Antitumor Activity of MEDI3726 (ADCT-401), a Pyrrolobenzodiazepine Antibody–Drug Conjugate Targeting PSMA, in Preclinical Models of Prostate Cancer

In the United States, prostate cancer is the most common cancer and the third leading cause of cancer deaths in men, with an estimated 161,360 new cases annually (1). Patients with metastatic prostate cancer initially respond to androgen deprivation therapy, but nearly all develop castration-resistant disease (2). Although recent therapeutic advances provide a survival benefit, not all patients respond, and those who do respond eventually develop resistance and experience disease progression (3). Therefore, there is a need for improved therapies for patients with metastatic castration-resistant prostate cancer (mCRPC).

Prostate-specific membrane antigen (PSMA), also known as glutamate carboxypeptidase II, has two known enzymatic functions. In the small intestine, it removes glutamates from dietary folates for absorption, and in the central nervous system, it hydrolyzes the most prevalent peptidic neurotransmitter, N-acetylaspartylglutamate (4). PSMA is also expressed in renal tubule epithelium and normal prostate tissue, although the function of PSMA in these tissues is not well understood (4). In contrast to the limited normal tissue expression at sites not accessible to an intact antibody (5), PSMA is highly expressed in the epithelium of nearly all primary and metastatic prostate cancers (6–9) and in the neovasculature that supplies most non-prostatic solid tumors, including carcinomas of the lung, colon, breast, kidney, liver, and pancreas as well as sarcomas and melanoma (10, 11). It is a clinically validated tumor-associated antigen for imaging prostate cancer, and PSMA-targeted anticancer therapies involving multiple modalities have demonstrated early signs of clinical activity (12–18).

Antibody–drug conjugates (ADC) consist of antibodies that specifically target tumor-associated antigens and are conjugated to small molecules that possess strong cytotoxic activity (19). Among the most potent cytotoxins are pyrrolobenzodiazepine (PBD) dimers, which can form highly cytotoxic DNA interstrand crosslinks (ICL; ref. 20). Because the crosslinks formed by PBD dimers are relatively nondistorting of the DNA structure, PBD dimer crosslinks are less susceptible to DNA repair mechanisms (21). Several ADCs under clinical evaluation in both hematologic and solid tumor indications use PBD dimers; e.g., rovalpituzumab tesirine and vadastuximab talirine are in phase III studies in small cell lung cancer and acute myeloid leukemia, respectively (22).

MEDI3726 (formerly known as ADCT-401, Fig. 1) is a PSMA-targeting ADC generated via site-specific conjugation of an engineered version of the anti-PSMA deimmunized (humanized) immunoglobulin G1 (IgG1) κ antibody J591 (23) to the PBD-based linker-drug tesirine (SG3249), which is the same PBD dimer used in rovalpituzumab tesirine (24). MEDI3726 is under evaluation in a phase I clinical trial as an anticancer treatment in patients with mCRPC (NCT02991911). Herein, we describe the in vitro and in vivo characterization of MEDI3726.

The antibody component of MEDI3726, J591, and the non-PSMA–targeting isotype control ADC contain the mutations C214S in the light chain and C226V and C229V in the heavy chain hinge (antibody numbering herein is according to EU numbering; ref. 25). These mutations allow the site-specific conjugation of maleimide at position C220 with the drug linker tesirine. The conjugation procedure and the analytical characterization of MEDI3726 as well as the generation and purification of the soluble recombinant extracellular PSMA domains (sPSMA) for human (shPSMA), cynomolgus monkey (Macaca fascicularis, scPSMA), and rat (srPSMA) are described in the Supplementary Methods section.

Two monomethyl auristatin E (MMAE)-based ADCs targeting PSMA (PSMA-ADC1-MMAE and PSMA-ADC2-MMAE) were generated as described in patent US 8114965 B2 by Concortis.

The ability of each sPSMA to displace MEDI3726 from shPSMA was determined by competitive ELISA. Briefly, Maxisorp microplates (Nunc) were coated with shPSMA (50 μL/well, 1.0 μg/mL in PBS) and incubated at 4°C overnight. A 2-fold dilution series of each sPSMA antigen was dispensed [50 μL/well, in PBS/Tween/BSA (PBS 0.05%, Tween 20 (v/v), 2% BSA (w/v)]. MEDI3726 was then added (50 μL/well of 0.2 μg/mL in PBS/Tween/BSA) and incubated for 1 hour at room temperature (RT). After the MEDI3726-shPSMA complex was washed, 50 μL of goat antihuman immunoglobulin G (IgG)–horseradish peroxidase (HRP) conjugate (Jackson ImmunoResearch; diluted 1:1,000 in PBS/Tween/BSA) was added and incubated for 1 hours at RT. After the wash, TMB-Turbo (Thermo Fisher Scientific) was added (75 μL/well) and the reaction was stopped with hydrochloric acid (HCl; 75 μL/well, 0.6 mol/L). Well optical densities were measured on the EnVision plate reader (PerkinElmer) at 450 nm, and data were analyzed by using GraphPad Prism (GraphPad Software, Inc.).

The LNCaP, CWR22Rv1, MDA PCa 2b, PC-3, and DU145 prostate cancer cell lines were obtained from American Type Culture Collection (ATCC) and maintained by following ATCC recommendations. The authentication of these mycoplasma negative lines was performed by ATCC to verify the identity of human cell lines and to rule out both intra- and interspecies contamination. The cell lines used in the assays were not reauthenticated or tested at ADC Therapeutics, but they were passaged for fewer than 6 months after the initiation of culture. The LNCaP C4-2 cell line was a gift from Dr. Neil Bander (Weill Cornell Medical College). Upon arrival, the cell line went authenticity and IMPACT testing (for murine viruses) by IDEXX Bioresearch. Cell-surface PSMA expression and antigen density were quantitated by flow cytometry on a panel of six human prostate cancer cell lines including LNCaP, LNCaP C4-2, CWR22Rv1, MDA PCa 2b, PC-3, and DU145. The specific antigen density for each cell line was calculated by using Quantum Simply Cellular Anti-Human IgG beads (Bangs Laboratories) according to the manufacturer's instructions, with the antibody component of MEDI3726. The percentage of PSMA-positive cells in the prostate cancer cell lines was calculated from the mean fluorescence intensity (MFI) of each cell line, based on the same data set, by using FlowJo v10 (FlowJo, LLC) for analysis.

The cytotoxic activity was determined by CellTiter-Glo. The cell suspension was diluted to the required seeding density (generally 10⁵/mL), dispensed into white 96-well flat-bottomed microplates (50 μL/well), and incubated overnight. Eight 10-fold serial dilutions of ADCs or SG3199 (the warhead component of tesirine) were dispensed (50 μL/well) into 4 replicate wells of a 96-well plate containing cells. Control wells received culture medium (50 μL/well) only. The cytotoxicity of SG3199, MEDI3726, and the nontargeted isotype control ADC was measured in each cell line after an incubation period ranging from 5 to 7 days at 37°C and 5% carbon dioxide. This incubation period was calculated to be at least three normal cell doubling times and was used for exposure to either warhead toxin or ADCs. After the ADC exposure period, cell viability was measured with the CellTiter-Glo assay (Promega) following the manufacturer's protocol. IC₅₀ values were calculated by using GraphPad to fit luminescence data to a sigmoidal dose–response curve with variable slope.

Immunofluorescence microscopy was used to detect internalization and colocalization of MEDI3726 or an isotype control ADC with the lysosomal marker, lysosome-associated membrane protein 1 (LAMP1). One cell line demonstrated to be PSMA positive and one demonstrated to be PSMA negative were incubated with MEDI3726 or the isotype control ADC. A detailed protocol is provided in the Supplementary Methods.

The modified single-cell gel electrophoresis (Comet) assay as described by Hartley and colleagues (26) was used to monitor the formation of DNA ICLs generated by the SG3199, MEDI3726, or the nontargeting isotype control ADC and the unhooking step of repair of these ICLs. Briefly, 10⁵ cells were seeded in 24-well plates and treated with MEDI3726 (6 nmol/L), isotype control ADC (4.7 nmol/L, an equimolar PBD concentration) or SG3199 (2.5 nmol/L) for 2 hours, washed, and incubated with drug-free culture medium for the required posttreatment time (up to 36 hours). At each time point, cells were harvested and stored at −80°C until they were used. Except for nonirradiated controls, thawed samples (3 × 10⁴ cells/mL) were irradiated (15 Gy) to introduce a fixed number of random DNA strand breaks and then processed for the Comet assay. The Olive Tail Moment (OTM) was determined as described previously (27), using the formula:

where TMdi is the tail moment of drug-treated, irradiated sample; TMcu is the tail moment of untreated, nonirradiated control; and TMci is the tail moment of untreated, irradiated control. Crosslinking was expressed as the percentage decrease in the OTM compared with control irradiated cells.

The LNCaP xenograft study was performed by Crown Bioscience, United Kingdom, according to ethical standards and practice. LNCaP cells (5 × 10⁶) were implanted subcutaneously with Matrigel (1:1; BD Biosciences) into the left flank of 4- to 8-week-old male MF1 nude mice (Hsd:Ola-MF1-Foxn1^nu, Envigo). Once the mean tumor volume reached ∼200 mm³, animals were randomized (based on tumor volume) into study groups (n = 10), and dosing with ADCs was initiated (day 1). The body weight of mice ranged from 22.6 g to 35.3 g. Each animal was euthanized when its tumor attained the endpoint tumor volume of 1,700 mm³ or on the final day (day 64), whichever came first.

The CWR22Rv1 (a castration-resistant prostate cancer model; ref. 28) and PC-3 studies were completed at Charles River Discovery Services, in an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) accredited facility in compliance with accepted standards for the care and use of laboratory animals. CWR22Rv1 cells (1 × 10⁷) were implanted subcutaneously with Matrigel (1:1) in the right flank of male C.B-17 SCID mice (Fox Chase SCID, C.B-17/Icr- Prkdc^scid/IcrIcoCrl, Charles River). When the mean tumor volume reached ∼120 mm³, animals were randomized (based on tumor volume) into study groups (n = 10) and treatment was initiated (day 1). Mice were 10 weeks old with a body weight range of 18.6 to 27.2 g on day 1 of the study. For the PC-3 xenograft study, 1-mm³ tumor fragments were implanted subcutaneously in the right flank of each female athymic nude mouse (Crl:NU(Ncr)-Foxn1^nu, Charles River). When the mean tumor volume reached ∼135 mm³, mice were randomized (based on tumor volume) into study groups (n = 10) and the treatment was initiated (day 1). Mice were 11 weeks old with a body weight range of 17.5 to 27.6 g on day 1 of the study. In the CWR22Rv1 and PC-3 studies, each animal was euthanized when its tumor volume exceeded 1,000 mm³, or on the final day (day 94 for the CWR22Rv1 study and day 60 for the PC-3 study), whichever came first.

All dosing materials were administered intravenously as a single dose except for the MMAE-based PSMA-targeted ADCs, which were administered once every 4 days for six doses total. Both tumor and body weight measurements were collected twice weekly, and the tumor volume was calculated using the formula:

Statistical analysis is described in the Supplementary Methods.

The LuCaP 35CR, 58, 70, 73, 77, 86.2, 96CR, and 147 prostate cancer PDX models were used (29). LuCaP 35CR and LuCaP 96CR are castration-resistant models established and maintained in castrated male mice (29). All LuCaP models used in this study express differing levels of androgen receptor (AR) splice variant AR-V7, and LuCaP 86.2 model expresses the AR-V567es variant (30, 31).

The general health of the mice was monitored daily, and all experiments were conducted in an AAALAC-accredited facility in accordance with MedImmune's Institutional Animal Care and Use Committee guidelines for humane treatment and care of laboratory animals. All PDX models were maintained in-life by continuous passage in 8- to 12-week-old male C.B-17/IcrHsd-Prkdc^scid mice (Envigo). Both the LuCaP 35CR and 96CR models were propagated in castrated male mice. To initiate tumor growth, 3-mm³ tumor fragments were implanted subcutaneously into the right flank of 8- to 12-week-old male C.B-17/IcrHsd-Prkdc^scid mice (Envigo) by using an 11-gauge trocar needle. Once the tumors reached a size of 150 to 250 mm³, the mice were randomized (based on tumor volume) into study groups of 6 mice per group. The body weight of mice ranged from 21.1 to 25.5 g at the initiation of dosing. MEDI3726 was given intravenously at doses of 0.1, 0.3, or 0.9 mg/kg, and the isotype control ADC was given at a dose of 0.9 mg/kg. Both ADCs were administered intravenously once every 3 weeks for three doses according to body weight. Tumor and body weight measurements and the statistical analysis were carried out as described above.

Detailed IHC protocols for detecting PSMA and phosphorylated Histone H2AX (γ-H2AX) and the quantification of γ-H2AX are provided in the Supplementary Methods. Briefly, the PSMA IHC assay used mouse monoclonal anti-human PSMA clone 3E6 (Dako) and mouse IgG1 isotype control (R&D Systems) and was performed on the Dako autostainer by using the Dako EnVision+ System-HRP (DAB) Kit. For the γ-H2AX IHC, LuCaP 73 xenograft-bearing animals were dosed once, and xenograft tumors were collected 48 hours after the dose for formalin-fixed paraffin-embedded generation. The rabbit anti-γ-H2AX (Ser139; 20E3; Cell Signaling Technology) was used on the Ventana Discovery ULTRA Staining Module. Stained slides were scanned with the Aperio Scanscope AT Turbo for the generation of digital images, and γH2AX-positive nuclei were quantified by using Aperio Nuclear Algorithm. Statistically significant differences were determined by using one-way ANOVA with post hoc Tukey for the multiple comparison. GraphPad Prism 7 was used for statistical analyses.

MEDI3726 had a monomer content of 98% as determined by size-exclusion chromatography and a drug-to-antibody ratio (DAR) as measured by reduced liquid chromatography mass spectrometry (LC/MS), close to 2. Both shPSMA and scPSMA competed for binding of MEDI3726 to shPSMA with similar half-maximal effective concentrations of 17.92 ng/mL and 16.89 ng/mL, respectively (Table 1). However, srPSMA did not compete for binding to MEDI3726, indicating a lack of rodent cross-reactivity for J591. As demonstrated by surface plasmon resonance, MEDI3726 bound to shPSMA and scPSMA with similar affinities of 1.67 and 3.11 nmol/L, respectively, consistent with the competitive ELISA data (Table 1). MEDI3726 maintained activity for up to 7 days in PBS or human or cynomolgus monkey plasma at 37°C as determined by in vitro cytotoxic activity on LNCaP cells, binding to sPSMA and determination of PBD containing ADC by anti-PBD antibody capture assays (Supplementary Fig. S1).

The LNCaP and LNCaP C4-2 prostate cancer lines had the highest antigen densities, whereas PC-3 and DU145 had antigen densities below the lower limit of quantitation (7,348) for the assay and were therefore classified as PSMA-negative (Table 2). More than 90% of LNCaP, LNCaP C4-2, and MDA PCa 2b cells were positive for PSMA expression, but fewer than half of the CWR22Rv1 cells were PSMA-positive (Table 2).

SG3199, the warhead component of tesirine, displayed potent cell-killing activity in all cell lines (IC₅₀ ranging from 0.04 to 1.05 nmol/L). Target-specific cytotoxicity of MEDI3726 was observed in the four PSMA-positive cell lines (LNCaP, LNCaP C4-2, MDA PCa 2b, and CWR22Rv1). For the two PSMA-negative cell lines, the IC₅₀ ratio of the isotype control ADC to MEDI3726 was approximately 1, indicating that no targeted cytotoxicity occurred in cell lines lacking PSMA expression (Table 2).

Cellular internalization and trafficking of MEDI3726 to the lysosome are essential for SG3199 to be released from other tesirine components by cleavage of the valine–alanine dipeptide linker. Within 24 hours of continuous ADC exposure, MEDI3726 was internalized and trafficked to LAMP-1–positive lysosomes in PSMA-positive LNCaP cells (Fig. 2A). Conversely, MEDI3726 was not internalized by the PSMA-negative PC-3 cells, indicating that uptake of MEDI3726 was specific to PSMA (Fig. 2B). The isotype control ADC did not bind and was not internalized by LNCaP or PC-3 cells (Fig. 2C and D, respectively).

Immediately after LNCaP cells were incubated with MEDI3726, a low level of ICL (5%) was observed and increased with time to a peak level of approximately 50% at 24 hours, which persisted at 36 hours (Fig. 2E). SG3199 induced ICLs more rapidly, with approximately 35% observed at the end of the incubation period and a peak of more than 60% by 6 hours. Peak level of ICL induced by SG3199 also persisted at 36 hours. The lower level of and longer time to maximum ICL formation by MEDI3726 likely arose from differential uptake and an extra processing hurdle of MEDI3726 to release the payload compared with SG3199, which can diffuse across the membrane and crosslink DNA without processing.

In the PSMA-negative cell line PC-3, MEDI3726 did not produce a significant level of DNA ICLs over 36 hours (Fig. 2F). However, SG3199 produced a high level of ICLs, similar to that observed in the LNCaP cell line. These results are consistent with the target-mediated internalization and cell killing by MEDI3726.

As determined by IHC, PSMA expression in the LNCaP xenografts was homogeneous, with nearly all tumor cells staining strongly positive for PSMA (Fig. 3A). In this model, MEDI3726 showed a dose-dependent antitumor activity (Fig. 3B), with an extended tumor stasis preceding regrowth after a single dose of 1 mg/kg, and a significant increase in the time-to-endpoint survival, compared with all other groups (Supplementary Table S1). Additionally, MEDI3726 was associated with a statistically significant increase in animal survival, compared with the naked antibody group, at 0.33 mg/kg and 1 mg/kg (Supplementary Table S1). The nontargeting isotype control ADC was also associated with a statistical increase in survival compared with the control groups, but only at the highest dose of 1 mg/kg (Supplementary Fig. S2; Supplementary Table S1).

Next, we investigated MEDI3726 in CWR22Rv1 xenografts, castration-resistant model (28). IHC analysis of the CWR22Rv1 xenograft tumors confirmed the heterogeneous expression of PSMA in this model, with approximately 25% to 50% of tumor cells staining positive for PSMA (Fig. 3C; Supplementary Table S2). Administration of MEDI3726 resulted in a dose-dependent antitumor activity (Fig. 3D), with a prolonged tumor regression (approximately 50 days) after a single 1 mg/kg dose (Supplementary Table S3). A single dose of the isotype control ADC at 0.75 mg/kg resulted in tumor growth comparable with that seen in vehicle-treated animals (Supplementary Fig. S3). Notably, two monomethyl auristatin E-based, PSMA-targeted ADCs, PSMA-ADC1-MMAE (DAR = 4) and PSMA-ADC2-MMAE (DAR = 4), did not show antitumor activity at 3 or 6 mg/kg. Nor did these ADCs show in vitro cytotoxic activity in CWR22Rv1 cells, although they did bind to PSMA-positive cells and showed strong cytotoxic activity on LNCaP cells (Supplementary Fig. S4).

In the PC-3 xenograft model, which did not express PSMA (Fig. 3E), administration of MEDI3726 and the isotype control ADC resulted in limited antitumor activity at the highest dose (1 mg/kg; Fig. 3F; Supplementary Table S4). This result might have occurred because of nontarget–specific uptake of the ADCs by the tumor cells and/or local release of PBD toxin caused by tumor-associated macrophages (32).

To further characterize the in vivo activity of MEDI3726, we assessed the efficacy of MEDI3726 against a panel of clinically relevant LuCaP prostate cancer PDX models. Among eight LuCaP models, six (LuCaP 70, 73, 77, 86.2, 96CR, and 147) showed strong membrane expression of PSMA in more than 75% of tumor cells, as determined by IHC (Fig. 4A; Supplementary Table S2). Of the remaining two, LuCaP 58 showed moderate PSMA staining intensity in 10% to 25% of tumor cells and LuCaP 35CR was negative for PSMA. Consistent with the results in the cell line xenografts, highly PSMA-positive LuCaP models responded in a dose-dependent manner to MEDI3726 (Fig. 4B–G). Compared with untreated controls, animals treated with 0.9 mg/kg MEDI3726 showed statistically significant responses in all highly PSMA-positive LuCaP models. Complete and durable tumor regressions were observed in five of the six highly PSMA-positive models (Fig. 4B–F). The LuCaP 147 model displayed a tumor regression followed by a prolonged tumor stasis (Fig. 4G). In the LuCaP 70, 73, and 96CR models, animals given 0.3 mg/kg MEDI3726 showed statistically significant tumor regressions compared with the untreated control group (Fig. 4B, C, and F). At this same dose, the LuCaP 77 model displayed a statistically significant prolonged tumor stasis (Fig. 4D). In contrast, the 0.3 mg/kg dose of MEDI3726 only inhibited tumor growth in the LuCaP 86.2 and 147 models (Fig. 4E and G). The 0.1 mg/kg dose of MEDI3726 did not significantly delay tumor growth, compared with the control in any of the highly PSMA-positive LuCaP models. MED3726 was not associated with tumor growth inhibition at any dose in the heterogeneously PSMA-positive LuCaP 58 model or the PSMA-negative LuCap 35CR models (Fig. 4H and I).

Next, we sought to investigate the phosphorylation of Histone H2AX (γ-H2AX), which occurs following the DNA damage induced by SG3199-generated ICLs. Tumors were collected from the LuCaP 73 xenograft-bearing mice 48 hours after the first single administration of each ADC. A dose-related increase in the number of γ-H2AX-positive nuclei was observed among each of the MEDI3726 treatment groups, compared with both the control and isotype control groups. However, statistical significance was achieved only at the highest MEDI3726 dose tested (Fig. 4J; Supplementary Fig. S5). It is important to note that the collection time point at 48 hours post single dose may not have been optimal to detect the differences in γ-H2AX–positive cells between the two lower dose levels, given that the tumor regression was not observed until the second dose of 0.3 mg/kg of MEDI3726.

MEDI3726-related toxicity, as determined by body weight measurements, was not observed in any of the cell line xenograft and PDX model studies (Supplementary Fig. S6). As noted earlier, MEDI3726 does not bind to rodent PSMA.

Cynomolgus monkey was selected as a relevant species for the toxicity study after MEDI3726 demonstrated comparable cross-reactive binding activity to human and cynomolgus monkey PSMA as determined by BIAcore (SPR) and competitive ELISA assays (Table 1). A repeat-dose, good-laboratory-practice (GLP) toxicology study of MEDI3726 was assessed in cynomolgus monkeys (5/sex/group) which were administered i.v. injections of 0.15, 0.3, or 0.6 mg/kg MEDI3726 on days 1 and/or 22. Animals were necropsied on days 29 (main study; 3/sex/group) and 71 (recovery; 2/sex/group).

No changes were noted in body weight, blood clinical chemistry, urine, coagulation parameters, body temperature, physical and neurologic exams, neurobehavior, ophthalmology, cardiovascular endpoints [electrocardiogram (ECG) and blood pressure], respiratory rate or organ weights.

Dose-dependent reversible myelosuppression was the primary toxicity, correlating with reversible findings in the bone marrow (hypocellularity of erythroid and myeloid cells and megakaryocytes) and transient neutropenia (Supplementary Fig. S7). Dose-dependent (late onset) reversible black skin discoloration, correlating with epidermal hyperpigmentation and hyperplasia was observed as well as reversible microscopic changes in lymphoid organs (decreased or absent germinal centers) and kidney (tubular degeneration, cellular hypertrophy/atypia, focal inflammation, and edema). Changes in the testis (seminiferous tubule degeneration, vacuolation, reduced sertoli cells, and loss of germ cells) were not fully recoverable during the study, but the presence of intact spermatogonia in the testis supported that full recovery would be evident with time (33). The highest non-severely toxic dose was the highest dose tested of 0.6 mg/kg.

MEDI3726 is a PSMA-targeting ADC generated via site-specific conjugation of an engineered version of the J591 antibody to the PBD-based linker-drug tesirine. In vitro, MEDI3726 demonstrated potent and specific cytotoxic activity. In the CWR22Rv1 cell line, where approximately 40% of the cells were positive for PSMA expression, MEDI3726 demonstrated potent in vitro activity, with close to 100% of the cells being killed. This maximum cell kill was likely achieved via bystander activity, which has been reported for other ADCs using tesirine (27, 34, 35). In agreement with the potent in vitro activity observed with the CWR22Rv1 cell line, durable tumor regressions was observed after a single MEDI3726 dose administration to mice subcutaneously implanted with the CWR22Rv1 cells. Thus, both in vitro and in vivo data suggest that MEDI3726 has robust bystander activity with an ability to kill neighboring non–PSMA-expressing tumor cells. Interestingly, MEDI3726 demonstrated a more potent activity in the CWR22Rv1 xenograft study compared with the LNCaP model, even though the LNCaP model had more intense and homogeneous PSMA expression. Many factors may explain this observation. For example, difference in the tumor stromal component could impact the ADC perfusion or the high, homogeneous PSMA expression found in LNCaP tumor cells could potentially prevent a more efficient tumor penetration and distribution due to the “binding site barrier” effect (36). Imaging studies investigating the tumor distribution of MEDI3726 may help us understand the differences in efficacy between these two models. Importantly, multiple doses of two auristatin-based PSMA-ADCs (US 8114965 B2; 3 mg/kg or 6 mg/kg, q4d × 6) did not result in any significant antitumor activity in the CWR22Rv1 xenograft model.

PDX models were used to assess the in vivo activity of MEDI3726. Such models are particularly important for assessing in vivo activity in prostate cancer, for which the number of available cell line xenograft models is limited. LuCaP PDX models have been derived from patient samples at various stages of disease and metastatic sites, and thus reflect the genetic and phenotypic heterogeneity of the disease (29). MEDI3726-responsive models represented various types of castration-resistant prostate cancer (LuCaP 70, 73, 77, 86.2, 96CR, 147) and metastasis (LuCaP 70, 77, 86.2, and 147). MEDI3726-responsive models also varied in AR expression; AR gain-of-function mutations (present in LuCaP 73 and 147) and the splice variants AR-V7 (31) and AR-V567es (present in LuCaP 86.2; ref. 30) which have been implicated in the resistance to AR inhibitors. Moreover, MEDI3726 appeared more potent compared with the auristatin-based PSMA-ADC, as a lower dose and less frequent administration showed comparable with superior antitumor activity in LuCaP 96CR and LuCaP 77 PDX models, respectively, in comparison with the published auristatin PSMA-ADC results in the these models (37). Interestingly, MEDI3726 had minimal antitumor activity in the LuCaP 58 model, which, like CWR22Rv1, showed heterogeneous PSMA expression. The difference in antitumor activity may have been due to fewer PSMA-positive tumor cells in the LuCaP 58 model compared with the CWR22Rv1 model.

As part of the Investigational New Drug Application enabling studies, MEDI3726 was tested in a repeat-dose, intravenous, GLP study in cynomolgus monkeys. The major adverse effects following administration of MEDI3726 were related primarily to the known safety profile previously reported in a phase I study of an ADC containing the same linker-drug teserine (38) and were consistent with the profile associated with the mechanism of action of the PBD dimer warhead. The major dose-limiting toxicity associated with MEDI3726 was myelosuppression, but this effect was readily monitorable. Overall, the toxicology profile of MEDI3726 supported further study of the drug in phase I clinical trials.

PSMA is a well-characterized tumor-associated antigen for prostate cancer and possesses many features that make it attractive for an ADC-mediated targeting approach to mCRPC. Two PSMA-targeting ADCs using tubulin-inhibiting payloads have been investigated in clinical studies in patients with mCRPC (16, 39). MLN2704, an ADC utilizing the unmodified J591 anti-PSMA antibody, has shown a limited therapeutic window in a phase I/II study in mCPRC patients attributed to the unstable disulfide linker, leading to rapid deconjugation of the DM1 warhead after administration to patients (39). PSMA-ADC has demonstrated antitumor activity in a phase II study in patients with mCRPC based on radiologic response and reductions in circulating tumor cells and prostate-specific antigen (16). Although neither of these tubulin inhibitor based ADCs is under current clinical development, MEDI3726 is the first PSMA-targeting ADC utilizing the highly potent DNA crosslinking PBD payload tesirine in clinical development for mCRPC (NCT02991911).

Preclinical studies have demonstrated that reduced expression of BRCA1 or BRCA2, key components of the homologous recombination (HR) pathway involved in the repair of PBD-mediated DNA damage, increases the sensitivity of tumor cells to PBD-mediated cell death (40, 41). Notably, recent data have shown that mutations in DNA repair pathway genes occur in 20% to 25% of prostate cancer cases (42). Thus, disruption in DNA repair pathway in prostate cancer may make difficult-to-treat tumors such as mCRPC especially sensitive to DNA damaging PBD payloads compared with the microtubule-inhibitor based–ADCs previously studied in clinical trials. Moreover, pharmacologic suppression of the AR pathway not only increases the expression of PSMA, but also decreases the expression of genes involved in the HR pathway (43, 44).

In conclusion, the emerging data showing frequent mutations of DNA damage response (DDR) genes in prostate cancer together with the decreased DDR gene expression induced by pharmacologic suppression of the AR pathway provide a strong rationale for targeting prostate cancer with MEDI3726, which combines specific targeting of highly expressed prostate cancer tumor antigen PSMA with a potent DNA damage inducer, not only as monotherapy, but also in conjunction with AR pathway inhibitors.

N.R. Monks has ownership interest (including stock, patents, etc.) in AstraZeneca. F. d'Hooge is ADC Project Manager at Gamamabs Pharma S.A. R. Fleming is a scientist at Medimmune LLC. N. Dimasi has ownership interest (including stock, patents, etc.) in AstraZeneca Stocks. M.J.M. Hinrichs has ownership interest (including stock, patents, etc.) in AstraZeneca. N.H. Bander is an advisor at, has ownership interest (including stock, patents, etc.) in, and is a consultant/advisory board member for BZL Biologics, LLC. P.W. Howard has ownership interest (including stock, patents, etc.) in ADCT. J.A. Hartley is the Director of Pre-Clinical Development at Spirogen Ltd, reports receiving a commercial research grant from ADC Therapeutics, has ownership interest (including stock, patents, etc.) in ADC Therapeutics, and is a consultant/advisory board for ADC Therapeutics. S. Coats has ownership interest (including stock, patents, etc.) in AstraZeneca. R. Herbst has ownership interest (including stock, patents, etc.) in MedImmune. P.H. van Berkel has ownership interest (including stock, patents, etc.) in ADC Therapeutics SA. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Cho, F. Zammarchi, D.G. Williams, N.R. Monks, N. Dimasi, M.J.M. Hinrichs, N.H. Bander, P.W. Howard, J.A. Hartley, D.A. Tice, P.H. van Berkel

Development of methodology: S. Cho, C.E.G. Havenith, N.R. Monks, P. Tyrer, R. Fleming, K. Vashisht, N. Dimasi, F. Bertelli, L. Adams, H.W. Reinert, C.E. Britten, M. Korade III, E. Corey, N.H. Bander, P.W. Howard

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Cho, N.R. Monks, P. Tyrer, F. D'Hooge, R. Fleming, K. Vashisht, N. Dimasi, S. Corbett, L. Adams, H.W. Reinert, S. Dissanayake, W. King, R. Tammali, K. Schifferli, P. Strout, M. Korade III, H. Liu, S. Kim, N.H. Bander,

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Cho, F. Zammarchi, C.E.G. Havenith, N.R. Monks, P. Tyrer, K. Vashisht, N. Dimasi, S. Corbett, L. Adams, H.W. Reinert, S. Dissanayake, C.E. Britten, W. King, K. Dacosta, P. Strout, M. Korade III, M.J.M. Hinrichs, N.H. Bander, D.A. Tice, P.H. van Berkel

Writing, review, and/or revision of the manuscript: S. Cho, F. Zammarchi, C.E.G. Havenith, N.R. Monks, P. Tyrer, R. Fleming, K. Vashisht, N. Dimasi, F. Bertelli, C.E. Britten, P. Strout, S. Chivers, E. Corey, N.H. Bander, P.W. Howard, S. Coats, D.A. Tice, R. Herbst, P.H. van Berkel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Cho, F. Zammarchi, C.E.G. Havenith, F. D'Hooge, R. Fleming, S. Dissanayake, C.E. Britten

Study supervision: S. Cho, F. Zammarchi, K. Vashisht, N. Dimasi, K. Schifferli, N.H. Bander, P.W. Howard, J.A. Hartley, S. Coats, D.A. Tice, R. Herbst, P.H. van Berkel

Other (acquisition of data through sample analysis): S. Dissanayake

This study was supported by MedImmune, the global biologic R&D arm of AstraZeneca, and ADC Therapeutics SA. The authors would like to thank Dr. Barry R. Davies, and Ms. Holly Nguyen for the input and support in selecting and establishing LuCaP models at MedImmune. In addition, we thank Dr. David W. Jenkins for contributions in the preclinical development of MEDI3726 and Dr. Rakesh Dixit for the critical review of the manuscript. Editorial support was provided by Frances McFarland, PhD, MA (funded by MedImmune).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.