A Novel Alphavirus Vaccine Encoding Prostate-Specific Membrane Antigen Elicits Potent Cellular and Humoral Immune Responses Free

Prostate cancer is the most prevalent type of noncutaneous cancer in American men with an estimated 234,000 diagnosed cases and ∼30,000 deaths in 2005 (1). Localized disease is generally treated with surgery or radiation therapy, and although recurrent or metastatic disease can be controlled temporarily with hormonal therapy, almost all prostate carcinomas eventually become androgen independent and then progress rapidly. Docetaxel in combination with prednisone is the only available nonpalliative therapy for androgen-independent prostate cancer. Treatment provides a modest survival benefit (2), and there is an urgent need for novel, molecularly targeted therapies. Metastatic prostate cancer is characterized by relatively small lesions within lymph node, bone marrow, and other sites within the immune system. These features may render prostate cancer more susceptible to vaccine immunotherapy than tumors that are characterized by bulkier or less accessible lesions.

Prostate-specific membrane antigen (PSMA) is the prototypic cell surface marker of prostate cancer. PSMA is an ∼100-kDa type II integral membrane glycoprotein. In normal tissues, PSMA expression is largely limited to prostate. However, PSMA expression is up-regulated in prostate carcinomas (3, 4) and increases with disease progression, becoming highest in metastatic, hormone-refractory disease (5, 6). In addition, PSMA is abundantly expressed on the neovasculature of many other solid tumors but not on normal neovasculature (7–11). For these reasons, PSMA is considered to be an attractive target for cancer immunotherapy (reviewed in refs. 12, 13).

Alphaviruses are positive-strand RNA viruses that can mediate high-level gene expression in mammalian cells, and members of this family have served as a basis for viral vector and DNA plasmid vaccines for infectious diseases and cancer (reviewed in ref. 14). Recent preclinical studies have shown potent immunogenicity and antitumor effects for alphavirus vaccines encoding self-antigens (15–17). A propagation-defective vaccine replicon particle (VRP) vector system has been developed based on an attenuated variant of Venezuelan equine encephalitis virus (18, 19). Using tyrosinase as a prototypical melanocyte differentiation antigen, we recently showed that Venezuelan equine encephalitis virus–based VRP induced robust B-cell and T-cell responses and improved survival in mice challenged with B16 melanoma cells. In this setting, more potent and protective responses were elicited using a syngeneic (murine) rather than a xenogeneic (human) form of tyrosinase (17). Based on these findings, we prepared VRP encoding human PSMA for preclinical evaluation as a potential vaccine therapy for prostate cancer in man.

In the present study, we show that PSMA-VRP mediate high-level expression of native, enzymatically active, cell surface PSMA in vitro and potently elicit Th-1–biased cellular and humoral immunity to PSMA in vivo. Anti-PSMA responses were induced following a single injection of PSMA-VRP, and these responses were boosted on repeated dosing. The cellular immune responses in BALB/c and C57BL/6 mice were determined to be directed toward different defined peptide sequences within PSMA. Anti-PSMA immunity was elicited with a favorable safety profile both in mice and in a rabbit toxicology study that was conducted in compliance with Good Laboratory Practice regulations.

Cell lines. BHK-21 [Syrian hamster kidney fibroblast (20)] and Vero [African green monkey kidney (21)] cell lines were obtained from the American Type Culture Collection. Dr. Michelle Sadelain (Memorial Sloan-Kettering Cancer Center, New York, NY) generously provided CT26 cells (22) transduced with PSMA-encoding and empty retroviral vectors (CT26-PSMA and CT26-vector cells). CT26 is a murine colon adenocarcinoma that is syngeneic to BALB/c mice. 3T3-PSMA cells, parental 3T3 cells (23), and C4-2 cells (24), an androgen-independent subclone of the LNCaP prostate cancer cell line, have been described previously. All cell lines, except for C4-2 cells, were cultured in DMEM (Invitrogen) containing 10% heat-inactivated fetal bovine serum (Hyclone), 1 mmol/L sodium pyruvate, 100 μmol/L MEM nonessential amino acids, and 2 mmol/L GlutaMAX Supplement (all from Invitrogen). C4-2 cells were grown in RPMI 1640 (Invitrogen) containing the same supplements as listed for DMEM.

Generation of VRP. VRPs were generated as described previously (18). Briefly, the genes for human PSMA or green fluorescent protein (GFP) were directionally cloned into the replicon expression vector (phosphorylated extracellular signal-regulated kinase) and fully sequenced to verify authenticity. Full-length PSMA and GFP replicon RNA transcripts were generated in vitro using the RiboMax RNA Transcription kit (Promega). For VRP production, PSMA or GFP replicon RNA transcripts were mixed with two helper RNA transcripts encoding the Venezuelan equine encephalitis virus envelope and capsid proteins and then coelectroporated into BHK-21 or Vero cells in PBS using a GenePulser II electroporator (Bio-Rad). Following the incubation of cells for 16 to 24 h at 37°C, PSMA-VRP or GFP-VRP were harvested from the culture supernatant, purified by heparin chromatography, and stored at −80°C. To determine the infectious titers of the VRP preparations, expressed in infectious units (IU) per mL, Vero cells were infected with serial dilutions of PSMA-VRP or GFP-VRP for 1 h, washed, and incubated overnight. GFP expression was detected by fluorescence microscopy, whereas PSMA expression was measured using PSMA monoclonal antibody (mAb) CYT-351 (Cytogen) as a primary detection reagent followed by detection with FITC-conjugated anti-mouse antibody (Caltag).

PSMA peptides and PSMA_ECTO. The deduced protein sequence of full-length human PSMA (Genbank accession no. AF007544) was spanned using 15-mer peptides that overlapped by 10 amino acids (SynPep). Peptides were synthesized to >90% purity and identity tested by mass spectroscopy. Three peptides (LCAGALVLAGGFFLL, GFFLLGFLFGWFIKS, and IHSTNEVTRIYNVIG) could not be purified to >90% and were pooled separately (Pool-11). Six additional peptides (LVLAGGFFLLGFLFG, NFTQIPHLAGTEQNF, NYISIINEDGNEIFN, DMKINCSGKIVIARY, VKNFTEIASKFSERL, and QIYVAAFTVQAAAET) could not be successfully synthesized and were excluded from analysis. Individual peptides were pooled into nonoverlapping pools and lyophilized as described previously (25). Pools contained 13 to 15 peptides, except for Pool-11. Peptide pools were solubilized in DMSO (Sigma) at 10 to 20 mg/mL total peptide, whereas individual peptides were resuspended at 5 to 10 mg/mL. Following reconstitution, individual peptides and pools were aliquoted and stored at −80°C before use. The same set of peptide pools was used throughout this study.

Purified, dimeric PSMA_ECTO protein was produced as described previously (26). Briefly, the extracellular domain of PSMA (amino acids 44-750) was expressed in secreted form in stably transfected Chinese hamster ovary cells and purified to homogeneity by column chromatography. Purified PSMA_ECTO was stored at −80°C before use.

Western blot. BHK-21 cells at >80% confluency were infected in a six-well dish with PSMA-VRP or GFP-VRP at a multiplicity of infection (MOI) of 1. After 18 h of incubation at 37°C, VRP-infected cells and LNCaP cells were lysed in radioimmunoprecipitation assay buffer [1 mmol/L EDTA, 1% NP40 (Sigma), 1% Triton X-100 (Sigma), protease inhibitor cocktail (Sigma) in Dulbecco's PBS (DPBS; Invitrogen)] and cell debris was removed by centrifugation. Aliquots of extracts were boiled for 5 min under reducing conditions and loaded onto 4% to 15% Tris-glycine gels (Bio-Rad). Gels were transferred to nitrocellulose using a Trans-Blot SD (Bio-Rad) for 45 min at 15 V. Blots were blocked for 1 h at room temperature in blocking buffer [5% nonfat dry milk in 10% SDS (Invitrogen), 0.1% Triton X-100 in DPBS]. Blots were incubated for 1 h in 1 μg/mL PSMA mAb 544 (Maine Biotechnology) in blocking buffer, quickly rinsed twice, and washed thrice for 10 min and then incubated for 1 h with secondary antibody (goat anti-mouse IgG-horseradish peroxidase; Pierce) at 1:5,000 in blocking buffer. Blots were rinsed twice and washed once for 45 min and once for 5 min in wash buffer (10% SDS, 0.1% Triton X-100 in DPBS). Blots were developed using Western Lightning Plus (Perkin-Elmer) and X-OMAT XB-1 film (Kodak).

Flow cytometry assay for PSMA expression by PSMA-VRP. BHK-21 cells at >80% confluency were infected with PSMA-VRP at a MOI of 0, 0.003, 0.03, 0.3, or 3.0. After 18 h at 37°C, cells were washed in DPBS, released with cell dissociation solution (Sigma), and washed once in staining buffer [0.25% bovine serum albumin (Sigma) in DPBS]. Following a 20-min incubation at 4°C in blocking buffer [10% heat-inactivated goat serum (Invitrogen) in staining buffer], samples were stained with 1 μg of PSMA mAb 3.9 (27) for 30 min at 4°C. Cells were washed and centrifuged once with 10 volumes of staining buffer at 4°C. FITC-conjugated anti-mouse IgG was added at 0.5 μg/sample in staining buffer for 20 min at 4°C. Samples were again washed and centrifuged in staining buffer and analyzed on a FACSCalibur flow cytometer (Becton Dickinson). Mean fluorescence intensities (MFI) were calculated using CellQuest Pro software (Becton Dickinson).

Immunohistochemistry of PSMA-VRP–infected cells. BHK-21 cells at >80% confluency were infected with PSMA-VRP at MOI of 0, 0.01, 0.1, or 1.0. After 18 h at 37°C, cells were washed in DPBS, released with cell dissociation solution, embedded with OCT medium (Miles), and frozen at −80°C. Sections 5 to 7 μm thick were cut on a Mictron HM505E cryostat at −10°C to −25°C and thaw mounted onto slides. Slides were incubated in Morphosave solution (Ventana Medica) for 15 min after fixation in acetone. Slides were washed twice in PBS, incubated for 20 min with 0.3% H₂O₂ in PBS, and washed once with PBS. After incubation with mouse PSMA mAb 3.1 (26) or isotype control antibody at 5 μg/mL for 30 min, slides were washed in Tris buffer, incubated with horseradish peroxidase–conjugated goat anti-mouse secondary antibody (DAKO) for 30 min, washed in Tris, and incubated twice for 5 min in 3,3′-diaminobenzidine (Sigma). Slides were washed again in Tris, counterstained in hematoxylin, rinsed in water, and finally dehydrated through a graded series of alcohols to xylene before analysis by light microscopy through a 20× Nikon Plan Apo objective using a Nikon Optiphot-2 with an Olympus Q-Color 3 camera system.

Folate hydrolase assay. BHK-21 cells were infected in T-150 flasks with PSMA-VRP or GFP-VRP at MOI of 1. After 18 h at 37°C, cells were harvested as described above, washed twice with PBS, and stored at −80°C before analysis. Cells were thawed, suspended in 50 mmol/L Tris (pH 7.5; 10 mL per T-150 flask), and disrupted with a Dounce homogenizer. Cell debris was removed by centrifugation at 3,000 × g for 5 min at 4°C, and the supernatant was then centrifuged at 70,000 × g for 35 min at 4°C. The pellet was washed in Tris buffer and homogenized in 1% Triton X-100 in Tris buffer. Protein concentration was determined by bicinchoninic acid assay, and 20 μg were combined with 5 μmol methotrexate diglutamate (Schircks Laboratories) in Tris buffer for 1 h at 37°C. The reaction was stopped by addition of Na₂HPO₄, and the mixture was analyzed by reversed-phase high-performance liquid chromatography as described previously (28, 29).

Mouse immunizations. BALB/c and C57BL/6 mice (Charles River Laboratories) were immunized by s.c. injections of PSMA-VRP or GFP-VRP in a total volume of 20 μL into the plantar surface of the left hind footpad using a Microfine-IV insulin syringe (Becton Dickinson) every 2 weeks. Tolerability was assessed by monitoring the appearance and weight of immunized animals. BALB/c mice (five per group) were immunized with 10 μg PSMA_ECTO adsorbed onto 250 μg Alhydrogel (alum, Brenntag Biosector) according to published methods (26). Vaccine was administered s.c. in the leg at weeks 0, 2, and 6. The 4-week rest period before the final immunization was selected based on our prior study of PSMA_ECTO/alum (26). Sera and splenocytes were collected for analysis 2 weeks after the final immunization.

Serology assays for PSMA antibodies. Blood was isolated from the mouse orbital sinus plexus using a nonheparinized glass capillary tube and collected into a CapiJect capillary blood collection tube (Terumo Medical Corp.) containing a gel/clot activator. Serum was isolated by centrifugation and stored at −20°C before analysis by flow cytometry and ELISA. For flow cytometry analyses, serum was diluted 1:100 in staining buffer and added to 2 × 10⁵ 3T3 or 3T3-PSMA cells in blocking buffer (10% heat-inactivated goat serum in staining buffer) for 30 min at 4°C. Cells were washed with 10 volumes of staining buffer at 4°C. Secondary FITC-conjugated anti-mouse IgG antibody was added for 20 min at 4°C. Samples were washed in staining buffer and analyzed by flow cytometry as described above.

For ELISA analyses, 3T3-PSMA cells were plated in Biocoat poly-l-lysine 96-well assay plates (BD Biosciences) overnight at 37°C. Wells were washed once with wash buffer (DPBS containing 5% fetal bovine serum) and blocked for 2 h at room temperature with DPBS containing 3% bovine serum albumin. Wells were decanted, and serum from immunized animals was added for 2 h at room temperature. Plates were washed thrice in wash buffer, and alkaline phosphatase–conjugated goat antibody to mouse IgG or mouse IgG subtypes (Caltag) was added at 1:1,000 dilution for 1 h at room temperature. Plates were washed thrice in wash buffer, developed for 1 h at 25°C with pNPP/DEA substrate (Pierce), and read on a SpectraMax 340 plate reader at wavelengths of 405 and 650 nm. End point titer was defined as the highest dilution of the immune serum yielding an absorbance of 0.100 above background.

Isolation of T lymphocytes. Spleens were harvested from immunized animals, and splenocytes were isolated by manual disruption of the spleen using aseptic technique. Single-cell suspensions were prepared by passage of cells through a 70-μm filter (BD Biosciences). Red cells were lysed using ACK lysing buffer (Cambrex) following the manufacturer's instructions, and splenocytes were resuspended into T-cell medium (RPMI 1640/MEM, α-medium 1:1 with 10% heat-inactivated fetal bovine serum, 10 mmol/L HEPES, 1 mmol/L sodium pyruvate, 100 μmol/L MEM nonessential amino acids, 2 mmol/L GlutaMAX Supplement, and 55 nmol/L 2-mercaptoethanol, all from Invitrogen). CD8⁺ T lymphocytes were isolated from bulk splenocytes by positive selection using anti-mouse CD8a (Ly-2) microbeads, whereas CD4⁺ T lymphocytes were isolated by positive selection with anti-mouse CD4 microbeads (Miltenyi Biotec) according to the manufacturer's guidelines. Purity of the isolated subsets was analyzed by flow cytometry and typically was >95%.

Enzyme-linked immunospot assay. Multiscreen nitrocellulose or Multiscreen HTS IP enzyme-linked immunospot (ELISPOT) plates (Millipore) were coated overnight at 4°C with 1 μg/well anti-mouse IFN-γ mAb (clone AN18; Mabtech) and blocked for 2 h at 37°C in T-cell medium (see above). Splenocytes (3 × 10⁵ per well) were allowed to settle onto the membrane for 30 min at 37°C. CT26-vector or CT26-PSMA cells (H-2^d positive, BALB/c syngeneic stimulator cells) were pretreated for 30 min at 37°C with 100 μg/mL mitomycin C (Sigma), washed, and allowed to recover for 1 h at 37°C before addition at 2 × 10⁵ per well. As controls, either medium alone or phorbol 12-myristate 13-acetate and calcium ionomycin (Sigma) were added at final concentrations of 1 and 0.5 μg/mL, respectively. Following 18 h of incubation at 37°C, cells were decanted and plates were washed six times with wash buffer [0.05% Tween 20 (Sigma) in DPBS without Ca/Mg (DPBS⁻; Invitrogen)]. Biotinylated anti-mouse IFN-γ mAb (clone R4-6A2; Mabtech) was added at 0.4 μg/mL in 0.5% bovine serum albumin in DPBS⁻ to each well and the plates were incubated for 2 h at 37°C. After washing the plates six times in wash buffer, 100 μL of enzyme complex (Vectastain Standard Elite avidin-biotin complex kit, Vector Laboratories) were added for 1 h at ambient temperature. Plates were washed thrice in wash buffer and thrice in PBS. Excess liquid was removed, and wells were developed for 4 min at room temperature with 3-amino-9-ethylcarbazole (Vector Laboratories). The substrate was decanted and the plates were washed with water before drying. Membranes were transferred to sealing tape (Millipore) and spots were counted on a KS ELISPOT System (Carl Zeiss). For the anti-interleukin-4 (IL-4) ELISPOT, plates were coated with anti-mouse IL-4 mAb (clone 11B11; Mabtech) and developed using biotinylated anti-mouse IL-4 mAb (clone BVD6-24G2; Mabtech) after a 40-h incubation at 37°C. Background was defined as the number of IFN-γ or IL-4 spot-forming cells (SFC) per number of cells plated in the presence of medium only (in all cases, background SFCs per 3 × 10⁵ cells were <10). A positive response was defined as greater than 3 SDs above the mean background signal.

For peptide-based ELISPOT assays, bulk splenocytes from immunized animals were plated at 3 × 10⁵ per well in IFN-γ-coated ELISPOT plates as described above. Alternatively, CD8⁺ or CD4⁺ T cells were isolated as described above and mixed in a 2:1 ratio with CD8/CD4-depleted splenocytes from nonimmunized mice to generate 3 × 10⁵ per well. Data are reported as SFCs per 3 × 10⁵ CD8⁺ or CD4⁺ cells for consistency. Cells were stimulated with individual or pooled PSMA peptides or pooled HIV gag peptides (p55, BD Biosciences) at a concentration of 2 μg/mL of each individual peptide, and the ELISPOT assay was developed as described above.

Dendritic cells were used as stimulators in some ELISPOT assays. Briefly, bone marrow cells were isolated from the tibia and femurs of naive BALB/c mice, and mononuclear cells were isolated using Lympholyte-M (Cederlane Laboratories) according to the manufacturer's instructions. Cells were grown in T-cell medium (see above) supplemented with 1,000 units/mL of mouse granulocyte macrophage colony-stimulating factor (GM-CSF) and mouse IL-4. Cells were cultured for 8 days in the presence of GM-CSF/IL-4 or for 6 days with GM-CSF/IL-4 followed by 2 days with GM-CSF/IL-4 plus 250 ng/mL murine tumor necrosis factor-α. Fresh cytokine-containing medium was added every other day during culture. Cytokines were purchased from R&D Systems. Cells were analyzed by flow cytometry for dendritic cell markers and infected at MOI of 5 with PSMA-VRP and GFP-VRP or left uninfected. After 18 h at 37°C, cells were washed and combined with 3 × 10⁵ splenocytes, and ELISPOT plates were processed as described above.

CTL assay. Splenocytes were isolated from immunized BALB/c mice as described above and incubated at 5 × 10⁶ cells/mL with PSMA peptide 16 (NFTQIPHLAGTEQNF) at 10 μg/mL for 5 days. CT26-PSMA or CT26-vector target cells were labeled with 250 μCi of Na₂⁵¹CrO₄ for 1 h and washed. Target cells and splenocyte effector cells were combined for 4 h in duplicate wells of a 96-well round-bottomed plate. Plates were centrifuged, and supernatants were harvested using a Skatron Supernatant Collection System (Molecular Devices). Specific lysis was determined following γ-counting (Perkin-Elmer) using the following formula: % specific lysis = 100 × (test release in the presence of effector cells − spontaneous release) / (maximal release − spontaneous release). Spontaneous release was determined based on wells containing labeled target cells only, whereas maximum release was determined by lysing labeled target cells with 1 N HCl.

Rabbit toxicology study. A Good Laboratory Practice–compliant toxicology study was done in male New Zealand White rabbits. Animals (n = 6) were inoculated s.c. with four doses of 2 × 10⁷ IU or 2 × 10⁸ IU of PSMA-VRP. A negative control group received vehicle only. Animals were immunized at 2-week intervals, and half of the animals were sacrificed for histopathology either 2 days or 2 weeks after the final injection. Systemic toxicity was evaluated by recording mortality/morbidity, body temperature, body weight, food consumption, and ophthalmic examinations. Hematopoietic toxicity was evaluated by quantitating cellular components of peripheral blood. Potential immune system and autoimmune toxicities were assessed by histopathologic examination of lymphoid organs, prostate, brain, duodenum, kidneys, and other tissues. Organ toxicity also was assessed by monitoring clinical chemistry variables. Local reactogenicity was evaluated by examining injection sites both grossly and microscopically. Preimmune and postimmune sera were tested at a 1:200 dilution for binding to 3T3-PSMA cells by flow cytometry as described above.

PSMA-VRP mediates high-level expression of enzymatically active PSMA. To examine the expression of PSMA by PSMA-VRP in vitro, BHK-21 cells were infected with PSMA-VRP or GFP-VRP (control VRP) at a MOI of 1. Whole-cell lysates were analyzed for PSMA expression by Western blotting (Fig. 1A). A single band of ∼100 kDa was identified in cells infected with PSMA-VRP, but not GFP-VRP, and this band comigrated with PSMA expressed in LNCaP cells, suggesting that PSMA-VRP directs expression of full-length PSMA with a normal extent of glycosylation.

BHK-21 cells were infected with PSMA-VRP and further analyzed by flow cytometry with a conformation-dependent anti-PSMA mAb (26). PSMA was detected at the surface of infected cells, and expression levels increased with increasing MOI ranging from 0.003 to 3.0 (Fig. 1B). No PSMA expression was detected in uninfected cells (Fig. 1B) or in cells infected with GFP-VRP (data not shown). In addition, immunohistochemistry was used to examine PSMA expression in BHK-21 cells infected with PSMA-VRP at MOI of 0.01, 0.1, and 1.0 (Fig. 1C). Consistent with the flow cytometry data, PSMA expression increased as the MOI was increased from over this range, and no staining was observed for uninfected cells (Fig. 1C) or for cells stained with isotype control antibody (data not shown). At the highest MOI (1.0), PSMA expression was comparable with or greater than that on PSMA-3T3 cells (data not shown), which express PSMA at >10⁶ copies/cell (30). These studies show that PSMA-VRP directs high-level surface expression of native PSMA in mammalian cells in vitro.

BHK-21 cells were evaluated for folate hydrolase activity following infection with PSMA-VRP. Uninfected cells and cells infected with GFP-VRP served as controls. Consistent, high-level enzyme activity was observed in cells infected with PSMA-VRP (Fig. 1D). The level of activity (∼50 nmol glutamate released/mg protein/h) in infected BHK-21 cells was comparable with that observed in C4.2 cells, an androgen-independent prostate cancer cell line with high endogenous expression of PSMA (29). In contrast, no significant activity was observed for uninfected cells or cells infected with GFP-VRP. The findings indicate that PSMA-VRP induces expression of enzymatically active PSMA.

PSMA-VRP induces potent, Th-1–biased humoral and cellular immune responses. BALB/c mice were vaccinated with either a single s.c. injection or repeated biweekly s.c. injections of PSMA-VRP, and serum was collected for analysis of the humoral response by flow cytometry and ELISA. A single injection with 2 × 10⁵ IU PSMA-VRP induced a significant anti-PSMA IgG response as measured by flow cytometry against 3T3-PSMA cells, whereas repeated immunizations substantially augmented the PSMA antibody response (Fig. 2A). In a second set of immunizations in BALB/c mice (n = 5/group), three doses of 2 × 10⁵ IU PSMA-VRP were compared directly with three doses of 10 μg PSMA_ECTO/alum, and sera were analyzed in parallel by flow cytometry. In this experiment, PSMA-VRP induced significantly (P = 0.012, two-sided t test) higher levels of anti-PSMA antibodies (MFI = 148 ± 29) than did PSMA_ECTO/alum (MFI = 94 ± 23).

As measured by ELISA, mean PSMA antibody end point titers increased from 1:200 following a single 2 × 10⁵ IU dose to 1:3,200 following three doses of 2 × 10⁵ IU PSMA-VRP (data not shown). We next used the PSMA-specific ELISA to characterize the isotype distribution of the anti-PSMA IgG antibodies induced by immunization with PSMA-VRP. The predominant isotypes were identified to be IgG2a and IgG2b (Fig. 2B). IgG2a was detected with 7-fold greater efficiency than IgG1, and this distribution is reflective of a Th-1 response.

Splenocytes from PSMA-VRP–vaccinated mice were analyzed by an IFN-γ ELISPOT assay to assess the induction of the cellular immune response to PSMA. Bulk splenocytes from immunized BALB/c mice were stimulated with syngeneic CT26-PSMA or CT26-vector cells. A significant PSMA-specific IFN-γ secretion (95 ± 33 SFCs per 3 × 10⁵ splenocytes) was observed from splenic T cells after a single 2 × 10⁵ IU immunization, and this response was increased 5-fold (456 ± 90 SFCs per 3 × 10⁵ splenocytes) following three biweekly s.c. vaccinations (Fig. 2C). In contrast, background levels of SFC were observed following three immunizations with GFP-VRP or using CT26-vector cells for stimulation (Fig. 2C). Significant anti-PSMA B-cell and T-cell responses were observed even 1 year after the final immunization (data not shown), indicating that the immunity is long lived.

In contrast to the robust PSMA-specific IFN-δ responses elicited by PSMA-VRP (Fig. 2C), no significant IL-4 secretion was observed by ELISPOT following stimulation of cells with CT26-PSMA (data not shown). This Th-1 cytokine profile is consistent with the Th-1 bias of the antibody response (Fig. 2B). In addition, no PSMA-specific IFN-γ response was observed for bulk splenocytes from animals immunized with three doses of PSMA_ECTO/alum (<2 SFCs per 3 × 10⁵ cells).

Dose dependency of the immune response elicited by PSMA-VRP. PSMA-specific cellular and humoral immune responses were evaluated in BALB/c mice following three biweekly s.c. injections of PSMA-VRP at dose levels of 10², 10⁴, and 10⁶ IU. Control animals received three injections of 10⁶ IU GFP-VRP. A 10² IU dose of PSMA-VRP induced measurable levels of PSMA antibodies in three of five mice, whereas all animals seroconverted at the two higher dose levels. The MFI values were 16 ± 11, 68 ± 31, and 112 ± 14 for animals immunized with 10², 10⁴, and 10⁶ IU PSMA-VRP, respectively, whereas MFI values were 6 ± 1 for animals immunized with 10⁶ IU GFP-VRP (Fig. 2D).

Splenic CD8⁺ T cells were isolated from these animals and tested for IFN-γ secretion by ELISPOT assay using CT26-PSMA or CT26-vector cells for stimulation. Compared with the humoral response, the cellular response to PSMA-VRP was similarly dose dependent (Fig. 2E). At a dose of 10² IU PSMA-VRP, the number of SFC (13 ± 3 per 3 × 10⁵ cells) was marginally but not significantly higher than that observed for animals immunized with 10⁶ IU GFP-VRP. However, a 10⁴ IU dose elicited significant PSMA-specific IFN-γ secretion (162 ± 13 SFCs per 3 × 10⁵ cells), and this response was 4-fold higher at 10⁶ IU (662 ± 65 SFCs per 3 × 10⁵ cells; Fig. 2D). The weight and appearance of animals were monitored at least twice weekly, and no sign of toxicity was observed by these measures.

PSMA-VRP elicits a functional CTL response. We next used a chromium release assay to examine the CTL activity elicited by PSMA-VRP. Splenocytes from BALB/c mice immunized with PSMA-VRP or vehicle were used as effectors, and syngeneic CT26-PSMA and CT26-vector control cells were used as targets. Potent CTL activity was observed for PSMA-VRP splenocytes, which lysed up to 60% to 70% of CD26-PSMA targets (Fig. 2F). Cell killing was PSMA specific, as <5% cell lysis was observed for PSMA-VRP splenocytes tested against CT26-vector controls or for splenocytes from vehicle-immunized animals tested against either CT26-PSMA or CT26-vector cells (Fig. 2F).

Mapping the anti-PSMA cellular immune response in BALB/c and C57BL/6 mice. To dissect the epitope specificity of the immune response induced by PSMA-VRP, peptides spanning the entire amino acid sequence of human PSMA were used to stimulate lymphocytes from immunized mice in an IFN-γ ELISPOT assay. IFN-γ ELISPOT data obtained on total splenocytes from PSMA-VRP immunized BALB/c mice identified a single peptide pool (PSMA Pool-6) that is responsible for the immune reactivity in this assay (Fig. 3A). The IFN-γ response directed to Pool-6 (115 ± 39 SFCs per 3 × 10⁵ cells) was similar in magnitude to that directed against CT26-PSMA stimulator cells (151 ± 20 SFCs per 3 × 10⁵ cells). In contrast, no significant IFN-γ responses were observed for splenocytes from animals immunized with PSMA_ECTO/alum (≤4 SFCs per 3 × 10⁵ cells for all peptide pools).

Bulk splenocytes and purified CD8⁺ T cells were next tested against the individual peptides that comprise Pool-6 (Fig. 3B). In this assay, a potent immune response (416 ± 33 SFCs per 3 × 10⁵ CD8⁺ T cells) was observed against a single 15-mer peptide, Peptide 16 (PSMA_76-90, NFTQIPHLAGTEQNF). The splenocyte response to Peptide 16 was similar in magnitude to that observed for Pool-6 (384 ± 12 SFCs per 3 × 10⁵ splenocytes), and no significant response was observed to the other peptides within Pool-6. The findings from the peptide-based IFN-γ ELISPOT assays indicate that the CD8⁺ T-cell response is targeted to an immunodominant determinant in Peptide 16.

Purified CD4⁺ T cells also readily responded to Peptide 16, albeit at a substantially lower level (51 ± 12 SFCs per 3 × 10⁵ CD4⁺ T cells; Fig. 3C), indicating that this peptide may contain a MHC class II determinant. The purity of isolated CD8⁺ and CD4⁺ T-cell subsets was typically >95% as determined by flow cytometry analysis (data not shown); however, we cannot exclude the possibility that the response observed in the CD4⁺ T-cell population against Peptide 16 was due to low-level presence of PSMA-specific CD8⁺ T cells. In addition, CD4⁺ T cells also showed a weak response (17 ± 5 SFCs per 3 × 10⁵ CD4⁺ T cells) against Peptide 146 (also a component of Pool-6; Fig. 3C). The sequence for Peptide 146, PSMA_726-740 (GEVKRQIYVAAFTVQ), partially corresponds to one previously predicted to bind to human HLA-DR class II (31).

We next examined whether dendritic cells could be transduced by PSMA-VRP in vitro and used to stimulate T cells from immunized animals. Dendritic cells from naive BALB/c mice were matured from bone marrow cells in vitro, infected with PSMA-VRP or GFP-VRP, and combined with splenocytes from mice immunized with PSMA-VRP or vehicle in an IFN-γ ELISPOT assay. Dendritic cells transduced with PSMA-VRP in vitro efficiently stimulated splenocytes from animals immunized with PSMA-VRP but not animals immunized with vehicle (Fig. 3D). Dendritic cells that were matured using GM-CSF plus IL-4 alone were more effective stimulators (82 ± 16 SFCs per 3 × 10⁵ splenocytes) than dendritic cells matured using GM-CSF, IL-4, and tumor necrosis factor-α (43 ± 8 SFCs per 3 × 10⁵ splenocytes). Regardless of the method of dendritic cell maturation, the number of SFC observed following transduction with PSMA-VRP was significantly greater (P < 0.02, two-sided t test) than that observed either with cells transduced with GFP-VRP or with uninfected cells (Fig. 3D).

A second mouse strain, C57BL/6, was vaccinated to determine if the breadth and potency of PSMA-VRP–induced immune response are MHC haplotype dependent. Similar to data obtained in BALB/c mice, 2 × 10⁵ IU PSMA-VRP elicited high levels of PSMA antibodies in C57BL/6 mice (123 ± 31 MFI; data not shown) and a vigorous cellular response in the peptide-based ELISPOT assay (178 ± 49 SFCs per 3 × 10⁵ splenocytes; Fig. 3E). However, the cellular response in C57BL/6 mice was focused toward a different PSMA peptide pool (Pool-7; Fig. 3E). In addition, responses of lower magnitude (<20 SFCs per 3 × 10⁵ splenocytes) were observed for other pools (1, 2, 3, 5, and 9) in C57BL/6 mice, indicating additional reactivities. Notably, the C57BL/6 splenocytes did not respond to Pool-6 (Fig. 3E), which was the focus of the response elicited by PSMA-VRP in BALB/c animals (Fig. 3A). The nonoverlapping determinants of IFN-γ responses in C57BL/6 mice (primarily Pool-7) and BALB/c mice (primarily Pool-6) provide additional specificity controls of our ELISPOT methods.

Using individual peptides from Pool-7, the major CD8⁺ T-cell reactivity in C57BL/6 mice was mapped to Peptide 127 (PSMA_631-645, SLFSAVKNFTEIASK, 414 ± 42 SFCs per 3 × 10⁵ CD8⁺ cells; Fig. 3E), whereas CD4⁺ T cells responded strongly to Peptide 47 (PSMA_231-245, PADYFAPGVKSYPDG, 181 ± 21 SFCs per 3 × 10⁵ CD4⁺ cells; Fig. 3E). CD8⁺ T cells also displayed low but detectable responses (18-37 SFCs per 3 × 10⁵ CD8⁺ cells) against several other peptides within Pool-7 (Fig. 3F). These additional responses were not detected in bulk splenocytes, possibly due to the smaller proportion of responding cells in the unfractionated population.

Whereas the CD4⁺ T cells in both BALB/c- and C57BL/6-vaccinated mice secreted IFN-γ on stimulation with PSMA peptides (Fig. 3), these cells did not secrete IL-4 in parallel ELISPOT assays (data not shown). However, these cells effectively secreted IL-4 when stimulated with mitogen (176 ± 34 per 2 × 10⁵ cells). This result indicates that lack of IL-4 secretion following stimulation with PSMA peptides was not due to an IL-4 deficiency of CD4⁺ T cells but rather reflects the Th-1 bias of the cellular response induced by PSMA-VRP.

Tolerability in rabbits. A Good Laboratory Practice–compliant toxicology study was conducted to examine the potential toxicity and immunogenicity of PSMA-VRP in male rabbits. Following four s.c. injections of 2 × 10⁷ IU or 2 × 10⁸ IU PSMA-VRP on a biweekly schedule, all animals survived to their scheduled sacrifice. There was no evidence of vaccine-related systemic, hematopoietic, immune, or other organ toxicity at any dose or time in the study. Prostate, brain, kidney, and small intestine were unaffected. Physical examination of the injection site showed minimal erythema that resolved within 1 to 2 days and was not related to PSMA-VRP. Anti-PSMA antibodies were observed in zero of six vehicle animals, five of six animals in the low-dose PSMA-VRP group, and six of six animals in the high-dose group. Therefore, immunogenic doses of PSMA-VRP were well tolerated in this study.

The present study represents the first preclinical evaluation of a novel alphavirus vector vaccine for prostate cancer. PSMA-VRP mediated high-level expression of PSMA in mammalian cells in vitro. The antigen reached the cell surface, showed carboxypeptidase activity, and was recognized by a conformation-specific antibody. Potent, Th-1–biased antibody responses and CTL responses were observed in mice, and the epitope specificity of the cellular immune response was mapped to specific peptide sequences within PSMA. The product was immunogenic and well tolerated in a rabbit toxicology study. The findings support further evaluation of PSMA-VRP as a new immunotherapeutic approach to treat advanced prostate cancer and/or to prevent disease recurrence.

PSMA-VRP elicited robust anti-PSMA immune responses in each of two mouse strains and in rabbits. Therefore, the immunopotency of PSMA-VRP is not restricted to mice but rather is also seen in a larger, nonrodent species. A dose of 100 IU of PSMA-VRP was sufficient to elicit PSMA-specific B-cell and T-cell immunity in mice. The highest responses were observed at doses of 10⁶ IU, and no overt toxicity was observed over this dose range in mice. Similarly, doses ranging to 2 × 10⁸ IU were well tolerated in a rabbit toxicology study, and this information is relevant to the selection of doses for human testing.

The utility of viral-based vaccines often is limited by antivector immunity, and such vaccines typically are used as a booster injection following priming with DNA or other vaccine approaches (17, 32, 33). Consequently, there is limited potential for continued immune stimulation following the booster injection. In the present study, PSMA-VRP was used without a heterologous prime, and repeat dosing with PSMA-VRP led to increased immunity to PSMA. This finding is consistent with that observed for other VRP vaccines for cancer (16, 17) and highlights the potential for PSMA-VRP to serve as a stand-alone immunization strategy that potentially could be used to periodically restimulate immune responses over the lifetime of the patient. The ability to generate and restimulate long-lived responses without toxicity provides a rationale to explore PSMA-VRP vaccination as a means to prevent relapse of early-stage disease.

The role of CD8⁺ and CD4⁺ T lymphocytes in tumor killing has been shown in a large body of preclinical studies (34). Like other VRP vaccines based on Venezuelan equine encephalitis virus (16, 35–37), PSMA-VRP induced Th-1 responses, which are important for activating CTL. Immune splenocytes showed potent CTL activity that specifically lysed PSMA⁺ cells ex vivo. In addition to IFN-γ secretion by CD8⁺ and CD4⁺ T cells, the Th-1 bias was reflected in the preferential elicitation by PSMA-VRP of antibody isotypes (mouse IgG2a/IgG2b) that can efficiently mediate effector functions. Fc effector functions have been shown to mediate antitumor effects in vivo in preclinical models of other solid tumors (38), and the elicitation of coordinated cellular and humoral responses is a desirable feature of PSMA-VRP.

In head-to-head testing, PSMA-VRP was a more potent immunogen than an adjuvanted PSMA protein vaccine, which did not elicit a significant cellular response. In addition, whereas both PSMA-VRP and PSMA protein elicited robust antibody responses, PSMA-VRP elicited higher levels of antibodies that bound cell surface PSMA. A prior study examined an optimized PSMA DNA plasmid vaccine approach in BALB/c mice (39). In terms of IFN-γ responses, PSMA-VRP compares favorably with the DNA vaccine. In addition, PSMA-VRP elicited robust antibody and CTL responses that were not reported for the DNA vaccine (39).

Our study is the first to explore the epitope specificity of the CD8⁺ and CD4⁺ responses induced by an alphavirus vaccine. As determined in our assay system, the PSMA-VRP–induced cellular immune response was focused on a limited number of peptide determinants reactive with CD8⁺ and CD4⁺ T lymphocytes, although different peptides were dominant in BALB/c and C57BL/6 mice. The response in BALB/c mice was directed primarily toward a peptide (PSMA_76-90, NFTQIPHLAGTEQNF) that may contain both MHC class I and class II epitope(s). In contrast, the CD8⁺ and CD4⁺ responses in C57BL/6 animals mapped to separate epitopes: PSMA_631-645 (SLFSAVKNFTEIASK) and PSMA_231-245(PADYFAPGVKSYPDG), respectively. For both strains of mice, the peptides responsible for the CD8⁺ responses were confirmed to contain the respective K^b and K^d MHC class I binding motifs by the SYFPEITHI epitope prediction algorithm (40).⁴

Human PSMA_631-645 and PSMA_231-245 peptides are, respectively, 80% and 87% identical with the corresponding murine sequences, and these values bracket the 85% identity shared over the entire protein sequences of human and murine PSMA. PSMA_631-645 and PSMA_231-245 are both slightly more homologous (93%) to their murine counterparts than is the protein as a whole (89% homologous). Therefore, the responses observed against the PSMA_631-645 and PSMA_231-245 peptides are not due to any obvious degree of interspecies sequence divergence in these regions.

The antigenic determinants of the PSMA-specific IFN-γ responses were determined following stimulation with 15-mer peptide pools in an ELISPOT assay. The peptide responses were similar in magnitude to those observed following stimulation with syngeneic PSMA⁺ cells; however, it is possible that additional reactivities could be detected using other methods of stimulation and detection. Furthermore, because nonoverlapping antigenic specificities were observed in two in-bred strains of mice, additional diversity could be expected in out-bred animals and in humans.

The present report adds to a growing body of data on the preclinical potency of alphavirus-based vaccines for cancer (16, 17, 41). Our prior study with tyrosinase-VRP showed that immune tolerance can be surmounted using Venezuelan equine encephalitis virus VRP encoding either syngeneic or xenogeneic forms of the tumor antigen. In the prior study, higher levels of immunity and tumor protection were observed for VRP encoding mouse rather than human tyrosinase (17), consistent with other studies showing that alphavirus-based DNA and VRP vaccines can overcome immune tolerance to self-antigens and thereby obviate the need for xenoimmunization strategies (16, 41). The potency of alphavirus-based vaccines has been linked to activation of IFN responses and other innate immune mechanisms by double-stranded RNA intermediates formed during the replication cycle (41). In addition, the potency of VRP-based vaccines may reflect their ability to specifically target dendritic cells, Langerhans cells, and other antigen-presenting cells that are critical for priming the immune system (42–45). In the present study, we showed that PSMA-VRP transduction of dendritic cells leads to efficient presentation of PSMA antigen in vitro.

The pattern of PSMA expression is not conserved across species. For example, PSMA is not expressed in the mouse prostate or on available mouse prostate cancer cell lines (46), and no relevant PSMA tumor challenge model is available. In addition, no human PSMA transgenic animal model has been developed to examine issues of tolerance to human PSMA. For these reasons and because alphavirus-based vaccines obviate the need for xenoimmunization, as discussed above, we did not evaluate murine PSMA-VRP as a research tool in mice but rather chose to generate and evaluate human PSMA-VRP as a candidate vaccine for human prostate cancer therapy.

PSMA-VRP mediated high-level expression of functional PSMA and strongly elicited long-lived humoral and cellular immune responses in this study. The specificity of the cellular response in mice was mapped to specific peptide determinants using novel and sensitive methods, and these methods are appropriate for immunomonitoring of patients treated with PSMA vaccines. The vaccine was well tolerated in a Good Laboratory Practice toxicology study in male rabbits. We believe that PSMA-VRP provides an attractive new agent for the treatment of progressive prostate cancer and/or the prevention of disease relapse.