CUB domain containing protein 1 (CDCP1) is a target for radioligand therapy in castration resistant prostate cancer including PSMA null disease

General Methods:

All data presented in this study are available upon request. All materials and chemicals were purchased from commercial vendors and used without further processing/purification. DU145, 22RV1, C4-2B, and PC3 cell lines were obtained from American Type Tissue Collection (ATCC) and cultured according to manufacturer’s recommendations. Cellular identity was authenticated by visually inspecting morphology and probing for signature expression markers on immunoblot. Mycoplasma contamination was tested within the first two passages after thawing cryostocks with the MycoAlert kit (Lonza). All cells were studied between passages 5 to 25. The monoclonal antibody 4A06 was expressed and purified in the IgG1 format as previously described (13). p-SCN-Bn-Deferoxamine (B-705) and p-SCN-Bn-DOTA (B-205) were purchased from Macrocyclics (Plano, TX). ⁸⁹Zr-oxalate was obtained from 3D Imaging, LLC (Maumelle, AR). ¹⁷⁷LuCl₃ was obtained from Oak Ridge National Laboratory. Iodine-125 was obtained from Perkin Elmer. ⁶⁸Ga-PSMA 11 was prepared by the radiopharmacy at UCSF according to previously reported protocol (17). LTL xenograft samples were acquired form Living Tumor Laboratory (Vancouver, BC) and the LuCaP xenograft series was provided by Dr. Eva Corey at University of Washington.

Analysis of Patient Biopsies:

Written informed consent was obtained prior to collecting patient biopsies. The sample collection and analysis were performed in accordance with the ethical guidelines stipulated in the Declaration of Helsinki, and the study was review and approved by the Institutional Review Board at UCSF prior to its start. Recently acquired metastatic biopsies from mCRPC patients clinically annotated with survival data was used for analysis (18, 19). Patient samples were obtained using image-guided core needle biopsy of metastatic lesion in the bone or soft tissue. Laser capture microdissection was used to isolate samples enriched for cancer, with cores freshly frozen for RNA sequencing, and separate core formalin-fixed and paraffin embedded for immunohistochemical analysis.

SCNC status was determined via unsupervised hierarchical clustering using a previously validated gene signature and confirmed via evaluation by three experienced pathologists blinded to the clinical and genomic features for determination of consensus pathologic subclassification(19). The following list of genes was used for hierarchical clustering to determine SCNC status: AR, TMPRSS2, GATA2, HOXB13, KLK3, FOXA1, NKX3–1, CHGB, FOXA2, SOX2, SCG2, NKX2–1, REST, SPDEF, NOTCH2, NOTCH2NL, ASCL1, ETV1, ETV4, ETV5, RB1, CDKN2A, E2F1. Linear regression between CDCP1 and FOLH1 expression was performed in R (v4.0.2), with the resulting P value and Pearson correlation coefficient reported.

The needle biopsies were fixed in 10% formalin overnight, transferred to 70% ethanol the next day, and then processed for paraffin embedding and sectioning. Sections of formalin-fixed, paraffin embedded (FFPE) tissue were placed into the Ventana Discovery Ultra automated slide stainer. Antigen retrieval was performed using heat-inactivated antigen retrieval buffer (Tris-EDTA) according to the manufacturer instructions (Roche) and then stained with the following primary antibodies: rabbit polyclonal CDCP1 (Cell Signaling Technology, #4115S, 1:50), mouse monoclonal PSMA (Dako Agilent, M3620 clone 3E6, 1:100) or rabbit monoclonal CD3 (Ventana clone 2GV6, #790–4341, 1:100) for 32 minutes at 36°C. Secondary antibodies (Anti-Rabbit HQ, Anti-Mouse alk phos, HQ-HRP, Purple HRP and Yellow AP for dual chromogenic stains, all from Ventana) were incubated for 12 min each, and DAB was used for detection for single stains. Slides were counterstained with hematoxylin per standard protocol. H-scores for CDCP1 and PSMA membrane staining were assigned by two independent pathologists (1+, 2+ and 3+ multiplied by the percentage), and the average of the H-scores was calculated.

Flow Cytometry:

Actively proliferating cells were lifted mechanically from a tissue culture plate and placed in the primary antibody solution diluted in PBS (4A06, 1:1,000). After a 30 min incubation, the cells were washed with PBS multiple times and placed in a fluorophore conjugated secondary antibody solution for 30 minutes on ice (1:500, 109-546-097, Jackson ImmunoResearch). Cells were collected and washed before being placed in PBS and passed through a cell strainer. Samples were taken to a flow machine (FACS CantoII). Data analysis was performed using FlowJo.

Immunoblot:

Tumor samples were added to Pierce RIPA lysis buffer (89900, ThermoFisher Scientific) with Halt protease and phosphatase inhibitors (1861281, ThermoFisher Scientific). The samples were homogenized using a probe sonicator (Omni TH-01) and then centrifuged for 15 minutes at 15,000 ×g. 4X SDS-loading buffer was added to protein lysates and 15 μg of lysate were resolved on a 4–12% Bis Tris gel (NW04120BOX, Invitrogen). Gels were transferred onto an Immobilon-P membrane (IPVH00010, Millipore). Membranes were blocked in 5% milk in TBST for 90 minutes before being placed in a primary antibody solution. Primary antibodies used were CDCP1 (4115, Cell Signaling, 1:1000), PSMA (12815, Cell Signaling, 1:1000), and B-Actin (A5441, Sigma-Aldrich, 1:5000). Membranes were then washed and incubated with a secondary antibody solution for 30 min at room temperature. Secondary antibodies used were goat anti-rabbit (65-6120, Invitrogen, 1:5000) and goat anti-rat (62-6520, Invitrogen, 1:5000). Proteins were detected using West Pico Chemiluminescent Substrate for 20 seconds (34578, ThermoFisher Scientific) and then exposed to film (30-507, Blue Devil). Each immunoblot was reproduced at least once with freshly harvested protein samples.

Saturation binding assays:

To prepare ¹²⁵I-4A06, 6.0 μL of ¹²⁵I in 1M NaOH (~ 340 uCi) was added into a vial containing 100.0 μL of HEPES buffer (0.5 M). This solution was transferred to an iodination tube (Peirce) and 4A06 antibody (300μg) was then added. The tube was incubated for 10 min at room temperature, and ITLC showed 85.0% radiolabeling efficiency (solvent: 20nM citric acid). ¹²⁵I-4A06 was purified using a G-25 column. The final yield of the purified ¹²⁵I-4A06 was 68.58% (specific activity =1.13 μCi/μg) and purity was > 99%.

DU145 and PC3 (0.6 × 10⁶ cell/well) were seeded on 12-well plates using DMEM (10% FBS). The cells were washed with PBS for the saturation binding assay. Total binding of ¹²⁵I-4A06 was determined by adding it to cell suspensions at seven concentrations from 0.025 nM to 10 nM. Non-specific binding was determined by adding 1000x cold 4A06 to ¹²⁵I-4A06/cell mixtures at 0.025nM, 0.3 nM, and 10 nM. In all cases, cells were incubated with ¹²⁵I-4A06 at room temperature for 1 hr, washed with PBS, and lysed by adding 1.0 M NaOH. The bound and unbound radioactive fractions were collected and measured on a Hidex Gamma counter (Turku, FI). Bmax was calculated using Prism v8.0.

Animal studies:

All animal studies, including housing and welfare monitoring, were conducted in compliance with Institutional Animal Care and Use Committee at UCSF. Animal imaging studies involving patient derived xenografts (PDX) utilized eight to ten week old intact male NOD SCID gamma (NSG) mice from Charles River Laboratory. PDX lines were obtained from the Living Tumor Lab at the Vancouver Prostate Centre. For tumor imaging or treatment studies with mCRPC xenografts from cell line implants, four to six-week-old intact male athymic nu/nu mice (Charles River) were utilized. Mice were inoculated subcutaneously (~1.5 ×10⁶ cells) in the flank with a slurry of cells in 1:1 mixture (v/v) of media (DMEM) and Matrigel (Corning). Xenografts were generally palpable within 3–4 weeks after injection. Tumor bearing mice received ~300 μCi of ⁸⁹Zr-406 or ~300 μCi of ⁶⁸Ga-PSMA 11 for imaging studies. Mice bearing subcutaneous DU145 tumors received ¹⁷⁷Lu-4A06 (400 μCi) or vehicle (saline) via tail vein at day 0 and day 5 of the study period, which was approximately 14–21 days post tumor inoculation. Mice bearing subcutaneous C4-2B tumors received ¹⁷⁷Lu-4A06 (300 μCi) or vehicle (saline) via tail vein at day 0 and day 5 of the study period, which was approximately 14–21 days post tumor inoculation. Mice were arranged in treatment arms using a simple randomization approach (20). Animals were weighed at the time of injection, and once weekly until the completion of the study. Tumor volume measurements were calculated with calipers. The study endpoints were death due to tumor volume >2000 mm³ or ≥ 20% loss in mouse body weight. The researcher performing the tumor volume and body weight measurements was blinded to the treatment arms.

Small animal PET/CT:

4A06 was functionalized with desferrioxamine (DFO) and subsequently radiolabeled with Zr-89 as previously described (14). Tumor-bearing mice received ~200 μCi of ⁸⁹Zr-4A06 in 100 μL saline solution volume intravenously using a custom mouse tail vein catheter with a 28-gauge needle and a 100–150 mm long polyethylene microtubing. Mice were imaged on a small animal PET/CT scanner (Inveon, Siemens Healthcare, Malvern, PA). Animals were typically scanned for 30 minutes for PET, and the CT acquisition was performed for 10 minutes.

The co-registration between PET and CT images was obtained using the rigid transformation matrix generated prior to the imaging data acquisition since the geometry between PET and CT remained constant for each of PET/CT scans using the combined PET/CT scanner. For microPET/CT data, PET images were reconstructed using the ordered subsets expectation maximization algorithm (OSEM) provided by the scanner manufacturer. The parameters for OSEM were 16 subsets and 4 iterations, and the resulting reconstructed image volume was in a matrix of 128×128×159 with a voxel size of 0.0776 mm × 0.0775 mm × 0.0796 mm. CT images for attenuation correction were reconstructed using a conebeam Feldkamp algorithm provided by the scanner manufacturer. The data were acquired using x-ray tube voltage of 80 kVp and current of 0.5 mA for 120 angular steps over 220 degrees, and 175 ms of exposure at each angular step. The reconstructed CT volume was in a matrix of 512×512×700 with a voxel size of 0.195 mm × 0.195 mm × 0.195 mm. The precalibrated scaling was used to convert the CT images to attenuation maps for correction in PET reconstruction.

For SUV computation, we used freeware software, Amide (amide.sourceforge.net), and used its automated SUV calculation tool by entering decay-corrected injected activity and the animal weight. For each volume of interest, a spherical VOI (2–3 mm diameter) was drawn and SUV was calculated by VOI statistics.

Small animal SPECT/CT: 4A06 was functionalized with DOTA and radiolabeled with Lu-177 as previously described (14).

At 48 hours post injection of the second dose of ¹⁷⁷Lu-4A06 (i.e. day 7 overall of the antitumor assessment study), the mice were imaged under anesthesia using a small animal SPECT/CT (VECTor4CT, MILabs, Utrecht, The Netherlands). Animals were typically scanned for 40 minutes for SPECT, and the CT acquisition was performed for 10 minutes. The co-registered CT was used for photon attenuation correction in the SPECT reconstruction. For microSPECT/CT data, SPECT images were reconstructed using the similarity-regulated OSEM (SROSEM) provided by the scanner manufacturer. The parameters for SROSEM were 128 subsets and 10 iterations in a base voxel size of 1.2 mm. After reconstruction, a Gaussian postfilter with 1.5 mm full-width at half maximum was applied to suppress image noises. SPECT data were acquired using a multipinhole collimator (HE-GP-RM) with an aperture diameter of 3.6 mm that was designed for general purpose to scan both mice and rats with the axial field of view (FOV) of 18 mm and transverse FOV of 28 mm. Multiple bed positions during the data acquisition were used, controlled by the scanner to cover the whole mouse volume. Both photopeaks (171 keV and 245 keV) for Lu-177 with +/−10% energy window was used from the list mode data, and triple window based scatter correction was applied. After the SPECT reconstruction, the SPECT images were registered to the CT images, and attenuation correction using CT-based attenuation map was applied. CT data were acquired using x-ray tube voltage of 55 kVp and current of 0.19 mA for 480 angular steps over 360 degrees, and 75 ms of exposure at each angular step. CT images were reconstructed using the manufacturer provided conebeam Feldkamp algorithm.

Biodistribution studies:

At a dedicated time after radiotracer injection, animals were euthanized by cervical dislocation. Blood was harvested via cardiac puncture. Tissues were removed, weighed and counted on a Hidex automatic gamma counter (Turku, Finland). The activity of the injected radiotracer was calculated and used to determine the total number of counts per minute by comparison with a standard of known activity. The data were background- and decay-corrected and expressed as the percentage of the injected dose/weight of the biospecimen in grams (%ID/g).

Digital autoradiography:

Post mortem, tumors were harvested, transferred to sample boats, and immersed in Tissue–Plus OCT compound (Scigen, Gardena, CA). The tissues were snap frozen at −80° C. The tumor tissue was sectioned into 20 μm slices using a microtome (Leica, Buffalo Grove, IL) and mounted on glass microscope slides. The slides were loaded onto an autoradiography cassette and exposed with a GE phosphor storage screen for 24–72 hours at −20° C. The film was developed and read on a Typhoon 9400 phosphorimager (Marlborough, MA). Images of whole sections were acquired on a VERSA automated slide scanner (Leica Biosystems, Wetzlar, Germany), equipped with an Andor Zyla 5.5 sCMOS camera (Andor Technologies, Belfast, UK). ImageScope software (Aperio Technologies, Vista, CA) is used for creating individual images. Photoshop CS6 software (Adobe Systems, McLean, VA) were used for montage and processing.

Statistical analysis:

Binary comparisons between two treatment arms were made with an unpaired, two-tailed Student’s t-test. Differences at the 95% confidence level (P < 0.05) were considered to be statistically significant. Differences at the 95% confidence level were considered statistically significant. Unless otherwise stated, all data were expressed as mean ± standard deviation.

Overall survival was measured from the date of metastatic biopsy in the mCRPC patients. Kaplan-Meier product limit method and log-rank test were used to investigate the relationship between CDCP1 expression (transcripts per million, TPM) and overall survival, or CDCP1 expression, PTEN mutation status and overall survival. The patient cohort was dichotomized above and below median CDCP1 expression as well as broken into quartiles of CDCP1 expression for survival analyses.

Data Availability:

All data are available upon request to the corresponding authors of the study.

CDCP1 is expressed in mCRPC, including SCNC and adenocarcinoma with low levels of PSMA:

We probed our recently published RNA-seq data set of mCRPC biopsies to assess CDCP1 expression (19). Unlike a previously reported expression analyses by Alimonti et al. (9), the majority of the mCRPC biopsies from this data set were obtained post-abiraterone and/or enzalutamide from bone and soft tissue lesions, reflecting current treatment patterns. CDCP1 was expressed in 90% of mCRPC biopsies (107 of 119 samples, see Figure 1A). Since Alimonti et al. showed that CDCP1 mRNA and protein levels are upregulated in PTEN null prostate cancer, we tested for a correlation between CDCP1 and PTEN. A Pearson analysis revealed that CDCP1 levels were modestly, but significantly, inversely correlated (Supplemental Figure 1). We also compared CDCP1 mRNA levels between patients with wild type PTEN and patients with any type of PTEN mutation (i.e. deletion, inversion, break-end). We found that CDCP1 is slightly upregulated in PTEN-mutant patients, though the effect is not statistically significant (Table 1).

To understand if CDCP1 is expressed in tumors with low or undetectable PSMA, we next evaluated the distribution of CDCP1 versus FOLH1 (PSMA) expressing tumors. A Pearson analysis comparing the CDCP1 and FOLH1 expression per tumor showed that there was no significant correlation between the gene expression levels among the patients in the cohort (Figure 1B). Of the 119 tumors, 14 (12%) were predicted to have low or absent FOLH1 expression and positive CDCP1 expression. In addition, 93 of 119 (78%) were predicted to be mutually positive, 3 of 119 (3%) were mutually negative, and 9 of 119 (8%) were predicted to be FOLH1 positive but CDCP1 negative.

We further evaluated the protein expression of CDCP1 in mCRPC biopsies. Two blinded pathologists reviewed the biopsies for CDCP1 and PSMA staining, and a mean H score was reported. Of 17 evaluable biopsies, 15 (85%) were positive for cell surface expression of CDCP1, with H-score range of 15 – 285 (Figure 1C). Four samples (23%) had a higher mean H-score for CDCP1 compared to PSMA, including one example (biopsy #10) that overexpressed CDCP1 (H-score 285) but did express PSMA. Of note, one sample (biopsy #15) was negative on staining for both CDCP1 and PSMA. Representative stains are shown in Figure 1D.

CDCP1 overexpression was previously shown to be associated with poorer overall survival in mCRPC; therefore, we assessed this relationship in our cohort. In contrast to the former study, we found no significant association between CDCP1 expression and overall survival in dichotomized cohorts above and below median expression (log-rank p value = 0.2; Supplemental Figure 2A) or by breaking into quartiles of CDCP1 expression (log-rank p-value = 0.1; Supplemental Figure 2B). We also found no differences in survival when subdividing the patient cohort into subgroups based upon CDCP1 expression (above vs. below median) and PTEN mutation status (wild type vs. mutation), (log-rank p-value = 0.5; Supplemental Figure 2C).

Since CDCP1 protein was already found to be expressed at high levels in the whole cell lysates of several human prostate cancer cell lines (11), we evaluated expression in human PDX whole cell lysates (Supplemental Table 1). Full length and/or cleaved CDCP1 was expressed in four of five adenocarcinoma PDX models that we tested from the LuCaP and Living Tumor Laboratory series (Figure 2A). The highest levels were observed in LuCaP70 CR and LTL-484, an AR-expressing adenocarcinoma sample with negligible PSMA expression. CDCP1 expression was detected in six of seven SCNC PDXs, with the highest expression observed in LTL-545 and LTL-370 (Figure 2B). In line with some recent reports suggesting that proteolytically cleaved CDCP1 may contribute to early tumor development and metastatic potential (21–23), this proteolytic isoform of CDCP1 was detected in 4 of the 12 PDX samples (LuCaP70 CR, LuCaP 77 CR, LTL484, LTL545).

We next confirmed surface CDCP1 expression in prostate cancer cells using flow cytometry (Figure 2C). The relative cell surface levels in six human prostate cancer cell lines (PC3 > DU145 > C4-2B > LNCaP-AR > CWR22Pc > 22Rv1) approximated previously reported relative levels of total expression on immunoblot (11). To quantify CDCP1 receptor number per cell, we performed saturation binding assays using 4A06, a monoclonal recombinant human antibody that recognizes a region of the CDCP1 ectodomain present in both full length and cleaved CDCP1 (13). We calculated PC3 cells to have a B_max of 22,377 fmol/mg (~1.2 × 10⁶ receptors per cell), DU145 to have a B_max of 6819 fmol/mg (~3.7 × 10⁵ receptors per cell) and C4-2B to have a B_max of 2897 fmol/mg (~1.5 × 10⁵ receptors per cell, see Figure 2D). These values equal or exceed receptor densities reported for PSMA in human prostate cancer cell lines like LNCaP and MDA PCa 2b (~1 × 10⁵ receptors/cell) (24).

Tumor autonomous expression of CDCP1 in mCRPC is detectable with ⁸⁹Zr-4A06 PET/CT:

We next assessed CDCP1 expression in vivo using ⁸⁹Zr-labeled 4A06, an IgG1 monoclonal antibody that we previously developed that recognizes an epitope in the ectodomain found on both full-length and cleaved forms of CDCP1 (13). 4A06 was functionalized with desferrioxamine B and radiolabeled to high yield with Zr-89 using our previously reported method (14). We then evaluated tumor uptake over time in intact male nu/nu mice bearing subcutaneous C4-2B tumors (n = 4). Tumoral uptake of ⁸⁹Zr-4A06 was significantly higher than blood levels at 24 hours post injection (Figures 3A and 3B). Although tumoral uptake did not significantly change between 24 and 48 hours post injection, region of interest analysis suggested the tumor to blood ratio improved from 24 to 48 hours post injection.

As our previous experience with ⁸⁹Zr-4A06 suggested tumoral uptake typically peaks from 48–72 hours post injection, we next profiled CDCP1 expression at 48 hours post injection in intact male nu/nu or NSG mice bearing DU145, 22Rv1, LTL-545, LTL-331, or LTL-484 tumor xenografts (n = 4/tumor). These tumors represent AR-positive adenocarcinoma (22Rv1, LTL-331, LTL-484) and AR-null disease (DU145, LTL-545). ⁸⁹Zr-4A06 PET showed clear evidence of radiotracer accumulation in tumors above background levels in blood or skeletal muscle (Figure 3C). Post mortem biodistribution measurements at 48 hours post injection showed that DU145 had the highest radiotracer uptake at 12.9 ± 1.9% ID/g (Figure 3D and Supplemental Figure 3). The tumor to muscle ratios were at least 7:1 and DU145 had the highest ratio at ~25:1 (Supplemental Figure 4). The tumor to blood ratios were at least 2:1 and the LTL-331 cohort had the highest mean ratio of ~70:1 (Supplemental Figure 4). The blood values in the NSG cohorts were lower than those observed in the nu/nu cohorts, which is likely due to high splenic uptake of the IgG as has been previously noted (25). Digital autoradiography was performed to assess radiotracer distribution within a tumor representing high (DU145) and comparatively lower (22Rv1) tracer uptake. We found that radiotracer distribution was qualitatively similar, and most abundant around the periphery of the tumor as expected for a large immunoglobulin like 4A06 (Supplemental Figure 5). Lastly, we compared ⁶⁸Ga-PSMA 11 uptake to ⁸⁹Zr-4A06 uptake in intact male mice bearing LTL-484 tumors (n = 4). Consistent with the protein expression patterns, ⁸⁹Zr-4A06 uptake in tumors was significantly higher than ⁶⁸Ga-PSMA 11 (Figure 3E).

Radioligand therapy with ¹⁷⁷Lu-4A06 inhibits mCRPC tumor growth:

We next tested if CDPC1 can be targeted for RLT in mCRPC tumors. We first tested antitumor effects in PSMA-null tumors. Intact male nu/nu mice bearing PSMA-negative DU145 tumors received ¹⁷⁷Lu-4A06 in two fractions of 400 μCi at day 0 and day 5 of the study. We chose this dosing regimen as we previously showed it to be more efficacious that a single fraction of the same total radioactive dose (i.e. 800 μCi) (14). SPECT/CT imaging showed effective tumor targeting at 72 hours after the first injection (Figure 4A). ¹⁷⁷Lu-4A06 treatment significantly delayed tumor growth compared to vehicle with no evidence for unhealthy weight loss (Figure 4B and Supplemental Figures 6–7). To test if CDCP1 RLT is efficacious against AR-positive, PSMA-positive tumors, we treated intact male nu/nu mice bearing subcutaneous C4-2B tumors with ¹⁷⁷Lu-4A06 (2 fractions of 300 μCi at day 0 and 5) or vehicle. ¹⁷⁷Lu-4A06 significantly inhibited tumor growth in this model as well, and tumor regressions were observed in two of five mice (Figure 4C–4D and Supplemental Figures 8 and 9).