Preclinical activity of the DLL3-targeted T cell engager MK-6070 in neuroendocrine prostate cancer

Abstract

Neuroendocrine prostate cancer is an aggressive variant of prostate cancer with limited therapeutic options. Delta-like ligand 3 (DLL3) is a cell surface protein and therapeutic target expressed in the vast majority of NEPC tumors. The DLL3-targeted T cell activating construct MK-6070 (formerly called HPN328) binds to both DLL3 on tumor cells and CD3 on T cells, as well as serum albumin to extend half-life. A phase 1/2 trial of MK-6070 is currently underway which includes an NEPC cohort (NCT04471727). Here we report the preclinical activity of MK-6070 in prostate cancer models, showing high specificity and anti-tumor activity in DLL3-expressing NEPC models both in vitro and in vivo, with T cell activation and tumor infiltration of T cells after treatment. MK-6070 also demonstrates anti-tumor activity in mixed tumors, impacting DLL3-negative prostate cancer cells after engagement with surrounding DLL3-expressing tumor cells, supporting a potential bystander effect. Overall, these data demonstrate promising activity of MK-6070 in NEPC preclinical models including heterogeneous tumors, supporting the clinical development of MK-6070.

Introduction

Neuroendocrine prostate cancer (NEPC) is an aggressive histologic variant of prostate cancer that rarely arises de novo but more commonly develops in patients previously treated for prostate adenocarcinoma as a mechanism of treatment resistance(1). The diagnosis of NEPC is based on tumor morphology demonstrating poorly differentiated neuroendocrine (NE) carcinoma, most commonly small cell carcinoma, histologically similar to other high grade NE carcinomas(2). Since treatment-emergent NEPC arises from a prostate adenocarcinoma precursor(3), mixed NE and adenocarcinoma tumor features can also be observed clinically(2). Despite an increased recognition in recent years, the management of NEPC (pure or mixed) remains challenging and prognosis is poor. Platinum-based chemotherapy is standardly given for NEPC and supported by guidelines(4) but responses are often short lived. Next line therapy after platinum-based chemotherapy is typically extrapolated from small cell lung cancer (SCLC) guidelines(4), as there have been very few prospective clinical trials for NEPC and no approved therapies. The delta-like ligand 3 (DLL3)-targeted T cell engager tarlatamab was recently approved by the United States Food and Drug Administration (FDA) for the treatment of patients with extensive stage SCLC after progression on platinum chemotherapy(5, 6), raising the question of whether a similar approach should also be used for NEPC.

DLL3 is a cell surface protein expressed in the vast majority of SCLC, NEPC, and other poorly-differentiated NE carcinomas, and not expressed in normal healthy adult tissues(7), supporting DLL3 as a potential pan-NE therapeutic target. High levels of DLL3 have been associated with advanced disease and poor prognosis across NE carcinomas(7). We previously reported up to 76% of metastatic NEPC tumors strongly express DLL3 and there is limited expression in prostate adenocarcinoma or benign prostate tissues(8). DLL3 is a negative regulator of Notch signaling, and Notch signaling is downregulated during prostate cancer lineage plasticity and the transition of adenocarcinoma towards NEPC; this corresponds with upregulation of both DLL3 and achaete-scute homolog 1 (ASCL1)(9). ASCL1 is an important lineage determining transcription factor for neuronal development and a key driver of neuroendocrine differentiation(10); ASCL1 can further drive DLL3 expression at the transcriptional level(11).

T-cell engager therapy to target cell surface antigens is an area of rapid clinical development in oncology(12). By bringing antigen-specific cancer cells and T cells into close proximity, this drug mechanism drives synapse formation and T cell activation to induce tumor cytotoxicity independent of MHC-I recognition(13, 14). Immune responses are therefore expected even in cancers without identifying neoantigens such as NEPC(15). MK-6070 (formerly HPN328) is a DLL3-targeted tri-specific T cell activating construct (TriTAC) that binds to DLL3 on tumor cells, CD3ε on T cells, and human serum albumin to extend half-life (Fig 1A). In an ongoing Phase 1/2 clinical trial of MK-6070 (NCT04471727), there are dedicated cohorts for SCLC, NEPC, and other DLL3-expressing NE carcinomas(16). In this study, we investigated the specificity and anti-tumor activity of MK-6070 in preclinical models of prostate cancer.

Figure 1. In vitro cytotoxicity of MK-6070.

(A) Schematic of MK-6070 construct and mechanism of action of T cell engager mediated cytotoxicity in DLL3-positive tumor cells. (B) DLL3-positive NCI-H660, (C) DLL3-negative 22Rv1, and (D) DLL3-positive 22Rv1DLL3 were used as representative NEPC and prostate adenocarcinoma models to evaluate in vitro cytotoxicity of MK-6070. 3,000 tumor cells were co-cultured with T-cells in a 1:10 ratio from independent healthy donors. Various concentrations of TriTAC-GFP (control) antibodies and MK-6070 were added in the presence of 15 mg/ml human serum albumin for 7 days. Relative growth rate was determined by Cell Titer-Glo and then normalized to no antibody control (0 nM). Experiments were conducted three times independently.

Methods

Human T cell isolation

Whole blood from healthy human donors was collected in leukophoresis blood collars by the Brigham and Women’s Hospital (BWH) Crimson Core (protocol no: T0781) and spun in Ficoll-Paque density medium to separate peripheral blood mononuclear cells (PBMCs). Twenty million PBMCs were cultured in T cell media with IL-2 (30U/ml, ThermoFisher) and CD3 (100ng/ml, BioLegend) for three days to stimulate T cell population(17), then T cells were cultured with IL-2 only for another 4–7 days.

Cell culture and growth assay

The 22Rv1 (RRID:CVCL_1045), C4–2 (RRID:CVCL_4782), and NCI-H660 (RRID:CVCL_1576) male prostate cancer cell lines were purchased from the American Type Culture Collection (ATCC) in 2019 and 2020. All cell lines were only in culture for less than 6 months or 20 passages. Cell authentication was performed using STR analysis (ATCC) and cells were routinely tested for mycoplasma contamination (InvivoGen). For growth assays, 3,000 cells per well were plated in a 96-well plate for 7 days. Relative cell growth was measured by CellTiter-Glo (Promega) per the manufacturer’s protocol and normalized to Day 1. All growth experiments were conducted at least twice biologically with multiple technical replicates.

MK-6070.

MK-6070 (formerly referred to as HPN328) structure, drug properties, and pharmacokinetics have been previously described (18).

Flow Cytometry

30,000 tumor cells and T cells were co-cultured in at varied ratios in the presence of 15mg/ml human serum albumin (Fujifilm Irvine Scientific) and either TriTAC-GFP antibody or MK-6070 for 2 days in a 24-well plate. Both tumor and T cells were collected and washed with cell staining buffer (BioLegend) twice and then resuspended in cell staining buffer. Cells were incubated with Fc Receptor Blocking Solution (BioLegend) for 5 minutes. FITC anti-human CD3 (1/500, BioLegend), APC anti-human CD25 (1/100, BioLegend), PE anti-human CD69 (1/100, BioLegend), BV421 anti-human PD1 (1/100, BioLegend) antibodies were added and then incubated in dark at room temperature for 15 minutes. Stained cells were then washed and resuspended in cell staining buffer (BioLegend). For DLL3 expression analysis, cells were fixed with 4% PFA for 20 minutes at room temperature and washed with cold PBS once. Cells were incubated with Fc Receptor Blocking Solution (BioLegend) for 10 minutes and then anti-DLL3 antibody (1/100, #71804, Cell Signaling Technology) for 20 minutes on ice. Alexa 488 secondary antibodies were added and incubated in dark for 20 minutes. To measure granzyme B levels in 22Rv1-RFP cells, cells were fixed in Fixation Buffer (BioLegend). Cells were then resuspended in Intracellular Staining Perm Wash Buffer (BioLegend) and stained with BV421 anti-granzyme B antibody (BioLegend) for 20 minutes at room temperature. Stained cells were washed and resuspended in the cell staining buffer (BioLegend). Flow cytometric analysis was performed using the LSRFortessa machine (BD Biosciences).

Immunoblot

Whole-cell lysates were prepared using RIPA buffer (Sigma-Aldrich), supplemented with a protease inhibitor cocktail (Roche) and 1× phosphatase inhibitor (ThermoFisher Scientific). Lysates (50 μg per sample) were separated by SDS-PAGE using Mini-PROTEAN^® TGX^™ gels (Bio-Rad), followed by transfer onto nitrocellulose membranes (Bio-Rad). Membranes were blocked with 5% Blotting-Grade Blocker (Bio-Rad) and incubated with anti-DLL3 (1/1000, #71804, Cell Signaling Technology), anti-INSM1(1/2000, SC-377428, Santa Cruz), anti-Vinculin antibodies (1/5000, #13901, Cell Signaling Technology) overnight at 4°C. Detection was performed using horseradish peroxidase (HRP)-conjugated secondary antibodies (Bio-Rad), and signal was visualized using Immobilon^® Classico Western HRP substrate (Millipore).

Immunohistochemistry

Immunohistochemistry was conducted on the Leica Bond III automated staining platform. Briefly, 4-μm thickness FFPE slides were sectioned, deparaffinized in xylene solution, and gradually rehydrated in ethanol. DLL3 staining was performed using Ventana DLL3 SP347 assay kit; anti-INSM1(Santa Curz, clone A-8) and anti-AR(Dako, AR441) antibodies were subjected to detect INSM1 and AR expression. Slides were visualized vis DAB and counterstained with hematoxylin. Slides were imaged by the NIS-elements imaging system (Nikon) fitted to a Nikon ECLIPSE Ti2 microscope.

Cytokines

IFN-γ, TNF-α, soluble granzyme B were measured using LEGEND MAX^™ ELISA kits (BioLegend) following manufacturer’s protocol. Tumor cells and T cells were co-cultured in the presence of 15mg/ml human serum albumin and either TriTAC-GFP antibody or MK-6070 for 2 days in a 24-well plate. Supernatants were collected and subjected to ELISA analysis. Two independent studies with technical duplicates were conducted.

In vivo studies

Castrated eight-week-old male NOD Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice (The Jackson Laboratory) were used for the entire study. WCM154 and WCM12 patient-derived xenograft models of NEPC were previously described(19, 20). PC7 was developed in the Balk Laboratory at BIDMC. 23-Rv-Lv was established from a liver metastatic lesion of a rapid autopsy (protocol no: 15–441; PI: David Einstein). The 22Rv1 xenograft model was established by injecting 5 million cells subcutaneously in the right flank. To evaluate anti-tumor activity of MK-6070, 2–5mm single tumor pieces were subcutaneously implanted in the right flank of castrated NSG mice. When the tumor size reached 100mm³ (tumor kinetics varied, depending on the individual model), mice were randomized. Twenty million pre-cultured CD3+ human T cells were intravenously injected in NSG mice intravenously prior the beginning of treatment. The following day, mice were treated with 100 μg/kg of either TriTAC-GFP (control antibody) or MK-6070 with 4% human serum albumin intravenously, twice per week, for a total of ten doses. MK-6070 was prepared weekly in phosphate buffered saline. After the completion of treatment, mice were kept in house until they reached the humane endpoints either by tumor size or body condition per IACUC guidelines. Tumor size was measured by a digital caliper twice a week and calculated using the following formula: volume (V) = length (L)² × width (W) × 0.5. Body weight was measured twice per week to monitor adverse effects. All experiments were performed in compliance with IACUC guidelines at Dana-Farber Cancer Institute (protocol no. 18–020).

Data Availability

Data generated in this study are available within the article and its supplementary data files.

Results

Cytotoxic activity of MK-6070 in vitro

To obtain human CD3+ T cells, we isolated peripheral PBMCs from whole blood of healthy donors and cultured them with an anti-CD3 antibody and IL-2 for 7–14 days in vitro to enrich for CD3+ T cell populations (Supp Fig1A–B). We then co-cultured T cells with the DLL3-expressing NEPC cell line NCI-H660 at various effective T cell/target cell (E/T) ratios in the presence of either TriTAC-GFP (control antibody) or MK-6070 (Fig 1A) and evaluated cell viability. A ratio of 10:1 E/T had the greatest suppression of cell viability (Supp Fig1C–E) and was applied to the following studies. We further tested MK-6070 activity in NCI-H660 with different healthy donor T cells showing that NCI-H660 cells consistently responded to MK-6070 in a dose-dependent manner (IC50 = 0.45–1.23nM) (Fig 1B). This was, in contrast, to the DLL3-negative castration resistant prostate adenocarcinoma cell lines 22Rv1 and C4–2 (Supp Fig1C–D), which did not respond to MK-6070 (Fig 1C & Supp Fig1F). We introduced exogenous DLL3 in 22Rv1 cells to generate an isogenic model; 22Rv1DLL3 responded to MK-6070 in a dose-dependent manner comparable to NCI-H660 (Fig 1D & Supp Fig1C–D). Overall, these results support high specificity of MK-6070 in targeting DLL3-expressing prostate cancer cells.

To examine for T cell activation, we measured CD25, CD69, and PD-1 levels after MK-6070 treatment. Each of these markers was upregulated in CD3+ T cells that were co-cultured with DLL3-expressing NCI-H660 and 22Rv1DLL3 cells, but not with DLL3-negative 22Rv1 cells (Fig 2A & Supp Fig2A&C). Moreover, cytokine levels (IFN-γ, TNF-α, soluble granzyme B) were also upregulated in a dose dependent manner in both NCI-H660 and 22Rv1DLL3 conditions but not in 22Rv1 (Fig 2B & Supp Fig2B). These data support MK-6070 activation of CD3+ T cells that may selectively mediate cytotoxicity in DLL3-expressing cells by secreting IFN-γ, TNF-α, and granzyme B.

Figure 2. T cell activation by MK-6070.

(A) T cells activation markers, CD25, CD69, and PD1, were measured by flow cytometry upon engagement of H660, 22Rv1, 22Rv 1DLL3 cells. (B) IFN-γ, TNF-α, and soluble granzyme B levels were determined by ELISA. 30,000 tumor cells were co-cultured with T cells in a 1:10 ratio in the presence of various concentrations of either TriTAC-GFP or MK-6070 and 15mg/ml human serum albumin for 2 days. Cells were collected for FACS analysis and CD3 was used to gate T cells. Supernatants were harvested to measure cytokines. Three independent experiments were performed.

Cytotoxic activity of MK-6070 in vivo

The patient-derived xenograft (PDX) models of NEPC WCM154 and WCM12 homogeneously express high levels of DLL3 and the PC7 NEPC PDX model expresses heterogeneous levels of DLL3; all three models express NE markers and lack androgen receptor (AR) expression (Fig 3A & Supp Fig3A). The 22Rv1 adenocarcinoma xenograft model is DLL3-negative and AR-positive (Fig 3A & Supp Fig3A). We tested the anti-tumor activity of MK-6070 in these xenograft models. CD3+ human T cells were injected in NSG mice intravenously prior the beginning of treatment. The following day, TriTAC-GFP (control) or MK-6070 was administrated intravenously twice per week for a total of 10 treatments. WCM154 NEPC tumors began shrinking after the second dose, fully regressed after the 4th or 5th doses (Supp Fig3B), then remained tumor-free for over six months after the completion of therapy. MK-6070 significantly improved survival of the WCM154 model compared with the control group (Fig 3B–D). No significant adverse events were observed. Similar to the WCM154 model, the NEPC model WCM12 also demonstrated rapid regression after MK-6070 and had prolonged the survival without significant toxicity (Supp Fig3C–E). Notably, after initial regression, WCM12 tumors relapsed after 5–6 weeks of treatment completion, which was not seen in the WCM154 model. This relapsed model (WCM12-relapsed) still expressed high and homogeneous levels of DLL3 which were comparable to parental untreated WCM12 tumors (Fig 3A & Supp Fig4A). To test if WCM12 relapsed tumors still respond to MK-6070, we engrafted relapsed tumors in a cohort of mice and re-treated with MK-6070. The majority of mice with relapsed tumors responded again to MK-6070 with tumor regression after 4 doses (Supp Fig4B); however, two mice did not respond to MK-6070, suggesting that these tumors might have acquired resistance to MK-6070 (Supp Fig4C). Notably, these resistant tumors still homogeneously expressed DLL3 (Supp Fig4D). When examining tumors for mouse myeloid cells that might have contributed to resistance to MK-6070(21) using the pan-myeloid cell marker Cd11b, we did not observe a noticeable difference between naive and resistant WCM12 tumors (Supp Fig4E). Antitumor activity of MK-6070 was not seen in control DLL3-negative 22Rv1 xenografts (Fig 3E–F). Overall, these data support the in vivo anti-tumor activity of MK-6070 in DLL3-expressing NEPC xenograft models.

Figure 3. In vivo anti-tumor activity of MK-6070.

(A) DLL3 expression was evaluated by immunohistochemistry. WCM154 and WCM12 expressed DLL3 homogeneously; PC7 expressed DLL3 heterogeneously; 22Rv1 did not express DLL3. Scale bar = 200μm. (B) WCM154 xenograft model was treated with either TriTAC-GFP (100μg/kg, n = 5) or MK-6070 (100μg/kg, n = 5). Tumor size was measured twice per week by a caliper. All mice in the MK-6070 arm remained tumor-free after week 4 until the end of the study. The mice in the TriTAC-GFP arm were euthanized when tumors reached 2,000 mm³. (C) Kaplan-Meier plots showing survival curves corresponding to (B). Statistical significance was determined by Log-rank using Prism. (D) Mice body weight was used as a surrogate to evaluate adverse effects corresponding to (B). (E) 22Rv1 model was treated with either TriTAC-GFP (100μg/kg, n = 8) or MK-6070 (100μg/kg, n = 9). Mice were euthanized when tumors reached 2000 mm³. (F) Kaplan-Meier survival curves with TriTAC-GFP or MK-6070 in 22Rv1 corresponding to (E). Statistical significance was determined by Log-rank using Prism.

PC7 and 23-Lv-R4 are NEPC PDX models established pre- and post-platinum/etoposide chemotherapy that express heterogeneous DLL3 (Fig 3A & Supp Fig5A). Despite overall lower and more heterogenous expression of DLL3, both PC7 and 23-Lv-R4 tumors had comparable sensitivity to MK-6070 as the DLL3-homogeneous NEPC models (WCM154, WCM12) (Fig 4A–B & Supp Fig5B–C). We posited that this could be mediated by a bystander effect where T cells activated by DLL3-positive cells can impact surrounding DLL3-negative cells(22). To further evaluate for a potential bystander effect, we labeled a pair of isogenic cell lines, 22Rv1 and 22Rv1DLL3 with GFP and RFP, respectively. After co-culturing in the presence or absence of MK-6070, as expected 22Rv1 did not respond to MK-6070 indicated by RFP expression; in contrast, 22Rv1DLL3 cells were sensitive to MK-6070 shown by reduced GFP+ cells (Fig 4C–D). We then mixed RFP-labeled 22Rv1 cells and GFP-labeled 22Rv1DLL3 cells and treated with MK-6070. Both RFP DLL3-negative and GFP DLL3-positive cells exhibited cell death and demonstrated increased granzyme B levels upon MK-6070 exposure (Fig 4D–E & Supp Fig5D–E). Evaluation of varied ratios of 22Rv1 and 22Rv1DLL3 cells (Supp Fig5F) suggested a dose dependent effect and that the proportion of DLL3+ cells contributes to the magnitude of MK-6070- mediated bystander effect.

Figure 4. MK-6070 displays anti-tumor activity in heterogeneous DLL3 tumors.

(A) PC7 expressed heterogeneous levels of DLL3 as shown in Fig 3A. Mice were treated with either TriTAC-GFP (100μg/kg, n = 5) or MK-6070 (100μg/kg, n = 4). Tumors treated with MK-6070 began shrinking after the 2nd dose and all 4 mice were tumor-free at the end of the study. (B) 23-Lv-R is another DLL3 heterogeneous PDX shown in Supp Fig5A. Mice were treated with either TriTAC-GFP (100μg/kg, n = 5) or MK-6070 (100μg/kg, n = 9). Tumors treated with MK-6070 arm began shrinking after the 2nd dose and 3 out of 9 mice were tumor-free at the end of the study. (C) Schematic of experimental design to evaluate bystander effect. Parental 22Rv1 cells were labeled with RFP and isogenic 22Rv1DLL3 cells were labeled with GFP. RFP and GFP cells were mixed in different ratios. Results are displayed in (D) and (E). (D) Parental 22Rv1 cells were insensitive to MK-6070, but 22Rv1GFP cells showed sensitivity. On the right panel, 22Rv1 cells also responded to MK-6070 when co-cultured with 22Rv1DLL3 cells. These are representative images of the same experimental data shown in (E). The magnification is 40X. Scale bar = 100μm. (E) 22Rv1 and 22Rv1DLL3 cells were mixed in a 1:3 ratio and treated with various concentration of MK-6070. Imaging analysis was utilized to evaluate the confluency of RFP or GFP cells using Celigo. The confluency percentage is determined by the area covered by cells over the total surface area of a well. In 22Rv1/22Rv1DLL3 mixing condition, only RFP cells (22Rv1) were measured to determined bystander effect induced by the presence of 22Rv1DLL3 cells. The magnification is 20X. Scale bar = 200μm.

Discussion

DLL3 is an established cell surface target in SCLC that is also expressed in other poorly differentiated NE carcinomas and not in normal healthy adult tissues(7) making it an attractive pan-NE carcinoma therapeutic target. While NEPC shares histologic and molecular features with SCLC, NEPC typically arises from prostate adenocarcinoma(1). Up to 15–20% of patients with late-stage prostate cancer develop NE features associated with loss of AR-dependence and aggressive disease(23, 24). During the lineage plasticity process and transition from prostate adenocarcinoma to NEPC, Notch signaling is downregulated and the Notch inhibitory ligand DLL3 becomes expressed(9). Because this process is dynamic, it is not uncommon for CRPC tumors to have heterogeneous features with both DLL3-negative adenocarcinoma and DLL3-positive NEPC cells seen within or across metastatic lesions.

The National Comprehensive Cancer Network (NCCN) guidelines recommend platinum-based chemotherapy for biopsy confirmed NEPC and refer to SCLC guidelines(4) After platinum chemotherapy, options for SCLC include lurbinectedin, topotecan, and more recently the DLL3-targeted bispecific T cell engager tarlatamab. There are no prospective data for second line SCLC chemotherapy regimens in NEPC. A recent retrospective report of 18 patients with NEPC treated with lurbinectedin reported modest benefit in some patients with median progression free survival (PFS) and overall survival (OS) of 3.35 and 6.01 months, respectively(25).

Tarlatamab was FDA approved in 2024 for SCLC based on the phase 2 DeLLphi-301 trial(6). In the follow-up Phase 3 DeLLphi-304 trial for patients with SCLC after progression on platinum- based chemotherapy, tarlatamab treatment with associated with improved PFS and OS compared with chemotherapy(5). Adverse events of grade 3 or higher were lower with tarlatamab compared with chemotherapy. Immune related toxicities such as cytokine-release syndrome are the most common adverse event seen with tarlatamab as also seen with other T cell engager therapies. Based on the approval of tarlatamab and NCCN guidelines referring to SCLC therapies, there is increased interest in using tarlatamab for NEPC and other extra-pulmonary small cell carcinomas. Preclinical studies have supported promising activity of tarlatamab in NEPC models(26), and interim results from a phase 1b trial of tarlatamab supports a manageable safety profile and anti-tumor activity in NEPC(27). The trial enrolled not only small cell NEPC histology, but also CRPC tumors with NE marker (synaptophysin, chromogranin) positivity by immunohistochemistry and/or two or more genomic alterations involving TP53, RB1, and/or PTEN even if adenocarcinoma morphology was present. In 32 tumors that were evaluable for DLL3 expression, 56% expressed DLL3. The ORR was 10.5% and median duration of response was 1.4 months, with activity limited to those expressing DLL3.

MK-6070 is similar to tarlatamab in that it is also a T cell engager, but it binds not only to DLL3 and CD3 but also serum albumin to extend half-life. An ongoing phase 1/2 trial is currently enrolling patients with SCLC, histologically confirmed NEPC, and other DLL3-expressing NE carcinomas (NCT04471727). Interim results have reported promising ORR ranging 39–47% and DCR 46–71%(28). Correlative studies and analyses of the NEPC cohort have not yet been reported.

Here, we complement the ongoing clinical study with a preclinical evaluation of MK-6070 in prostate cancer models. Overall MK-6070 demonstrated high specificity against DLL3-positive NEPC and robust anti-tumor activity when co-administered with human T cells. This cytotoxicity is likely driven by T cell activation, corresponding with elevated levels of IFN-γ, TNF-α, and soluble granzyme B, and T cell infiltration within treated tumors. Although the use of different donor T cells showed similar efficacy, it is likely that variable T cell activity and immunogenicity in individual patients could impact clinical response. T cell exhaustion, driven by upregulation of immune checkpoints, has been identified as a resistance mechanism in preclinical models and in patients treated with other T cell engagers(13). Although this study did not comprehensively assess T cell exhaustion induced by MK-6070 in NEPC, the ongoing Phase 1/2 trial (NCT04471727) includes a SCLC cohort evaluating the combination of MK-6070 with the PD-L1 antibody atezolizumab.

There are limited prostate cancer preclinical models that represent the diversity of tumor phenotypes observed clinically. In this study, we used unique PDX models that represent NEPC and mixed histology tumors. However, these models have limitations as they are immunodeficient and required drug treatment in combination with exogenous human T cells. Additionally, we did not fully investigate cytokines and chemokines profiles of our preclinical models to elucidate how cytokines/chemokines may impact T cells infiltration upon MK-6070 engagement. A comprehensive assessment of immune-mediated responses, downstream pathways, resistance mechanisms, and the impact on the tumor microenvironment could not be investigated in these immunodeficient models.

In addition to T cell engagers, other approaches targeting DLL3 have been developed including antibody-drug conjugates (ADCs), chimeric antigen receptor T cells (CAR-T), and radionuclide therapy(29). Rovalpituzumab tesirine (Rova-T) was the first DLL3-targeted ADC that showed promising results in early phase trials but was ultimately discontinued in part due to payload-mediated toxicities(30). Although Rova-T was discontinued, next generation DLL3-targeted ADCs still represent a promising therapeutic modality with more suitable linkers and payloads. For instance, ZL-1310 is an ADC composed of a humanized anti-DLL3 monoclonal antibody with a cleavable linker to a camptothecin derivative topoisomerase I inhibitor. Interim data from the phase 1 trial (NCT06179069) showed promising antitumor activity(31) and ZL-1310 has been fast-tracked by the FDA; of the 19 patients with extensive stage SCLC who received at least one post-treatment tumor assessment, the ORR was 74% (95% CI, 48.8%−90.9%). DLL3-targeted CAR-T approaches such as AMG119 and LB2102 have also advanced to the clinic with preliminary safety and efficacy data reported(32, 33). Several DLL3-targeted radionuclides have shown preclinical activity in SCLC models(34–36), and the DLL3-targeted radionuclide [225Ac]Ac-ETN029 is currently undergoing phase 1 evaluation in SCLC with an expansion cohort that includes NEPC (NCT07006727).

Tumor heterogeneity is a key challenge that limits the efficacy of targeted therapies in oncology, which is especially relevant in CRPC/NEPC as there is often mixed adenocarcinoma and NE tumor cells seen within and across metastases. We found that MK-6070 has activity in models that express DLL3 heterogeneously, with responses that were comparable to DLL3 homogeneous tumors. Similar to what has been observed in other contexts with T cell engagers(22), it is possible that MK-6070 exhibits a bystander effect that could kill surrounding DLL3-negative cells. Understanding how DLL3 heterogeneity impacts acquired resistance to DLL3-targeted therapies in preclinical models as well as in patients with NEPC will be important.

Developing methods to capture heterogeneity could also aid in the development of future biomarkers to improve patient selection. For instance, DLL3-positron emission tomography (PET) imaging(26, 37, 38), potentially in conjunction with prostate-specific membrane antigen (PSMA)-PET, might improve the diagnosis of NEPC as well as patient selection for DLL3-targeted therapies. In the Phase 1/2 trial of the DLL3-targeted imaging tracer [⁸⁹Zr]Zr-DFO-SC16.56 in patients with metastatic SCLC and other NE carcinomas including NEPC, DLL3 PET imaging was feasible and showed strong tumor-specific uptake with a range in SUVmax that associated with corresponding tumor expression of DLL3 by IHC(38). Plasma cell free DNA (39) and circulating tumor cell platforms(40) can also capture and track tumor phenotype and might provide dynamic readouts of lineage plasticity and infer target expression. Understanding the biologic mechanisms driving NEPC transformation, DLL3 expression, and co-expression patterns with other emerging prostate cancer cell surface targets (e.g., B7-H3, KLK2, STEAP1) could also lead to rational combination strategies.

In summary, DLL3 is a promising therapeutic target for NEPC. We show robust preclinical activity of MK-6070 in models of NEPC including heterogenous tumors, supporting the ongoing clinical development of MK-6070.