DTPA-SC16 radioimmunotherapy (DLL3-targeted)DLL3-targeted 1⁷⁷Lu-SC16 showed curative effects in NEPC xenografts; no inhibition in adenocarcinoma xenografts—supports DLL3 as theranostic target in NEPC modelsAssay heterogeneity: multiple studies do not fully report antibody clone/platform, scoring system, or biopsy site; where not available in the original report, this is marked as not reportedScope: table includes tissue-based expression, single-cell/spatial/CTC biology, and DLL3 imaging/theranostic feasibility relevant to NEPCAR androgen receptor, CgA chromogranin A, CRPC castration-resistant prostate cancer, CTC circulating tumor cells, DLL3 Delta-like ligand 3, FFPE formalin-fixed paraffin-embedded, FIH first in human, IDC-P intraductal carcinoma of the prostate, IHC immunohistochemistry, mCRPC metastatic castration-resistant prostate cancer, NEN neuroendocrine neoplasm, NEPC neuroendocrine prostate cancer, NF-κB nuclear factor-kappa B, NSE neuron-specific enolase, OS overall survival, PC prostate cancer, PSA prostate-specific antigen, PSMA prostate-specific membrane antigen, SCNC small cell neuroendocrine prostate cancer, WES whole exome sequencing Translational studies evaluating DLL3 expression Assay heterogeneity: multiple studies do not fully report antibody clone/platform, scoring system, or biopsy site; where not available in the original report, this is marked as not reported Scope: table includes tissue-based expression, single-cell/spatial/CTC biology, and DLL3 imaging/theranostic feasibility relevant to NEPC AR androgen receptor, CgA chromogranin A, CRPC castration-resistant prostate cancer, CTC circulating tumor cells, DLL3 Delta-like ligand 3, FFPE formalin-fixed paraffin-embedded, FIH first in human, IDC-P intraductal carcinoma of the prostate, IHC immunohistochemistry, mCRPC metastatic castration-resistant prostate cancer, NEN neuroendocrine neoplasm, NEPC neuroendocrine prostate cancer, NF-κB nuclear factor-kappa B, NSE neuron-specific enolase, OS overall survival, PC prostate cancer, PSA prostate-specific antigen, PSMA prostate-specific membrane antigen, SCNC small cell neuroendocrine prostate cancer, WES whole exome sequencing Table 2Preclinical evaluation of DLL3-targeted therapeutic and imaging strategies in prostate cancerAgent/platformModalityDLL3 assayExperimental modelKey findingsReferences177Lu-DTPA-SC16DLL3-targeted radioimmunotherapyDLL3 expression assessed by IHC (SC16 ab) in SCLC and NEPC (H660, DLL3 high) xenografts; in vitro SC16 binding assayIn vivo efficacy evaluated in NEPC CDX (H660, DLL3 high); DU145 adenocarcinoma xenografts (DLL3 negative) as negative controlCurative responses and tumor eradication in DLL3 high NEPC xenografts; no activity in adenocarcinoma modelsKorsen et al. 2022 [98][177Lu]Lu-DTPA-CHX-A″-TDI-Y-010DLL3 radiopharmaceuticalDLL3 expression assessed by IHC (SC16.56 ab) and flow cytometry in SCLC and NEPC (H660, DLL3 high) xenografts; in vitro SC16.56 binding assayIn vivo efficacy evaluated in NEPC CDX (H660, DLL3 high), SCLC CDX (NCI-H82, DLL3 high) and SCLC PDX (Lu149-DLL3 medium)Complete responses in 7/9 DLL3 high xenografts with favorable biodistribution and durable antitumoral activityTendler et al. 2026 [61]FZ-AD005DLL3 antibody–drug conjugateDLL3 expression assessed by IHC (anti-DLL3 mAB FZ-A038) and flow cytometry in SCLC and NEPC CDX and PDXIn vivo efficacy evaluated in NEPC CDX (NCI-H660, DLL3 high), SCLC CDX (NCI-H82, NCI-H889; DLL3 high) and SCLC PDX (LU-5236)Complete tumor growth inhibition across dose levels in DLL3 high tumors Guo et al. 2024 [51] MK-6070 (HPN328)DLL3 T-cell engager (tri-specific: DLL3 × CD3 × albumin)DLL3 expression assessed by IHC (anti-DLL3 clone SP347) and flow cytometry in NEPC cell lines and PDXIn vitro cytotoxicity assays performed on NEPC CDX (NCI-H660, 22Rv1, 22Rv1-DLL3); in vivo efficacy evaluated in human T-cell-reconstituted NEPC xenografts models: WCM154 (high DLL3), WCM12 (high DLL3), PC7 (heterogeneous DLL3). For NEPC PDX, not specified whether models are de novo SCNC or t-SCNCDLL3-dependent T-cell-mediated cytotoxicity; strong tumor regression in DLL3⁺ NEPC xenografts; activity proportional to fraction of DLL3⁺ cellsKu et al. 2026 [31]AMG 757 (tarlatamab)DLL3 × CD3 Bi-Specific T-cell engagerDLL3 expression assessed by IHC (anti-DLL3 clone SP347) and flow cytometry in NEPC cell lines (H660, DLL3 high) and xenograftsIn vitro cytotoxicity assays performed on NEPC cell-line (H660); in vivo efficacy evaluated in NEPC CDX (H660, DLL3 high) and PDX: LTL-545 (DLL3 high), LTL-331R (DLL3 low/heterogeneous), LTL-370). For NEPC PDX, not specified whether models are de novo NEPC or t-NEPCDLL3-dependent cytotoxicity confirmed in vitro. Tumor growth suppression and survival prolongation in NEPC CDX and PDX Durable responses occurred in a subset of models and correlated with DLL3 expression (e.g., 20% of CRs in LTL-545 PDX); loss of DLL3 identified as a potential resistance mechanism to AMG-757; CD8⁺ T-cell infiltration; Preliminary clinical activity observed: objective response in a patient with t-NEPC (DLL3⁺ by IHC, H-score 125), partial response ongoing at 10 months Chou et al. 2023 [37] 89Zr-DFO-SC16DLL3 immunoPETDLL3 expression assessed by IHC (SC16 ab) in NEPC (H660, DLL3 high) xenografts; radiolabeled SC16 binds DLL3In vivo PET imaging performed on NEPC CDX (H660, DLL3 high); DU145 adenocarcinoma xenografts (DLL3 negative) as negative controlSelective uptake in DLL3 high NEPC CDX Korsen et al. 2022 [83] Preclinical studies in neuroendocrine prostate cancer (NEPC) cell lines and xenograft models demonstrate selective activity of DLL3-directed ADCs, T-cell–engaging biologics, and radiopharmaceuticals in DLL3-expressing tumors, with minimal effects in DLL3-negative adenocarcinoma models. Imaging studies using DLL3-targeted immunoPET agents confirm tumor-specific uptake and support the development of DLL3-based theranostic approachesADCs antibody-drug conjugates, CR complete response, DLL3 delta-like ligand 3, IHC immunohistochemistry, mAB monoclonal antibody, NEPC neuroendocrine prostate cancer, PDX patient-derived xenograft, SCLC small cell lung cancer, SCNC small cell neuroendocrine prostate cancer Preclinical evaluation of DLL3-targeted therapeutic and imaging strategies in prostate cancer Preclinical studies in neuroendocrine prostate cancer (NEPC) cell lines and xenograft models demonstrate selective activity of DLL3-directed ADCs, T-cell–engaging biologics, and radiopharmaceuticals in DLL3-expressing tumors, with minimal effects in DLL3-negative adenocarcinoma models. Imaging studies using DLL3-targeted immunoPET agents confirm tumor-specific uptake and support the development of DLL3-based theranostic approaches ADCs antibody-drug conjugates, CR complete response, DLL3 delta-like ligand 3, IHC immunohistochemistry, mAB monoclonal antibody, NEPC neuroendocrine prostate cancer, PDX patient-derived xenograft, SCLC small cell lung cancer, SCNC small cell neuroendocrine prostate cancer Delta-like ligand 3 was detected in CTCs from patients with t-NEPC, often co-occurring with classical NE markers (chromogranin A, NSE) [38, 92]. Longitudinal and autopsy-based analyses similarly document transition from DLL3-negative adenocarcinoma to DLL3-high small-cell morphology, paralleling NE trans-differentiation [28, 37]. Across clinical studies, DLL3 assessment in advanced PC has been based predominantly on IHC of metastatic tumor tissue, although assays, scoring systems, and positivity thresholds have varied across trials. In the Phase 1/2 Rova-T study, DLL3 was assessed at baseline on fresh or archival samples using the Ventana SC16.65 assay, with eligibility defined by staining in ≥ 1% of tumor cells and exploratory analyses of higher-expression groups such as ≥ 50% [40]. The DeLLpro-300 tarlatamab study similarly used the Ventana DLL3 (SP347) assay with a ≥ 1% cutoff, identifying DLL3 expression in 56% of metastatic NEPC tumors, enriched in small-cell morphology and absent in tumors defined only by genomic AVPC features without NE marker expression [41]. In the Phase 1/2 HPN328 (gocatamig) study, DLL3 positivity was generally defined as > 0% tumor-cell staining by IHC for most NE carcinoma cohorts, whereas NEPC enrollment relied mainly on clinicopathologic evidence of NE disease [42]. By contrast, the Phase 1 obrixtamig study used a more stringent enrichment strategy, requiring centrally confirmed DLL3-high expression, defined as ≥ 50% of evaluable tumor cells with moderate-to-strong membranous and/or cytoplasmic staining by VENTANA DLL3 (SP347) [43]. Delta-like ligand 3-targeted immunoPET with [⁸⁹Zr]Zr-DFO-SC16.56 has also shown high concordance with tissue IHC, supporting the feasibility of non-invasive DLL3 assessment [44–46]. Despite consistent evidence of DLL3 enrichment in NEPC, important knowledge gaps remain. Delta-like ligand 3 prevalence has not been systematically characterized across the full aggressive-variant spectrum (clinically and molecularly-defined AVPC), and most available datasets derive from small, NEPC-enriched metastatic cohorts. Assay methodology is heterogeneous, with variable antibodies, scoring cut-offs, specimen types, and limited reporting of key pre-analytical factors (biopsy site, archival vs fresh tissue, and intrapatient heterogeneity). Moreover, no validated thresholds currently define biologically or therapeutically meaningful “DLL3-high” PC. Clinical Studies of DLL3‑Targeted Therapies in Prostate Cancer Clinical development of DLL3-targeted therapies in PC has focused predominantly on de novo and t-NEPC, in which DLL3 expression is most prevalent [28, 40–42, 47]. Investigational strategies include ADCs, TCEs, and radiopharmaceuticals, although clinical experience remains limited to a small number of early-phase studies. Preclinical studies of these agents in NEPC models had demonstrated selective activity in DLL3-expressing tumors, with minimal effects in DLL3-negative adenocarcinoma models (Table 2). Imaging studies using DLL3-targeted immuno-PET agents confirm tumor-specific uptake and support the development of DLL3-based theranostic approaches. To date, published PC data are available from three completed DLL3-directed studies—Rova-T, tarlatamab, and HPN328 (gocatamig)—while preliminary results have been reported for an additional agent—obrixtamig (Table 3). Across these trials, enrichment strategies have combined DLL3 IHC with clinicopathologic definitions of NEPC/AVPC and, in some cases, molecular features associated with AR-independent lineage states, although eligibility criteria varied substantially across studies. In the Phase 1/2 Rova-T trial, enrollment required DLL3 expression in fresh or archived tumor tissue by IHC, generally using a cutoff of ≥ 1% tumor-cell staining. In DeLLpro-300, patients with metastatic de novo or t-NEPC were eligible on the basis of small-cell morphology, NE marker expression, or combined TP53, RB1, and/or PTEN alterations, whereas DLL3 expression was evaluated retrospectively by IHC using a ≥ 1% membranous staining threshold. In the Phase 1/2 HPN328 (gocatamig) study, patients with relapsed or refractory NE carcinomas, including NEPC, were enrolled; DLL3 IHC positivity was required only for non-prostatic NE neoplasms, while NEPC patients were included regardless of DLL3 status. Table 3Clinical studies of DLL3-targeted therapies in NE prostate cancerAgent, NCT (trial status)ModalityPhaseStudy design and populationDLL3 inclusion/cut offDLL3 assayKey findingsReferencesTarlatamab (AMG 757)NCT04702737DLL3×CD3 Bi-Specific T-cell engager1bMetastatic de novo NEPC and t-NEPCEligible patients: (i) morphologic evidence of small cell NEPC (ii) positive staining for chromogranin and/or synaptophysin by IHC in > 50% of the tumor sample, and/or (iii) ≥ 2 genetic alterations in tumor protein p53 (TP53), RB1, and/or PTEN observed by IHC or genomic analysis (molecular AVPC)DLL3⁺ not requested for eligibility; DLL3 expression assessed retrospectively on submitted tumor samples; Positive membranous DLL3 expression, defined as ≥ 1% DLL3 tumor positivity by IHC;Ventana DLL3 (SP347) assayORR 22.2% in DLL3⁺ tumors (4/18); 0/22 responses in DLL3−/unknown; 0/5 in genomic-only (AVPC-like without NE markers). Median rPFS 3.5 vs 1.7 months and median OS 9.8 vs 5.5 months for DLL3⁺ vs DLL3−/unknown; CRS generally manageable with mitigation;Aggarwal et al. 2025 [99]Gocatamig(MK-6070/HPN328)NCT04471727(active, not recruiting)Tri-specific DLL3×CD3×albumin engagerstep-up dosing followed by target doses (6, 12, or 24 mg. QW or Q2W)1/2Refractory/relapse metastatic de novo NEPC and t-NEPC, SCLC and other DLL3⁺ malignanciesDLL3⁺ requested for eligibility for other malignancies but not for NEPC; DLL3 expression defined as ≥1% DLL3 tumor positivity by IHCNAIn the NEPC subset reported (10 treated, 6 evaluable): ORR 50% (3/6 evaluable); CRS noted as a dose-limiting toxicity in 2 patients;Across all doses: ORR NEPC 15%; other NEN 46%. 24 mg Q2W cohort: ORR NEPC 29% (n = 7), other NEN 50% (n = 12)Grade 3–4 TRAEs: NEPC 40%, other NEN 53%, no Grade 5 TRAEs. DLL3 expressed in 88% of NEPC, response did not correlate with DLL3 levelsBeltranet al. 2024 [79] and Beltran et al. 2026 [42]Rovalpituzumab tesirineNCT02709889DLL3 antibody-drug conjugate (SC16.65 ADC)1/2Advanced solid DLL3+ tumors incl. NEPC (21 NEPC among 101 NEC/NET)DLL3⁺ requested for eligibility (central assessment); DLL3 expression defined as ≥ 1% DLL3 tumor positivity by IHC; exploratory analysis by DLL3 level (DLL3-high vs DLL3-low) using ≥ 50% tumor cell staining as thresholdVentana DLL3 (SC16.65 mouse antibody) assayIn pooled NENs: ORR 13% (9/69); trend toward higher response rate among patients with high DLL3 expression, including NEPC high rates of clinically significant toxicity (including serosal effusions) limited feasibility; program discontinued Mansfield 2021 [40] Obrixtamig (BI 764532)NCT04429087 (ongoing)DLL3/CD3 IgG-like T-cell engager1DLL3⁺ SCLC, LCNEC and epNEC (incl. NEPC)DLL3⁺ requested for eligibility (central assessment); positivity cut-off for enrollment NA; exploratory analysis by DLL3 level (DLL3-high vs DLL3-low) using ≥ 50% tumor cell staining as thresholdVentana DLL3 (SP347) assayN=60 epNEC (30% GU epNEC); DLL3-high epNEC: ORR 40%, median DoR 7.9 mo; higher ORR/DCR than DLL3-low; responses enriched in GU and GEP epNEC; manageable toxicity profile, CRS was mostly low grade; grade ≥3 CRS occurred in 3.3% of both DLL3-high and DLL3-low patients)Capdevila et al.ASCO 2025 [39] abstract Early-phase clinical trials in patients with metastatic NE prostate cancer (NEPC) demonstrate DLL3 expression in a substantial subset of tumors and provide preliminary evidence of antitumor activity with DLL3-targeted T-cell–redirecting therapies. DLL3-directed antibody–drug conjugates have shown limited efficacy and notable toxicityADCs antibody-drug conjugates, AVPC aggressive-variant prostate cancer, CR complete response, CRS cytokine release syndrome, DLL3 delta-like ligand 3, epNEC extra-pulmonary neuroendocrine carcinomas, GEP gastroenteropancreatic, GU genitourinary, IgG immunoglobulin gamma, IHC immunohistochemistry, LCNEC large cell neuroendocrine carcinoma, mAB monoclonal antibody, NEN neuroendocrine neoplasm, NEPC neuroendocrine prostate cancer, ORR objective response rates, PDX patient-derived xenograft, QW every week, SCLC small cell lung cancer, SCNC small cell neuroendocrine prostate cancer Clinical studies of DLL3-targeted therapies in NE prostate cancer Tarlatamab (AMG 757) NCT04702737 Metastatic de novo NEPC and t-NEPC Eligible patients: (i) morphologic evidence of small cell NEPC (ii) positive staining for chromogranin and/or synaptophysin by IHC in > 50% of the tumor sample, and/or (iii) ≥ 2 genetic alterations in tumor protein p53 (TP53), RB1, and/or PTEN observed by IHC or genomic analysis (molecular AVPC) Gocatamig (MK-6070/HPN328) NCT04471727 (active, not recruiting) Tri-specific DLL3×CD3×albumin engager step-up dosing followed by target doses (6, 12, or 24 mg. QW or Q2W) In the NEPC subset reported (10 treated, 6 evaluable): ORR 50% (3/6 evaluable); CRS noted as a dose-limiting toxicity in 2 patients; Across all doses: ORR NEPC 15%; other NEN 46%. 24 mg Q2W cohort: ORR NEPC 29% (n = 7), other NEN 50% (n = 12) Grade 3–4 TRAEs: NEPC 40%, other NEN 53%, no Grade 5 TRAEs. DLL3 expressed in 88% of NEPC, response did not correlate with DLL3 levels Rovalpituzumab tesirine NCT02709889 Obrixtamig (BI 764532) NCT04429087 (ongoing) Capdevila et al. ASCO 2025 [39] abstract Early-phase clinical trials in patients with metastatic NE prostate cancer (NEPC) demonstrate DLL3 expression in a substantial subset of tumors and provide preliminary evidence of antitumor activity with DLL3-targeted T-cell–redirecting therapies. DLL3-directed antibody–drug conjugates have shown limited efficacy and notable toxicity ADCs antibody-drug conjugates, AVPC aggressive-variant prostate cancer, CR complete response, CRS cytokine release syndrome, DLL3 delta-like ligand 3, epNEC extra-pulmonary neuroendocrine carcinomas, GEP gastroenteropancreatic, GU genitourinary, IgG immunoglobulin gamma, IHC immunohistochemistry, LCNEC large cell neuroendocrine carcinoma, mAB monoclonal antibody, NEN neuroendocrine neoplasm, NEPC neuroendocrine prostate cancer, ORR objective response rates, PDX patient-derived xenograft, QW every week, SCLC small cell lung cancer, SCNC small cell neuroendocrine prostate cancer Antibody-Drug Conjugates (ADCs) Preclinical evidence provides strong mechanistic and pharmacologic support for DLL3-directed ADC therapy in NEPC. Antibody-drug conjugate binding to DLL3 on the tumor-cell surface in NEPC results in rapid antibody-mediated internalization, enabling efficient intracellular delivery of conjugated cytotoxic payloads [28, 48–50]. Delta-like ligand 3-targeted ADCs exhibit strictly antigen-dependent cytotoxicity across DLL3-positive PC cell lines and patient-derived xenograft (PDX) models, with minimal activity in DLL3-negative controls, confirming target specificity. In NEPC PDX, these agents induce durable tumor regression or complete growth inhibition at tolerated exposures and demonstrate pharmacodynamic evidence of DNA damage-mediated cell death consistent with their topoisomerase-I or PBD payload classes [28, 51]. Importantly, in preclinical studies, ADC activity correlates with the proportion and intensity of DLL3-expressing tumor cells, supporting DLL3 as a quantitative therapeutic biomarker [51]. To date, only one study has explored a DLL3-targeted ADC in NEPC in humans—the Phase 1 trial of Rova-T in which 101 patients with DLL3-positive NEN were treated with the trial drug, including 21 with NEPC (the study did not report detailed subclassification into de novo versus t-NEPC disease). Patients were heavily pretreated (≈ 55% received ≥ 3 prior systemic lines), and trial eligibility required at least 1% DLL3 expression by tumor IHC. Objective response rates (ORRs) were modest overall (ORR 13%, 9/69 evaluable patients). Exploratory analyses suggested higher activity in tumors with high DLL3 expression (≥ 50% tumor-cell staining), with an ORR of 17.1% (6/35) compared with 8.8% (3/34) in the DLL3-low subgroup (1–49% staining). However, this difference in response rate did not translate into clearly improved survival outcomes, with similar median progression-free survival (PFS) (4.3 vs 3.3 months) and overall survival (OS, 7.4 vs 7.1 months) between the DLL3-high and DLL3-low groups [40]. Responses were typically short-lived (median duration of response [DOR] in the whole population 3.1 months) and accompanied by substantial ADC-related toxicity, limiting sustained dosing. These findings derive from a heterogeneous NEN cohort (predominantly SCLC), and no dedicated analysis in NEPC has been reported to date. These data indicate that early DLL3-ADC efficacy signals in PC have been explored only in morphologically defined NEPC with currently insufficient evidence to support its relevance in the broader AVPC population. T Cell Engagers (Bi- and Tri-Specific T Cell–Redirecting Antibodies) In preclinical models, DLL3-directed TCEs (tarlatamab, MK-6070/HPN328) elicit robust, DLL3-dependent T-cell activation and tumor regression in NEPC models including mixed DLL3-positive/negative tumors, by redirecting CD3⁺ T cells to DLL3-expressing tumor cells, and inducing antigen-dependent cytotoxicity, where bystander killing extends activity to DLL3-negative cells [30, 31]. Delta-like ligand 3-directed immunoPET tracers demonstrated accurate in vivo detection of DLL3-positive prostate tumors, enabling non-invasive target assessment and integration of DLL3 imaging as a translational biomarker [37]. Early clinical activity of tarlatamab in prostate cancer was demonstrated in the Phase 1b DeLLpro-300 study, which enrolled a broader metastatic AVPC/NEPC-spectrum population that included de novo NEPC, treatment-emergent NEPC (t-NEPC), and related aggressive-variant phenotypes. Among the 40 enrolled patients, tumor histology was evaluable in 37, including small-cell carcinoma (pure or mixed; n = 17), adenocarcinoma with NE features (n = 15), and adenocarcinoma with at least two alterations in TP53, RB1, and/or PTEN, consistent with m-AVPC (n = 5). Delta-like ligand 3 surface expression was evaluable in 32 tumors overall and was detected in 56% (18/32), with positivity rates of 73% (11/15) in small-cell tumors, 50% (7/14) in adenocarcinomas with NE features, and 0% (0/3) in tumors meeting only m-AVPC criteria. The population was heavily pretreated (median 3 prior lines; 78% prior platinum; 83% prior AR-targeted therapy), and 56% of evaluable tumors were DLL3-positive by IHC (≥ 1%) [41]. Objective responses occurred exclusively in DLL3-positive tumors (4/18; ORR 22.2%), with no responses in DLL3-negative/unknown tumors (0/22) or in tumors defined only by genomic AVPC-like features without NE marker expression (0/5). Clinical benefit was associated with DLL3 status: median radiographic PFS was 3.5 months in DLL3-positive versus 1.7 months in DLL3-negative/unknown tumors, and median overall survival was 9.8 versus 5.5 months, respectively [41]. Responses were enriched in morphologic small-cell NEPC (3/11 DLL3⁺ small-cell cases), mirroring the preclinical observation that DLL3 expression—and therapeutic sensitivity—tracks with NE lineage state. Preliminary results from a Phase 1 dose-escalation trial of obrixtamig (BI 764532), a DLL3 × CD3 IgG-like T-cell engager (NCT04429087) were reported. The study enrolled patients with DLL3-positive pulmonary and extrapulmonary NE carcinomas (epNEC), including a genitourinary subset [43]. Among 60 patients with epNEC treated with optimized step-up dosing regimens (RB2/RB3), clinical activity was strongly dependent on DLL3 expression. The objective response rate (ORR) was 40.0% in DLL3-high tumors (≥ 50% tumor cells stained) compared with 3.3% in DLL3-low tumors, with a median duration of response of 7.9 months versus 2.8 months, respectively. The disease control rate favored DLL3-high tumors (66.7% vs 26.7%). Notably, responses were most frequent in DLL3-high gastro-enteropancreatic and genitourinary epNEC (ORR 50.0% and 60.0%, respectively) [43]. Preliminary results were also reported for the Phase 1/2 monotherapy study of the DLL3-targeted trispecific T-cell engager gocatamig (MK-6070-001 study, NCT04471727), after enrolling 20 patients with NEPC and 37 with other high-grade NEN (including 12 with small-cell carcinoma of the bladder). All participants were heavily pretreated (median 3 prior lines for NEPC; 93% prior platinum) [42]. Delta-like ligand 3 positivity was required for non-NEPC enrollment but not for NEPC. Across all dose groups, the confirmed ORR was 15% in NEPC and 46% in other NEN; at the 24 mg Q2W target dose, ORR reached 29% in NEPC (n = 7) and 50% in other NEN (n = 12). Median DOR was not reached in either cohort (range 3.7–7.8+ months in NEPC). Delta-like ligand 3 was expressed in 88% of NEPC tumors available for analysis, and, in contrast to tarlatamab and obrixtamig, response did not correlate with DLL3 expression level, a discrepancy that warrants further investigation [42]. Collectively, across tarlatamab, gocatamig and obrixtamig, available evidence indicates that meaningful antitumor activity of DLL3-directed TCEs supports the presence of DLL3 (at least 1% tumor cells stained) as a predictive biomarker for DLL3 TCE therapy in PC; however, whether a gradient effect and exact IHC cut off to optimize patient selection is still unknown. Safety Profiles and Class-Specific Toxicities of DLL3-Targeted Therapies Delta-like ligand 3-ADC‑related toxicities are often driven by the cytotoxic payload (e.g., pyrrolobenzodiazepine or topoisomerase I inhibitors) [52]. Prostate cancer-specific DLL3-ADC safety data remain limited, because NEPC cases were typically embedded within broader NE/solid-tumor cohorts rather than studied in dedicated prostate-only ADC trials. In the Phase 1/2 Rova-T study in DLL3-expressing advanced solid tumors (n = 200 enrolled; 145 treated at 0.3 mg/kg every 6 weeks [q6w]), grade 3/4 AEs occurred in 54% (78/145) and treatment discontinuation due to treatment-emergent adverse events (TEAEs) occurred in 21% (31/145) [40]. At the RP2D (0.3 mg/kg q6w ×2 cycles), grade 3/4 events included anemia (17%), thrombocytopenia (15%), and AST elevation (8%); common all-grade TEAEs included fatigue (52%), nausea (37%), and thrombocytopenia and pleural effusion (33% each) [40]. Importantly, at 0.4 mg/kg q6w (n = 12), 1/12 (8%) experienced a dose-limiting toxicity (DLT) of grade 4 thrombocytopenia, and the cohort included a grade 5 hepatic failure event, supporting selection of the lower RP2D [40]. Delta-like ligand 3-directed T-cell-redirecting therapies have a toxicity profile dominated by on-target immune activation, most notably cytokine release syndrome (CRS), with a smaller but clinically relevant incidence of immune effector cell-associated neurotoxicity syndrome (ICANS). In the Phase 1b DeLLpro-300 study of tarlatamab in metastatic NEPC treated with a one-step regimen (1 mg lead-in then 100 mg q2W), TEAEs occurred in 100% of patients; grade ≥ 3 TEAEs occurred in 45%. Cytokine release syndrome was frequent (82.5% overall) but predominantly grade 1–2; grade 3 CRS occurred in 5%, and there were no grade 4–5 CRS events. Immune effector cell-associated neurotoxicity syndrome or related neurologic events occurred in 12.5%, including one grade 4 event (2.5%) that represented a dose-limiting toxicity and led to discontinuation; there were no fatal TEAEs. Treatment-emergent adverse events led to dose interruption or reduction in 22.5% and treatment discontinuation in 7.5% of patients [41]. In the Phase 2 DeLLphi-301 trial in previously treated SCLC receiving tarlatamab 10 mg Q2W, updated safety analyses reported CRS in 53% of patients, again mostly low grade, and ICANS/neurologic events in 10%; treatment-related AEs led to dose interruption in 15% and discontinuation in 3% [53]. Notably, DeLLpro-300 explicitly reports a numerically higher CRS incidence in NEPC than in DeLLphi-301, and differences in CRS grading criteria (2014 and 2019 Lee criteria) as well as target dose level may contribute to these cross-study discrepancies [41, 54, 55]. In the Phase 1/2 MK-6070-001 study of gocatamig (MK-6070; HPN328) monotherapy (20 patients with NEPC and 37 with other high-grade NEN), any-grade and grade 3–4 TRAEs occurred in 100% and 40% of NEPC patients, respectively, and no treatment-related deaths were reported. Obrixtamig (BI 764532), an IgG-like DLL3×CD3 T-cell engager, has been evaluated in an ongoing Phase 1 dose-escalation trial (NCT04429087) in DLL3-positive pulmonary and extrapulmonary NEC, including a genitourinary subset. Among 60 patients with epNEC treated in optimized step-up regimens (RB2/RB3), safety was again dominated by low-grade immune toxicities. Treatment-emergent adverse events occurred in 100% of DLL3-high and 90% of DLL3-low patients, with grade ≥ 3 TEAEs in 23.3% and 20.0%, respectively. Cytokine release syndrome was reported in 70.0% (grade ≥ 3: 3.3%) of DLL3-high and 60.0% (grade ≥ 3: 3.3%) of DLL3-low patients, while neurotoxicity, including ICANS, occurred in 16.7% (grade ≥ 3: 6.7%) and 10.0% (grade ≥ 3: 3.3%), respectively [43]. Overall, across DLL3-targeted modalities evaluated to date, no consistent or unexpected toxicity pattern uniquely attributable to PC biology has been identified. Observed adverse events largely mirror those seen in SCLC and other high-grade NE malignancies and appear primarily modality- and dose-dependent rather than disease-specific. Notably, ICANS appears to occur in a non-negligible proportion of patients, with grade ≥ 3 events also reported, highlighting the importance of dedicated monitoring and careful neurological surveillance. Emerging Therapeutic Approaches and Ongoing Clinical Development of DLL3-Targeted Treatment in Advanced Prostate Cancer Next‑generation biologics as tri‑specific or multi-specific T‑cell constructs and bispecific antibodies co‑targeting DLL3 plus other antigens are being developed to improve depth and breadth of response and address heterogeneity and antigen escape [25, 56]. However, several challenges may constrain their translation into NEPC. These include the relatively small, heavily pretreated, and biologically heterogeneous patient populations encountered in prostate cancer compared with SCLC, as well as marked spatial and temporal variability in DLL3 expression and substantial phenotypic heterogeneity within and across metastatic lesions [57]. Such complexity may result in mixed target states, including DLL3+/PSMA+ lesions or DLL3+/DLL3− subclones, underscoring the need for dynamic biomarker-guided strategies [37]. In addition, the immunosuppressive microenvironment of advanced PC may limit the efficacy of T-cell-based therapies and support the development of armored constructs or rational combinations. Delta-like ligand 3-directed cellular and emerging immune platforms are designed to deliver sustained, tumor-restricted cytotoxicity by either genetically engineering effector cells to recognize surface DLL3 or by coupling DLL3-binding molecules to cytotoxic or immunomodulatory payloads. Unlike soluble TCEs, these approaches aim to achieve prolonged immune activation, enhanced tumor infiltration, and deeper eradication of DLL3-expressing cell populations [58]. As of now, there are no prostate‑specific DLL3 CAR‑T trials, but these platforms provide a template for future NEPC‑focused programs. Across solid-tumor CAR-T programs targeting DLL3 and related antigens, the principal safety concerns remain CRS and ICANS, driven by in vivo T-cell expansion and cytokine activation [59]. Delta-like ligand 3-targeted radioligand strategies using lutetium-177-labeled anti-DLL3 antibodies significantly prolonged survival in lung and prostate NE carcinoma xenograft models with predominantly mild and largely transient toxicity observed in treated animals, thereby supporting the feasibility of DLL3-directed theranostic strategies [60, 61]. Furthermore, DLL3 PET imaging with ⁸⁹Zr‑labeled antibodies or single-chain variable fragments (scFv) has shown feasibility and safety in NEN and NEPC models and could guide selection and response assessment for DLL3‑directed therapies [44]. Ongoing clinical trials, which have yet to report their findings, are listed in Table 4. The current DLL3 clinical portfolio relevant to advanced PC is still early-stage with one notable translational imaging study aiming to validate in vivo DLL3 detection (111In-ABD147 SPECT) and its correlation with tumor DLL3 by IHC, a potentially important enabler for biomarker-guided selection in a biologically heterogeneous setting (NCT07052760). On the therapeutic side, the most advanced and conceptually consistent platform is DLL3 × CD3 T-cell engagement, represented by multiple ongoing bispecific programs (tarlatamab, peluntamig, obrixtamig, and gocatamig) that include de novo and t-NEPC cohorts while explicit m-AVPC inclusion is reported in a more limited subset (notably NCT07111507); these studies are the most likely to deliver the interpretable prostate signals on safety, dosing, and preliminary efficacy, although many remain basket-enrolled with limited prostate-specific power (NCT07111507, NCT05652686, NCT06893783, NCT04471727, NCT05882058). Beyond TCEs, mechanistic diversification is underway through a DLL3-targeting ADC program (BL-M14D1, NCT07080242) and DLL3 radiopharmaceutical approaches including Lu- and Ac-based constructs (177Lu-DTPA-SC16.56, NCT06941480; 225Ac-ETN029, NCT07006727), with at least one protocol explicitly incorporating AVPC-defined populations (NCT06941480). Notably, the DLL3-targeting ADC zocilurtatug pelitecan (ZL-1310) has demonstrated promising efficacy and a favorable safety profile in heavily pretreated SCLC; a Phase 1b/2 study in patients with NEC is ongoing (NCT06885281), although its relevance to NEPC remains to be defined [62]. Table 4Ongoing/recruiting clinical trials of DLL3-targeted strategies including prostate cancer populationsNCTAgent/modalityStudy design and populationDLL3 requirementProstate cancer inclusionStatusNCT05652686Peluntamig (PT217, DLL3 BiTE) ± chemo/IOPhase 1/2 basket in NEC (SCLC, LCNEC, ep-NEC)DLL3⁺ not requested for eligibilityNot specifiedRecruitingNCT07111507Tarlatamab (DLL3×CD3 Bi-TE)Phase 2 metastatic PC focusedDLL3⁺ requested for eligibility; DLL3 expression defined as ≥ 50% DLL3 tumor positivity by IHC, using CLIA certified assay)Any histologic PC subtype is allowedRecruitingNCT06893783Tarlatamab (DLL3 × CD3 Bi-TE)Phase 2 ep-NEC; t-NEPC are allowed (cohort 2)DLL3⁺ not requested for eligibilityNEPC (both de novo and transformed)RecruitingNCT07006727225Ac-ETN029 (DLL3 radioligand) ± 111In imagingPhase 1 basket in DLL3⁺ NEC (SCLC, LCNEC, ep-NEC)DLL3⁺ requested for eligibility; DLL3 expression defined as ≥ 1% one measurable lesion that shows 111In-ETN029 uptake higher than surrounding tissues on SPECT/CT (assessed by the investigator)NEPC (both de novo and transformed)RecruitingNCT07080242BL-M14D1 (DLL3 ADC)Phase 1 basket in NEC (SCLC, LCNEC, ep-NEC); Merkel cell carcinoma or other NEN are allowed if DLL3⁺;DLL3⁺ not requested for eligibilityNot specifiedRecruitingNCT06941480177Lu-DTPA-SC16.56 (DLL3 radioligand)Phase 1 theranostic in SCLC and NEPCDLL3⁺ required for eligibility; DLL3 expression defined as SUVmax greater than that of normal liver in ≥ 80% of progressing tumor lesions ≥ 2 cmNEPC/AVPC: Includes NEPC (de novo or transformed), m-AVPC, or c-AVPC. Key criteria: histologic evidence of small cell or NE carcinoma, chromogranin positivity, ≥ 2 pathogenic mutations among PTEN, RB1, or p53, visceral-only disease, PSA < 10 with > 20 bone lesions, or DLL3 positivity on available tissueNot yet recruitingNCT07052760111In-ABD147 SPECT (DLL3 imaging)Imaging study in DLL3-expressing solid tumorsDLL3⁺ not requested for eligibility; DLL3 expression assessed by imaging and IHC (correlation analysis planned)CRPC and NEC (including NEPC)Not yet recruitingNCT04429087Obrixtamig (BI764532)DLL3xCD3 IgG like T cell engager (bispecific)Phase 1 basket in DLL3⁺ NEC (SCLC, LCNEC, EP-NEC)DLL3⁺ requested for eligibility; positivity cut-off for enrollment NANot specifiedRecruitingNCT04471727Gocatamig (HPN328) ± IO/I-DXd)Phase 1/2 basket in SCLC, NEPC or DLL3⁺ solid tumorsDLL3⁺ not requested for eligibility in SCLC and NEPCNEPC (both de novo and transformed)Active, not recruitingADCs antibody-drug conjugates, AVPC aggressive-variant prostate cancer, BiTE bispecific T-cell engager, CR complete response, CRS cytokine release syndrome, DLL3 delta-like ligand 3, epNEC extra-pulmonary neuroendocrine carcinomas, GEP gastroenteropancreatic, GU genitourinary, IgG immunoglobulin gamma, IHC immunohistochemistry, LCNEC large cell neuroendocrine carcinoma, mAB monoclonal antibody, NEN neuroendocrine neoplasm, NEPC neuroendocrine prostate cancer, ORR objective response rates, PDX patient-derived xenograft, PTEN phosphatase and tensin homolog, QW every week, SCLC small cell lung cancer, SCNC small cell neuroendocrine prostate cancer Ongoing/recruiting clinical trials of DLL3-targeted strategies including prostate cancer populations Obrixtamig (BI764532) DLL3xCD3 IgG like T cell engager (bispecific) ADCs antibody-drug conjugates, AVPC aggressive-variant prostate cancer, BiTE bispecific T-cell engager, CR complete response, CRS cytokine release syndrome, DLL3 delta-like ligand 3, epNEC extra-pulmonary neuroendocrine carcinomas, GEP gastroenteropancreatic, GU genitourinary, IgG immunoglobulin gamma, IHC immunohistochemistry, LCNEC large cell neuroendocrine carcinoma, mAB monoclonal antibody, NEN neuroendocrine neoplasm, NEPC neuroendocrine prostate cancer, ORR objective response rates, PDX patient-derived xenograft, PTEN phosphatase and tensin homolog, QW every week, SCLC small cell lung cancer, SCNC small cell neuroendocrine prostate cancer Across programs, the main design variable affecting interpretability is how DLL3 selection is operationalized: some trials require DLL3 positivity (by IHC or imaging), one radioligand protocol (with 177Lu-DTPA-SC16.56) requires imaging-based DLL3 positivity, whereas others treat DLL3 status as exploratory, creating cross-trial heterogeneity that can slow convergence on a prostate-transferable threshold and complicate comparisons even if activity is observed (NCT06941480, NCT07111507, NCT07006727). Delta-like ligand 3 could function as both a therapeutic target and a selection framework in lineage-plastic PC but the main near-term risk is that basket-heavy enrollment and limited phenotype-resolved reporting (NEPC vs AVPC; de novo vs t-NEPC) might leave outcomes suggestive rather than decision-ready for prostate-specific sequencing. Challenges in DLL3-Targeted Therapy for Neuroendocrine Prostate Cancer Delta-like ligand 3 is a strong lineage-associated target in NEPC, yet its prostate-specific clinical utility remains limited by several disease-intrinsic and trial-design challenges. Screening for NEPC is difficult, as histologic NE features are not easily identified in routine practice and many patients are recognized only late in the course of rapidly progressive disease, when clinical deterioration and short survival make trial enrollment and completion particularly challenging. In addition, DLL3 expression may itself be dynamic under treatment pressure, potentially switching on or off with lineage plasticity and ongoing therapeutic selection, thereby complicating biomarker-based patient selection. These limitations are further compounded by lung-derived development paradigms, underpowered and biologically heterogeneous prostate cohorts, and non-harmonized biomarker definitions. This chapter outlines a prostate-centered optimization agenda: phenotype-resolved enrollment (de novo vs t-NEPC and clinically aggressive AVPC states), standardized quantitative DLL3 assessment, and endpoints that reflect real-world benefit and deliverability in frail, heavily pretreated men by prioritizing durability, symptom improvement, and feasibility over response rate alone. Finally, it argues for trial designs that move beyond signal-diluting baskets toward prostate-powered, decision-ready cohorts capable of defining selection and sequencing rules—thereby positioning DLL3 agents as credible, practice-shaping options in NEPC and specific AVPC subsets (Fig. 2). Challenges in delta-like ligand 3 (DLL3)-targeted therapy for NE prostate cancer. The clinical utility of DLL3-targeted therapy in NE prostate cancer and aggressive-variant prostate cancer depends on matching the appropriate therapeutic platform to the appropriate disease phenotype at the appropriate time. Major barriers include uncertainty in patient selection across de novo NE prostate cancer (NEPC), treatment-emergent NEPC, and aggressive-variant prostate cancer (AVPC) states; substantial spatial, temporal, and inter-lesional heterogeneity of DLL3 expression; lack of standardized tissue- and imaging-based assays or validated prostate-specific thresholds for therapeutic DLL3 positivity; lineage plasticity and antigen escape; immunosuppressive microenvironmental constraints; and platform-specific limitations affecting antibody-drug conjugates, T-cell engagers, and radiopharmaceuticals. The figure also highlights trial design priorities for future prostate-specific development, including phenotype-resolved enrollment, harmonized biomarker assessment, serial biomarker integration, treatment-history capture, and endpoints that emphasize durability, feasibility, symptom benefit, hematologic recovery, and retreatment potential Patient Selection Most DLL3-enabled prostate trials pragmatically pool rare de novo NEPC with t-NEPC and only a limited number broaden eligibility to AVPC. Aggressive-variant prostate cancer sits at the interface between “clinically recognizable aggressiveness” and true NE lineage biology, and it is therefore an imperfect surrogate for DLL3 targetability. The overlap between NE histology and molecular AVPC is incomplete in both directions: only about 35% of NEPC are m-AVPC, and only about 27% of m-AVPC are NEPC, meaning that AVPC labels alone cannot substitute for contemporaneous NE/DLL3 confirmation when the therapeutic mechanism is antigen-directed [63–66]. In practice, this is further complicated by the fact that screening for NEPC is difficult: histologic NE features are not always readily identified, repeat metastatic biopsy is often challenging, and the rapid clinical deterioration typical of these patients narrows the window for referral, molecular work-up, and trial enrollment. Accordingly, AVPC criteria could be best used as triage tools to raise suspicion and accelerate referral when repeat biopsy is difficult, but current DLL3 status should be confirmed on contemporary metastatic tissue whenever feasible. Biologically, de novo NEPC and t-NEPC are convergent but non-identical. They share a core NE lineage/genomic program with near-ubiquitous TP53/RB1 loss, frequent PTEN loss, activation of NE lineage transcription factor networks (e.g., ASCL1/BRN2/SOX2/FOXA2) and reinforcement by epigenetic regulators such as EZH2/PRC2, while transcriptomic and single-cell frameworks place NEPC on a continuum from AR-high adenocarcinoma toward AR-low NE states enriched for proliferative/replication-stress programs and canonical NE markers (e.g., CHGA, SYP, PEG10) [10, 15, 67–70]. Where this matters for trial design is heterogeneity. In t-NEPC, prospective data indicate these tumors retain early adenocarcinoma events, with AR amplification in ~ 67% and AR protein expression in ~ 75% in a treatment-emergent small-cell series, supporting a more mosaic architecture in which DLL3 may mark only part of the metastatic burden [71]. This is a strong rationale to pre-specify and report outcomes separately for de novo NEPC versus t-NEPC. Finally, SCLC cannot be used as a simple translational proxy for NEPC even though they share major NE programs and DLL3 is tightly linked to ASCL1-driven states [10, 36, 70, 72, 73]. Small cell lung cancer appears to more commonly include non-NE/NOTCH-active and immune-interactive fractions [67, 72–74], whereas prostate NEPC is among the most immune-depleted tumor states; in one immunogenomic analysis only 9/36 (25%) NEPC were CD8+ T-cell-inflamed [74]. In a multi-cohort analysis, DLL3 was positive in about 77% of CRPC-NE biopsies but only 12.5% of aggressive CRPC-adenocarcinoma, supporting DLL3-gated enrollment and arguing against broad “aggressive” inclusion without objective DLL3/NE evidence [10, 28, 71]. This becomes even more important in t-NEPC, where state-linked heterogeneity, temporal acquisition, and the potential plasticity of DLL3 expression under treatment pressure may limit the reliability of archival tissue for patient selection. Quantifying DLL3 Expression Delta-like ligand 3-directed therapeutics require adequate target density and sufficiently uniform expression of DLL3 in the tumor [32]. A key constraint in detecting DLL3 expression by IHC is antibody choice as cross-clone concordance is poor. Currently Ventana SP347 is the best-supported decision-grade assay because it has been validated against DLL3-positive and DLL3-negative controls, and dominates the standardized SCLC evidence base [75]. In SCLC, SP347 has been used across major therapeutic development programs and large real-world cohorts. In NEPC, most trial-relevant clinical work has converged on SP347-on-Ventana instruments, whereas earlier biologic/proof-of-concept studies frequently used non-Ventana clones (e.g., SC16.56/SC16.65) that are not interchangeable with SP347 for prevalence comparisons [37, 44]. Therefore, decision-grade DLL3 testing in both SCLC and NEPC should preferentially use SP347 on validated Ventana platforms with central or tightly controlled local testing; other clones should remain exploratory until formally cross-validated versus SP347 using standardized interchangeability studies. In the largest SCLC SP347 real-world cohort (N = 1050), DLL3 positivity was 85% at ≥ 25% (895/1050) and “high” DLL3 was 68% at ≥ 75% (719/1050). These distributions explain why many SCLC trials could enroll patients broadly and still retain analytic power for DLL3-high subgrouping [76–79]. In NEPC, the numerical picture is less mature and harder to compare across studies because scoring systems vary and high cut-off strata are often not reported and the prevalence of DLL3-high disease at clinically meaningful cut-offs remains largely unquantified [80]. This uncertainty blocks evidence-based adoption of SCLC-derived thresholds in PC and makes enrichment strategy harder to optimize prospectively. In NENs, DLL3 expression appears relatively stable across disease sites. Delta-like ligand 3 positivity is highly concordant between primary tumors and matched metastases (~ 92 to 93%) and apparent stability has been shown across therapy lines [76, 81]. However specific NEPC data are less consistent, on one side lesion-level discordance has been documented -including cases with DLL3-negative metastases by both IHC and immuno-PET while other lesions are positive—consistent with greater spatial complexity and, particularly in t-NEPC, temporal evolution along a plasticity trajectory [26, 44, 82]. This spatial heterogeneity is further supported by first-in-human immuno-PET imaging with [⁸⁹Zr]Zr-DFO-SC16.56, which showed markedly variable tracer uptake across lesions within individual patients, with SUVmax values ranging approximately from 3 to 67. Non-avid lesions were uniformly DLL3-low by IHC, indicating that different tumor sites in the same patient may express substantially different levels of DLL3 surface antigen [44–46]. These findings have direct therapeutic implications: in patients with mixed DLL3-positive and DLL3-negative disease, DLL3-directed therapies may control antigen-expressing lesions while antigen-negative sites progress, resulting in divergent responses across disease sites. Bystander killing by T-cell engagers may partially mitigate this limitation, as DLL3-directed TCEs have shown activity against adjacent DLL3-negative tumor cells in preclinical NEPC models [32]. Interestingly, concordance has been shown between metastatic biopsies and CTCs [34], supporting its use for biomarker-driven trial design and longitudinal monitoring. Therefore, in NEPC studies a recent metastatic formalin-fixed paraffin-embedded (FFPE) biopsy should be preferred when feasible, specimen age and prior therapies should be recorded, and re-biopsy considered when archival tissue is old or clinical behavior suggests NE evolution. Molecular imaging with Zirconium-89-labeled anti-DLL3 antibodies or single-chain fragments ([⁸⁹Zr]-DFO-SC16, [⁸⁹Zr]-DFO-DLL3-scFv) enables PET imaging that selectively visualizes DLL3-positive NEPC lesions while largely sparing AR-dependent and non-NEPC lesions [37, 83]. These tracers can function as non-invasive biomarkers of target expression, aid in patient selection, and support pharmacodynamic assessment. First-in-human imaging with [89Zr]Zr-DFO-SC16.56 in SCLC and NEPC demonstrated feasibility and high PET–IHC concordance (about 90–94% of evaluable patients), with highly heterogeneous uptake (SUVmax approximately 3–67); importantly, non-avid tumors were uniformly DLL3-low by IHC, supporting PET’s ability to identify biologically meaningful target absence [44–46]. In early SCLC studies, tumor-to-background ratios > 5 were associated with response to DLL3-ADC [84]. However, PET is not yet interchangeable with IHC because SUV-based positivity thresholds and their linkage to therapeutic efficacy are not standardized; therefore, PET should currently be treated as complementary with IHC remaining the eligibility anchor until predictive PET metrics are prospectively validated. In this framework, PSMA PET may also have value as a selection tool, as PSMA-low or PSMA-negative cases could define a clinically relevant subgroup in which complementary DLL3 PET may help identify candidates for DLL3-targeted strategies. Overcoming Resistance to DLL3-Targeted Therapies The main resistance mechanism for DLL3-targeted therapies in PC is likely to arise from tumor composition and spatial heterogeneity: NEPC/AVPC can behave as a mosaic of DLL3-high NE clones and DLL3-low/negative populations [32, 37, 79, 80]. A DLL3 agent may control some lesions while other lesions progress, resulting in patient-level primary resistance even when target expression is “present” somewhere. Thus, multi-site assessment and in vivo mapping (DLL3 imaging) are important to both patient selection and response interpretation. Class-specific resistance may be less of a problem with DLL3 TCEs/bi-specifics and mixed-antigen settings through bystander killing, supporting the rationale that TCE-like mechanisms can be more heterogeneity-tolerant than payload-delivery modalities, which require stable surface antigen per cell [31, 37]. Acquired resistance via antigen modulation was demonstrated in a small-cell NE carcinoma PDX treated with AMG 757, which had late relapses with reduced DLL3 expression, consistent with selection pressure driving DLL3 downregulation [37]. Resistance to DLL3 ADCs may emerge via antigen loss or altered trafficking/drug efflux, as seen with Rova-T in SCLC [32]. Moreover, lessons from SCLC treated with Rova-T indicate that DLL3 membrane stability could be low after chemotherapy (TRINITY trial), making “chemo + ADC” biologically vulnerable [84]. Resistance to DLL3 radioligand/radioimmunotherapy could arise from the therapeutic window and marrow reserve because NEPC/AVPC patients are frequently cytopenic from prior platinum and diffuse marrow involvement, creating a real “resistance-to-delivery-intensity” problem that must be engineered around in trial design and sequencing [60, 83]. Combination strategies could be relevant to overcoming resistance. Because direct evidence on combination strategies in NEPC is still lacking, current considerations on how to overcome resistance to DLL3-targeted therapies rely largely on data from SCLC. However, SCLC and NEPC remain distinct diseases with different molecular contexts and tumor microenvironments; therefore, SCLC findings should be viewed as hypothesis-generating rather than directly translatable to NEPC. Preclinical work on SCLC showed that adding anti-PD-1 dramatically improves efficacy of a DLL3 bispecific/TCE, while providing no added benefit when combined with DLL3 CAR-T [85]. Importantly, the Phase 1b DeLLphi-303 study testing tarlatamab plus PD-L1 inhibitor as maintenance after first-line chemo-immunotherapy showed a manageable safety profile with promising anticancer activity, resulting in the ongoing Phase 3 trial for SCLC [86]. By contrast, chemotherapy-based combinations raise greater concern. In SCLC PDX models, chemotherapy reduced DLL3 membrane localization and impaired Rova-T binding, and in the TRINITY trial fewer than 30% of post-chemotherapy biopsies retained stable DLL3 expression [84]. Rova-T combinations with nivolumab plus ipilimumab (NCT03026166) or with cisplatin/etoposide (NCT02819999) were also discontinued because of toxicity and lack of clear added benefit [18, 84]. Together, these data provide an important negative translational signal for concurrent chemo- or immunotherapy combinations, especially with first-generation ADC approaches. Conversely, next-generation DLL3 ADC constructs may better address these limitations. Given that SEZ6 and B7-H3 are also expressed in NEPC, and that B7-H3 is additionally expressed in CRPC/adenocarcinoma, such dual-target strategies may be particularly attractive for heterogeneous AVPC/NEPC states [87, 88]. In preclinical SCLC models, DLL3 × SEZ6 and DLL3 × B7-H3 bispecific ADCs carrying a TOP1 inhibitor payload showed enhanced internalization and robust antitumor activity, with the DLL3 × B7-H3 construct also outperforming monospecific comparators and DS-7300-based approaches in vivo [89, 90]. Overall, these findings suggest that dual-target and next-generation DLL3-directed platforms may help overcome resistance related to antigen heterogeneity, reduced target accessibility, or impaired internalization, but their relevance to NEPC still requires validation in disease-specific preclinical models. Trial Design Priorities for NEPC/AVPC Delta-like ligand 3 development in NEPC/AVPC should move away from lung-centric, tumor-agnostic basket studies toward prostate-specific trials capable of generating clinically actionable evidence for patient selection and treatment sequencing. Because NEPC/AVPC is biologically heterogeneous, interpretability requires mandatory prospective phenotypic stratification rather than descriptive labeling, including separation of AVPC states (c-AVPC and m-AVPC) from histologically confirmed NEPC, as well as de novo versus t-NEPC. These strata should be incorporated into randomization and interaction analyses. Delta-like ligand 3 assessment should be harmonized through centralized IHC using continuous reporting metrics, including H-score, percentage of positive cells, and a prespecified threshold for DLL3-high disease. This issue is particularly relevant in t-NEPC, which is generally more heterogeneous and may show marked spatial and temporal variation in DLL3 expression, together with mixed subclonal states such as DLL3+/PSMA+, DLL3+/DLL3-, or DLL3+/STEAP1+ populations. In this setting, evaluating DLL3 alongside additional markers such as PSMA and STEAP1 may provide a more complete view of disease biology and help refine patient selection for biomarker-directed therapies [37, 44]. A critical unanswered question is the optimal position of DLL3-targeted therapy within the treatment sequence for advanced PC; specifically, whether it should be deployed prior to, instead of, or after platinum-based chemotherapy. Current evidence remains insufficient to draw definitive conclusions, but several considerations inform this question. First, platinum-based chemotherapy (typically carboplatin or cisplatin plus etoposide) remains the established standard of care for de novo and treatment-emergent NEPC, and no DLL3-directed agent has yet demonstrated superiority or non-inferiority to chemotherapy in a randomized trial [4, 15]. Second, DLL3 surface expression is tightly linked to the ASCL1-driven NE transcriptional state, as tumors are exposed to successive lines of therapy and undergo phenotypic evolution, including partial or complete loss of NE identity under therapeutic selective pressure. As a result, DLL3 expression may become heterogeneous or attenuated at the clonal level, potentially narrowing the therapeutic window for DLL3-directed agents in heavily pre-treated patients [28, 32]. This provides a biological rationale for earlier deployment before extensive phenotypic drift occurs. Third, patients who have received multiple prior lines of chemotherapy and/or alpha- or beta-emitting radioligand therapies have limited marrow reserve, which may constrain the tolerability of myelosuppressive DLL3-directed regimens and reduce feasibility of full-dose delivery. Taken together, these considerations favor exploring DLL3-targeted therapy either in earlier treatment lines (e.g., after ARPI failure but before or concurrent with first platinum exposure, or as maintenance after platinum response) or as an alternative to platinum rechallenge in patients with prior platinum exposure. The Phase 1b DeLLphi-303 experience with tarlatamab as maintenance after chemo-immunotherapy in SCLC provides a potential precedent for this maintenance strategy that warrants prospective evaluation in NEPC. Future prostate-specific trials should pre-specify the treatment-line context and capture prior platinum and radioligand therapy exposures to enable meaningful sequencing analyses. Endpoints should be designed around what makes DLL3 strategies practice-changing in a frail population and not ORR alone, such as durability, symptoms, and feasibility. The SCLC tarlatamab experience supports elevating DOR and fixed-time PFS (6 and 12 months) as key endpoints, with parallel capture of pain and quality-of-life measures (including the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire-Core 30 and the Quality of Life Questionnaire Core 13) to demonstrate patient-relevant benefit [53, 91]. Feasibility should be treated as efficacy-linked, including cycle completion (≥ 4 to 6), unplanned hospitalizations, and dose modifications or discontinuations due to CRS, neurotoxicity, or myelosuppression; radioligand studies should additionally quantify hematologic recovery, cumulative delivered dose, and retreatment feasibility [44, 45]. In short, prostate-specific utility will come from powered prostate cohorts, phenotype-resolved enrollment, harmonized biomarkers, and endpoints that prove durable, deliverable benefit, not from extrapolation across histologies. Conclusion Delta-like ligand 3 is a strong lineage-associated target in NEPC, yet prostate-specific clinical utility remains limited by lung-derived development paradigms, underpowered clinical trials, biologically heterogeneous prostate cohorts, and non-harmonized biomarker definitions. Delta-like ligand 3 prevalence has not been systematically characterized across the full spectrum of AVPC. No validated thresholds currently define biologically or therapeutically meaningful “DLL3-high” PC. Nevertheless, clinical trials of DLL3-targeted therapies show promising results and are redefining the management of DLL3-expressing PC. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (PDF 323 KB) Funding Open access funding provided by Università della Svizzera italiana. No external funding was used in the preparation of this manuscript. Declarations Conflict of interest Silke Gillessen received personal honoraria for participation in advisory boards for Sanofi, Orion, Roche, Amgen, MSD; other honoraria from RSI (Televisione Svizzera Italiana); invited speaker for ESMO, Swiss group for Clinical Cancer Research (SAKK), Swiss Academy of Multidisciplinary oncology (SAMO), Orikata academy research group, China Anti-Cancer Association Genitourinary Oncology Committee (CACA-GU); Speaker’s bureau for Janssen Cilag; travel grant from ProteoMEdiX; institutional honoraria for advisory boards for Bayer, Janssen Cilag, Roche, AAA International including Independent Data Monitoring Committee and IDMC and Steering Committee member for Amgen, Menarini Silicon Biosystems, Astellas Pharma, Tolero Pharmaceuticals, MSD, Pfizer, Telixpharma, BMS and Orion; patent royalties and other intellectual property for a research method for biomarker WO2009138392. Fabio Turco and Martino Pedrani received a travel grant from Bayer. Gunhild von Amsberg received personal honoraria for consultancy, advisory boards or discussion of off-label drug use from Amgen, Astellas, AstraZeneca, Bayer, BMS, Ipsen, Janssen (Johnson & Johnson), Lilly, MSD, Merck, Novartis, Pfizer, Roche. She received institutional research funding for industry sponsored clinical trials by Roche Pfizer, BMS, MSD/Merck, AstraZeneca, Pfizer, Exelixis, Ipsen, Lilly, Novartis, AvenCell, Bayer, Astellas. Himisha Beltran has served as consultant/advisory board member for Janssen, Merck, Pfizer, Amgen, Bayer, Daiichi Sankyo, Astra Zeneca, Novartis, Abbvie, and has received research funding (to institution) from Janssen, Bristol Myers Squibb, Circle Pharma, Daiichi Sankyo, Novartis, Abbvie. Markus Joerger received institutional honoraria for advisory roles from Novartis, Astra Zeneca, Bayer, BMS, MSD, Roche, Sanofi; research funding from Swiss Cancer Research; travel grants from Roche, Sanofi, PharmaMar. Tarek Taha reports institutional support for attending meetings and/or travel from Pfizer, and receiving honoraria from BMS, MSD, Merck-Serono and Pfizer. Ilaria Colombo declares institutional funding for clinical trials as PI from AstraZeneca, Merck Sharp & Dohme, Vivesto, Tolremo, Orion, Bayer, Incyte, Debiopharm, PharmaMar; personal honoraria for consultancy for Biontech; institutional honoraria for consultancy/advisor role from AstraZeneca, GlaxoSmithKline, Merck Sharp & Dohme, AbbVie, Incyte, Beigene. Fabio Turco reports payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Silvio Grasso Consulting (institutional), SAKK (institutional), Merck (institutional), Novartis (institutional), Bayer (individual) and MSD (individual); support for attending meetings or travel from Bayer (individual); and participation on a data safety monitoring board or advisory board for Bayer (individual) and Telix (individual). Giovanna Pecoraro, Alberto De Giorgi, Giulia Montelatici, Giuseppe Salfi, Hui-Ming Lin, Johann de Bono, Rahul Raj Aggarwal, and Martina Imbimbo declare the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Ethics approval Not applicable. Consent to participate Not applicable. Consent for publication Not applicable. Availability of data and material Not applicable. Code availability Not applicable. Author Contributions Study concept and design: Pedrani, Pecoraro, Gillessen. Acquisition of data: Salfi, Pedrani, Pecoraro, Montelatici, De Giorgi. Analysis and interpretation of data: Pedrani, Pecoraro, Gillessen. Drafting of the manuscript Pecoraro, Pedrani, Lin. Critical revision of the manuscript for important intellectual content: Turco, Taha, Beltran, de Bono, Aggarwal, von Amsberg, Joerger, Imbimbo, Colombo, Gillessen, Pedrani. Statistical analysis None. Obtaining funding None. Administrative, technical, or material support None. Supervision Pedrani, Gillessen. References