|Home | About | Journals | Submit | Contact Us | Français|
Prostate-specific membrane antigen (PSMA) is a prototypical differentiation antigen expressed on normal and neoplastic prostate epithelial cells, and on the neovasculature of many solid tumors. Monoclonal antibodies specific for PSMA are in development as therapeutic agents. Methodologies to actively immunize against PSMA may be limited by immunologic ignorance and/or tolerance that restrict the response to self-antigens. Our studies have previously shown that xenogeneic immunization with DNA vaccines encoding melanosomal differentiation antigens induces immunity in a mouse melanoma model. Here we apply this approach to PSMA to establish proof of principle in a mouse model. Immunization with xenogeneic human PSMA protein or DNA induced antibodies to both human and mouse PSMA in mice. Monoclonal antibodies specific for mouse PSMA were generated to analyze antibody isotypes and specificity for native and denatured PSMA at the clonal level. Most antibodies recognized denatured PSMA, but C57BL/6 mice immunized with xenogeneic PSMA DNA followed by a final boost with xenogeneic PSMA protein yielded autoantibodies that reacted with native folded mouse PSMA. Monoclonal antibodies were used to confirm the expression of PSMA protein in normal mouse kidney. These results establish the basis for clinical trials to test PSMA DNA vaccines in patients with solid tumors that either express PSMA directly or that depend on normal endothelial cells expressing PSMA for their continued growth.
PSMA (or folate hydrolase 1-FOLH1) is a 100–120 kDa type II transmembrane glycoprotein that is a glutamate-preferring carboxy-peptidase.1,2 PSMA is an attractive target for immunotherapy because it is highly expressed in normal and neoplastic prostatic tissue and the neovasculature of solid tumors, with limited expression in normal tissue.3–7
PSMA-specific monoclonal antibodies (MAb) have been used for in vivo diagnostic and therapeutic applications without adverse events.8–11 Biochemical and objective measurable disease responses were reported in some patients treated with radionuclide conjugates of a humanized MAb, J591, that binds to the extracellular domain of PSMA.12–18 The safety profile of passive immunotherapy with J591 and its derivatives, along with specific localization of radionuclide-labeled MAb to tumor sites in patients with metastatic prostate cancer or other solid tumors, supports the potential clinical benefit of antibodies and T cells specific for PSMA induced by active immunization. In addition to these studies, clinical trials of active immunotherapy have been completed with no evidence of toxicity,19–22 and some anecdotal declines in serum prostate-serum antigen (PSA) were reported.
Immunization against self-antigens such as PSMA may be achieved using plasmid DNA encoding the human orthologues of self-antigens,23–26 which in preclinical models resulted in tumor protection and, in some cases, rejection of established tumors.27–30 Outbred dogs with spontaneous melanoma immunized with xenogeneic tyrosinase DNA vaccines showed prolonged survival compared to historical controls.31 Preclinical studies in rats led to a clinical trial in which some patients immunized with rat prostatic acid phosphatase (PAP) responded to both rat and human PAP proteins.32
Using this paradigm, we have immunized mice with xenogeneic human PSMA (hPSMA), which shares 85% identity with mouse PSMA (mPSMA) at the amino acid level and has a similar tissue distribution.33,34 This pattern of expression supports the clinical relevance of a murine model that targets mPSMA.
We have previously demonstrated the induction of CD8 T cells specific for hPSMA but not mPSMA in mice immunized with hPSMA DNA vaccines.35 Here, we report that mice immunized with hPSMA protein and DNA vaccines also produce autoantibodies to both linear mPSMA epitopes and to naturally folded mPSMA epitopes present on the cell surface. These studies formed the basis for a clinical trial of PSMA DNA vaccines that are being tested in patients with solid tumors, which either express PSMA directly or depend on PSMA expressed by the neovasculature for their continued growth.
LNCaP, Sp2/0-Ag14 and NIH 3T3 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA). LNCaP cells were grown in RPMI supplemented with 7.5% fetal calf serum (FCS), 100 U/ml penicillin and 0.1 mg/ml streptomycin (P/S), L-glutamine and nonessential amino acids. SP2/0-Ag14 cells were grown in hybridoma serum-free medium (HyClone, Logan, UT) + 10% FCS. NIH 3T3 cells were grown in Dulbecco’s Modified Eagle’s Media, high glucose (DMEM HG) plus 10% Cosmic calf serum (HyClone, Logan, UT) and P/S. NIH 3T3 cells were transduced with recombinant retroviruses encoding human or mouse PSMA, and transduced cells were selected in 10 mg/ml puromycin. The parental retrovirus vector, SFG, has been previously described.36 Expression of PSMA in transduced cells was confirmed by immunostaining, using polyclonal and monoclonal antibodies specific for mouse and human PSMA.
Native human PSMA was purified from an LNCaP cell lysate by immunoprecipitation using CYT-356 (Cytogen Corporation, Princeton, NJ) and goat-anti-mouse IgG-agarose (Sigma Chemical Co., St. Louis, MO).4 The complex was washed 4 times with TNEN buffer (20 mM Tris, pH 7.5, 5 mM EDTA, 15 mM NaCl and 1% NP-40), 4 times with 0.1 × TNEN + 0.5 M NaCl, and once with water, after which it was resuspended in phosphate-buffered saline (PBS) and then injected intraperitoneally with Freund’s adjuvant. Mice were immunized 5 times at weekly intervals.
In other experiments, mice were immunized with DNA vaccines in which full-length human or mouse PSMA cDNAs were subcloned in the expression vector pcDNA3 (Invitrogen Life Technologies, Carlsbad, CA) or the closely related vector, pING.31 Plasmids were purified from bacterial lysates using Qiagen columns (Qiagen, Valencia, CA), and DNA was precipitated on gold, coated on plastic tubing and injected in the skin of mice, as described previously.26,37
An immunized mouse was rested for several months, boosted once with PSMA protein eluted from anti-mouse-agarose beads and 3 days later splenocytes were fused to the myeloma Sp2/0-Ag14 using a polyethyleneglycol-based fusion protocol modified by the Monoclonal Antibody Core Facility at MSKCC. Hybridoma supernatants were screened beginning 7 days after the fusion.
Recombinant hPSMA extracellular domain (residues 1–750) and mPSMA extracellular domain (residues 46–752) were overproduced in a baculovirus expression system using the Bac-N-Blue™ transfection and pBlueBacHis2 Xpress™ Kits (Invitrogen Life Technologies). Plaque purification, viral titers and expression of recombinant proteins were performed as described by the manufacturer. A third recombinant baculovirus expressing the extracellular domain of human tyrosinase in pBlueBacHis2A was the generous gift of Jason Fontenot and Philip Livingston (MSKCC). All recombinant proteins included an epitope tag for immunostaining (Tag) and a poly-HIS tag for affinity chromatography at the amino-terminus, which were contributed by the expression vector pBlueBacHis2A.
Recombinant proteins were expressed in High Five™ insect cells (Invitrogen, San Diego, CA) grown as a monolayer. Cells were infected at a multiplicity of infection of 5:1 and harvested at 62–72 hr. The recombinant proteins were almost entirely insoluble and were isolated from inclusion bodies in 8 M urea, 150 mM NaCl and 1 mM EDTA.
A soluble form of hPSMA was also prepared in CHO cells for the final immunization of the C57BL/6 mouse used in the second fusion. The hPSMA extracellular domain (ECD; aa 40–750) was inserted with a GST tag and a thrombin cleavage site into the pMelBacA vector (Invitrogen Life Technologies). The protein was purified from cell culture media by ammonium sulfate precipitation, glutathione 4B column binding, thrombin cleavage and size exclusion. Purity was assessed by SDS-PAGE (>99%). hPSMA (10 mg resuspended in 50 μl saline) was injected intravenously 3 days prior to the fusion.
Whole cell lysates were prepared by vortexing 107 cultured cells/ml lysis buffer [50 mM HEPES, pH 7.5, 15 mM MgCl2, 10% glycerol, 1 mM EGTA, 2 mM Na-Vanadate, 1% TritonX-100, 0.5 μg/ml leupeptin, 10 μg/ml aprotinin, 1 μg/ml pepstatin and 100 μg/ml phenylmethylsulfonyl fluoride (PMSF)]. Prior to use in immunoblotting or in the immunoprecipitation Western blot (IP-Western) assay, lysates were precleared for 10 min by centrifugation at 14,000 rpm in a microcentrifuge at 4°C.
For immunoblotting, whole cell lysate (10 μl) was separated by SDS-PAGE on 7.5 or 10% PAGE Gold Tris-glycine gels (Bio-Whittaker, Walkersville, MD), and transferred electrophoretically to Immobilon membranes (Millipore, Billerica, MA). Membranes were blocked in PBS + 5% nonfat dry milk and then incubated with mouse sera, hybridoma supernatant or with anti-Tag MAb (Invitrogen Life Technologies) for 1 hr at room temperature. After 4 washes with PBS + 0.1 % Tween20 (PBST), a combination of horseradish peroxidase (POD)-conjugated goat-anti-mouse IgG1, IgG2a and IgG2b (Southern Biotech, Birmingham, AL), secondary antibodies were added for 1 hr at room temperature. After 4 washes with PBST, POD-conjugated antibodies were detected by ECL™ (Amersham Biosciences, Piscataway, NJ) and autoradiography.
For immunoprecipitation, 50 μl LNCaP cell lysate was incubated with 50 μl hybridoma supernatant and incubated for 1 hr on ice. The immune complex was precipitated by addition of 10 μl packed volume of a goat-anti-mouse IgG-agarose bead (Sigma Chemical Co.), with a capacity of 30 μg mouse IgG. After 1 hr on ice, immune complexes were washed 4 times in cold PBS and eluted by boiling in SDS-PAGE sample buffer and the entire sample was loaded in 1 lane of a 7.5% Tris-glycine gel (BioWhittaker). After electrophoresis and transfer to Immobilon membranes, as described above, hPSMA was detected on the membrane with 6 μg/ml of a modified 7E11-C5 monoclonal antibody, CYT-356 (Cytogen Corporation, Princeton, NJ), followed by incubation with goat-anti-mouse IgG1-POD and developed with ECL™.
Corning ELISA plates were coated with 0.2 μg recombinant protein in 8 M urea, 150 mM NaCl for at least 12 hr, rinsed twice with PBS and blocked with PBS + 3% BSA for at least 1 hr. Plates were incubated with mouse sera or hybridoma supernatant and developed with a cocktail of isotype-specific goat-anti-mouse (anti-IgG3, anti-IgG1, anti-IgG2b and anti-IgG2a) alkaline phosphate-conjugated antibodies (Southern Biotech) and the substrate p-nitro-phenyl phosphate (Sigma Chemical Co.) in 10 mM diethanolamine (pH 9.5) and 0.5 mM MgCl2.
The 3T3 cells (105) were incubated with 1–5 μl sera or with 50 μl hybridoma supernatant for 1 hr on ice. Cells were washed with PBS + 1% bovine serum albumin (BSA), incubated 30 min with goat-anti mouse IgG-FITC (Southern Biotech) and after washing were read in a FACSCAN cytometer (Becton Dickinson, San Jose, CA). In some cases, isotype-specific secondary reagents were used in place of the goat anti IgG-FITC.
The peptides were chosen based on hydrophilicity and flexibility as predicted by Peptool software (Bio Tools, Inc., Edmonton, Canada) and by their location near the outer surface of the catalytic domain of mPSMA. The structure of the catalytic domain of hPSMA (residues 287–587) has been predicted based on the crystal structure of 2 bacterial dipeptidases from Streptomyces griseus and Vibrio proteolyticus.38 Peptides were synthesized as multiple antigenic peptides (MAP) and rabbits were immunized with 1 of 2 peptides, 434 STE (STEWAEEHSRLLQERS) or 483 QSP (QSPDEG-FEGKSLYDSWKEKSPSPE) (Research Genetics).
Tissues were fixed in 4% paraformaldehyde, embedded in paraffin and 8 mm thick sections were prepared. The immunohistochemical detection was performed using the MOM kit (Vector Laboratories, Burlingame, CA). Either rabbit sera at 1/4,000 dilution or mouse monoclonal antibodies (1 μg/ml) were used. Quenching of endogenous peroxidase was performed using H2O2, followed by citric buffer antigen retrieval for epitope unmasking. Blocking, incubation with the biotinylated secondary antibody and the detection steps were performed as recommended by Vector Laboratories.
Strategies for generating autoantibodies to mPSMA were investigated using 2 sources of antigen: 1) immune complexes containing hPSMA protein and 2) plasmid DNA encoding full-length mouse or human PSMA. Sera from immunized mice were tested for reactivity to human and mouse PSMA by ELISA, using denatured recombinant proteins produced in baculovirus. Because the recombinant protein used for the ELISA was isolated from inclusion bodies in 8 M urea and was therefore completely denatured on the assay plates, the binding antibodies most likely recognized linear determinants. This class of antibody may not bind native PSMA and would be expected to have limited utility in targeting tumor cells or tumor neovasculature. Antibodies to native folded PSMA on living cells were therefore assayed separately by flow cytometry using NIH 3T3 cells transduced with hPSMA, mPSMA or control (empty retroviral vector).
All mice immunized with an hPSMA DNA vaccine (10/10) or with immune complexes containing hPSMA protein (9/9) produced antibodies to both recombinant human and mouse PSMA proteins, as measured by ELISA (Fig. 1a and results not shown). There was no response to hPSMA or mPSMA by ELISA in mice immunized with either mPSMA DNA (9/9) or the control vector pcDNA (10/10). Western blots confirmed the specificity of the response to the hPSMA DNA vaccine (Fig. 1b). No response was elicited against an unrelated recombinant glycoprotein (human tyrosinase), which was produced in parallel in baculovirus. Immune responses detected by ELISA and evaluated by Western blot were always concordant.
As expected, all mice immunized with an immune complex containing hPSMA produced antibodies to native hPSMA. Both BALB/C and C57BL/6 mice produced antibodies to native mPSMA after immunization with an hPSMA DNA vaccine; however, the frequencies differed. In data combined from 2 separate experiments, sera from 1/15 BALB/C mice and 9/12 C57BL/6 mice immunized with hPSMA DNA reacted with native cell surface mPSMA (Fig. 2 and results not shown). The intensity of the staining for mPSMA was typically lower then that seen for hPSMA. This could either be due to a lower affinity of the antibodies or to lower expression of mPSMA compared to hPSMA in the transduced cells. In all cases, there was no staining of control NIH 3T3-vector cells. C57BL/6 mice immunized with an mPSMA DNA vaccine did not produce antibodies to human or mouse PSMA (results not shown). These results suggest that different genetic backgrounds can respond differently to xenogeneic DNA immunization against PSMA.
To dissect the fine specificity and isotypes of the antibody response to PSMA, we generated a panel of MAbs from a BALB/C mouse immunized with an immune complex containing xenogeneic hPSMA protein. An immunized mouse was rested for several months and then boosted once intraperitoneally with hPSMA immune complexes in saline buffer. Hybridoma supernatants were screened in 2 independent assays, by ELISA using recombinant proteins overproduced in a baculovirus expression system, and by flow cytometry using NIH 3T3-hPSMA cells.
Each supernatant was screened in triplicate wells coated with recombinant hPSMA, mPSMA, or human tyrosinase proteins. Supernatants that recognized either human or mouse PSMA, but not tyrosinase, were scored positive and were retested. In the primary screen, 47 of 1511 tested wells (3%) were confirmed positive (absorbance at least 5 times higher than that of human tyrosinase) and 42 clones were studied further (Table I).
A high proportion of stable MAbs identified by ELISA were cross-reactive with mPSMA (36/42, or 86%). There were no clones reactive to mPSMA alone. The specificity of representative clones was confirmed by Western blot analysis of lysates prepared from PSMA-positive LNCaP prostate carcinoma cells (results not shown), or from NIH 3T3-hPSMA or NIH 3T3-mPSMA (Fig. 3, Table I). These results eliminated the possibility that the determinants seen by these antibodies are an artifact of the insect cell expression system. After the initial analysis, 6 cross-reactive hybridomas were subcloned and studied further by immunohistochemistry. Results for clones 3E2 and 13D6 are detailed below.
Hybridoma supernatants were pooled in groups of 3 and screened by flow cytometry using NIH 3T3-hPSMA cells. Individual supernatants present in positive pools were retested. In the primary screen, 101 out of 970 (10.4%) supernatants tested were confirmed positive, showing obvious preferential staining of NIH 3T3-hPSMA cells relative to NIH 3T3-vector cells, and 59 hybridomas with the strongest and most persistent signal were retained. Each positive supernatant was then tested by flow cytometry for binding to NIH 3T3-mPSMA, NIH 3T3-vector and NIH 3T3-hPSMA cells. None of the hybridomas produced MAbs that recognized native mPSMA, while binding to native folded hPSMA was reconfirmed. In addition, hybridomas that recognized denatured mPSMA by ELISA did not bind to native mPSMA at the cell surface. Representative data are shown in Figure 4. Five hybridomas were chosen for subcloning and further characterization, and clones 5H12 and 7C12 are described in greater detail below.
Clones initially detected by either ELISA (denatured PSMA) or flow cytometry (native membrane hPSMA) were assayed by Western blot analysis, ELISA, flow cytometry and by immunoprecipitation of native hPSMA from LNCaP cell lysates (Table I). In our study, we identified 2 distinct classes of antibodies against PSMA: those reactive in the ELISA, which were predominantly cross-reactive and recognized denatured PSMA, and those reactive by flow cytometry, which were entirely hPSMA-specific and recognized the native folded protein.
The isotypes of individual hybridoma supernatants were determined either by ELISA or by flow cytometry (Table I). Of 24 cross-reactive clones, 17 (71%) were IgG1, 6 (25%) were IgG2b and 1 (4%) was IgG2a. Of 34 clones that recognized native hPSMA, 22 (65%) were IgG1, 11 (32%) were IgG2b and 1 (3%) was IgG2a. This pattern of isotype usage is consistent with a TH2-type response to both human and mouse PSMA.
Prior reports of human and mouse PSMA tissue distribution demonstrated expression primarily in the human prostate and in human and mouse kidney, brain and intestines. In our experience, lysates prepared from mouse kidney had the greatest PSMA enzymatic activity (NAALADase, results not shown). Normal mouse kidney sections were embedded in paraffin and stained with 2 independent rabbit anti-mPSMA peptide sera raised against peptides STE and QSP. Both sera strongly stained the luminal face of a subset of proximal tubules (Fig. 5a,b and results not shown).
A matching pattern of staining was seen in kidney sections stained with the cross-reactive MAb 3E2 (Fig. 5c). Three other cross-reactive MAbs tested, 13D6, 28D9 and 9C6, showed a diffuse cytoplasmic staining of all proximal tubules (Fig. 5d and results not shown). In contrast, clones specific for native hPSMA, such as 7E11, 5H12 and 7C12, did not stain proximal tubules in the mouse kidney (Fig. 5e and results not shown). In the case of 7E11, this is presumably due to differences in the primary amino acid sequence in the epitope recognized by 7E11, MWNLLH. The corresponding sequence in mPSMA is MWNALQ. For the other antibodies, the fine specificity is unknown and the lack of binding could either represent differences between the mPSMA and hPSMA primary sequences, or could result from a loss of conformational determinants of the native folded protein after fixation.
The molecular weight of mPSMA in the normal kidney was determined by Western blot analysis. Monoclonal antibodies 3E2 and 13D6, but not the human-specific MAbs 5H12 and 30A9, recognized a 100 kDa protein in a mouse kidney lysate (Fig. 6).
We were unable to identify monoclonal antibodies to native mPSMA following vaccination with xenogeneic hPSMA-antibody immune complexes. Therefore we turned to a prime-boost strategy using xenogeneic vaccination for priming followed by a final boost with xenogeneic hPSMA protein. Female BALB/C (n=15) and C57BL/6 (n=9) mice were immunized 5 times at weekly intervals with an hPSMA DNA vaccine. Sera were tested by flow cytometry, comparing the mean fluorescence intensity of control 3T3-vector cells versus 3T3-mPSMA cells stained with the same sera, under the same conditions. The best responder was identified as the animal with the highest binding to 3T3-mPSMA in the absence of binding to 3T3-vector cells. This C57BL/6 mouse was boosted intravenously with 10 μg purified recombinant hPSMA protein prior to fusion. Hybridoma supernatants were screened by flow cytometry using 3T3-mPSMA. Two positive pools, 9C1 (IgG2b) and 11C8 (IgG2b), were identified among 2,082 wells, which contained approximately 4,000 hybridomas. Hybridomas specific for mPSMA were cloned from each of the positive wells, and both of these pools were found to be specific for native mouse and human PSMA, with no recognition of 3T3-vector cells (Fig. 7).
To compare the number of wells specific for human and mouse PSMA, 284 representative wells from the fusion were also tested by flow cytometry for recognition of native hPSMA using 3T3-hPSMA. In this screen, we identified 3 positive samples out of 284 (1%). Normalized to the entire fusion, we estimate that 22 wells out of 2,082 recognized native hPSMA, and 2 of these (10%) cross-reacted with mPSMA.
Immunization of mice with xenogeneic hPSMA DNA or protein resulted in production of auto-antibodies to mPSMA. Mice did not respond to a mPSMA DNA vaccine, showing an inherent tolerance to mPSMA. All mPSMA-specific MAbs also recognized hPSMA, indicating that they arose from a response to epitopes in hPSMA. As the PSMA ECD was used to screen the fusion, the cross-reactive MAbs must recognize denatured linear determinants in the mouse and human PSMA ECD that are not found in naturally folded protein at the cell surface. In contrast, all MAbs specific for native epitopes in hPSMA failed to recognize native mPSMA, implying that B cells that recognized any immunodominant epitopes shared by native human and mouse PSMA have been deleted from the repertoire of BALB/C mice. Therefore, the remaining immunodominant epitopes were all restricted to hPSMA.
The predominant response in BALB/C mice was restricted to epitopes found only in fully denatured mPSMA protein, while C57BL/6 mice exhibited a broader response, including clones that recognize native mPSMA on the cell surface. This strain difference is most likely due to presentation of different class II-restricted peptides derived from hPSMA. While high affinity B cells specific for native mPSMA may be deleted or anergic in both strains, presentation of additional high avidity helper peptides in C57BL/6 mice may enhance B cell stimulation, driving the generation of cross-reactive B cells from low affinity immature cells via multiple rounds of somatic hypermutation.
Only 2 cross-reactive clones that recognize native mPSMA were obtained from a C57BL/6 mouse in the second fusion. This may be due to the relatively poor immunogenicity of DNA vaccines compared to immune complexes, to a poor overall response to native mPSMA in B6 mice, or to the TH1 bias of C57BL/6 mice. We estimated that 22 hPSMA-specific clones were present in the same fusion, and therefore we conclude that while the overall yield of hPSMA-specific MAbs in the C57BL/6 fusion was far below that of the first fusion from a BALB/C mouse, the percentage of antibodies to native mPSMA was higher. This is supported by our observation that a greater proportion of C57BL/6 mice immunized with hPSMA made antibodies to native mPSMA.
In one PhaseI/II trial, patients with advanced disease were immunized with PSMA peptides pulsed onto autologous antigen presenting cells,21,22 and minor T cell reactivity was observed. Other patients immunized with naked PSMA DNA vaccines alone or in combination with adenovirus encoding PSMA20,39 developed DTH responses to PSMA, while immune sera showed no reactivity to PSMA protein when tested by Western blot. Despite the lack of clinically meaningful immunity, these trials demonstrated that prostate cancer patients can be safely immunized with PSMA vaccines, and that limited T cell responses were elicited.
Our findings have led us to initiate a clinical trial using a xenogeneic DNA vaccine to break tolerance to PSMA. Our study shows the importance of analyzing patients’ sera with assays based on native protein, such as flow cytometry, and not Western blot analysis, using denatured proteins. The varying response to PSMA DNA vaccines in different mouse strains suggests the importance of MHC presentation of PSMA peptides in shaping antibody responses. In our clinical trials, the patient population will have multiple HLA alleles and therefore we expect to see a broad serological response to native hPSMA.
Protein for the final boost (second fusion) was provided by A. Wong and N. Pavletich (MSKCC).
Grant sponsor: CaPCURE; Grant sponsor: The Burke Foundation; Grant sponsor: Swim Across America; Grant sponsor: Abrons Foundation; Grant sponsor: NIH; Grant numbers: P50 92629, DK47650.