Prostate cancer is the most frequently diagnosed cancer and the second leading cause of cancer death in men in the United States. In spite of increasing attention and accumulating knowledge, advances in treatment that improve survival remain elusive. Prostate specific antigen (PSA) is the most widely used marker of prostate cancer. Immunoassays for PSA using monoclonal (MAb) or polyclonal antibodies have clinical applications, such as monitoring and early detection of prostate cancer. However, PSA is not a perfect tumor marker because serum levels often are elevated in men with benign prostatic hyperplasia, prostatitis and other nonmalignant disorders and also because levels are not always elevated in individuals with early prostate cancer.1
In an effort to identify additional proteins that could be of value for the diagnosis and treatment of prostate cancer, we have recently described the generation of a new prostate cancer-reactive MAb, designated 107-1A4, using a two-phase immunization protocol involving the prostate cancer cell line, LNCaP.2 This MAb recognizes an antigen, which appeared to be distinct from those previously described and shows specific tumor targeting in preliminary in vivo studies. In part because MAb 107-1A4 recognized a conformational epitope and, thus, could not be used in Western blots, we were unable to characterize its antigen using conventional approaches.2
In this report we describe the identification of the target antigen of MAb 107-1A4 using ProteinChip® array, surface-enhanced laser desorption/ionization (SELDI) technology from Ciphergen Biosystems (Fremont, CA). SELDI, a concept introduced by Hutchens and Yip3 combines ProteinChip technology with time-of-flight mass spectrometry and offers the advantages of speed, simplicity, sensitivity and accuracy. Briefly, each ProteinChip array has a number of spots that contain functional groups, chemical or biological “docking sites,” for the selective binding and washing of proteins/peptides from complex mixtures. After the sample is applied to the surface, unbound proteins and interfering substances are washed away. A solution containing laser energy-absorbing molecules, often referred to as a matrix, is then added and allowed to dry. In this fashion, the laser energy absorbing matrix molecules co-crystallize with the adsorbed proteins. The captured proteins are then detected using laser desorption ionization time-of-flight mass spectrometry (LDI-TOF MS). A more comprehensive description of the overall history and recent advances in SELDI technology and comparisons with other laser-based mass spectrometry techniques can be found in a recent review by Merchant and Weinberger.4
Herein we describe the application of SELDI to the peptide mass mapping of in-gel tryptic digests of the immunoaffinity-purified antigen. The resulting fragment data were used for a ProFound search of human protein sequences (http://prowl.rockefeller.edu/cgi-bin/ProFound). On the basis of the peptide masses derived from the SELDI spectrum, the protein is identified as prostate-specific membrane antigen (PSMA). PSMA is a type II membrane glycoprotein of Mr ∼100,000 that has a short intracellular N-terminal domain (residues 1 to 18), a transmembrane region (residues 19 to 43) and a large extracellular domain consisting of residues 44 to 750. PSMA was initially characterized by MAb 7E11, which binds to the cytoplasmic domain,5 and this antibody, like MAb 107-1A4, was derived after immunization of mice with preparations from the LNCaP human prostate cancer cell line. An alternatively spliced PSMA variant, termed PSM′, has also been described. PSM′ is cytoplasmic and consists of residues 60 to 750 based on the PSMA sequence; thus it lacks the intracellular domain, the transmembrane domain and a short segment of the extracellular domain.6 A number of other anti-PSMA MAbs have been developed recently and are typically categorized as being specific for the extracellular domain, if they bind intact target cells, or for the intracellular domain if they do not. Based on the results reported herein and on those of our previous study, MAb 107-1A4 is believed to recognize a distinct conformational epitope in the extracellular domain of PSMA.
There has been much interest in characterizing additional MAbs to PSMA in terms of their potential use in tumor targeting,7 particularly since the MAb that is currently approved for this indication (ProstaScint, Cytogen) binds the cytoplasmic domain of the molecule and is unable to target viable cells. In terms of in vitro diagnostic applications, the shortcomings of the PSA assay have led to considerable interest in additional markers for prostate cancer. Thus, much effort has been expended on configuring and evaluating immunoassays for PSMA.8 However, the results of these studies remain controversial. The use of SELDI to detect several potential prostate cancer markers including PSMA is of particular interest relative to our report.9, 10
MATERIAL AND METHODS
MAbs 107-1A4 and the control antibody, MOPC-21 were purified from ascites as described previously.2 Biotin-labeled MAb 107-1A4 and MOPC-21 were prepared with the Protein Biotinylation System kit (GIBCO BRL, Gaithersburg, MD). The following monoclonal antibodies, which are known to be specific for PSMA, were generously provided by Drs. Harry Rittenthouse and Roger Sokoloff of Hybritech (San Diego, CA): PM2J-004.5, PM2M-440, PM2M-474, PEQ-226.5 and PM2E-343.4.
Preparation of MAb 107-1A4-agarose and MOPC-21-agarose
One milligram of a solution of each MAb (1.5 mg/mL) was dialyzed at 4°C against 2 × 500 mL of 0.1 M NaHCO3 (pH 8.8) and coupled to 2 mL each of CNBr-activated agarose (Sigma, St. Louis, MO) according to the manufacturer's instructions. The coupling efficiency was 81% for MAb 107-1A4 and 83% for MOPC-21 based on the amount of unbound protein detected in their respective resin washes.
In a typical preparation a pellet containing approximately 100 × 106 LNCaP cells was produced as described previously and kept frozen at −20°C until use. The pellet was dispersed in 2.0 mL of cold 150 mM NaCl, 50 mM Tris-HCl (pH 7.4) containing 1% Nonidet P-40, 0.5% sodium deoxycholate and a mixture of protease inhibitors (Boehringher Mannheim- Complete). The cell lysate was prepared in the cold room on a rocker for 2hr. The cell lysate was microfuged at 10,000 rpm for 10 min in the cold, and the supernatant reserved for immunoabsorbtion.
Analytical scale immunoprecipitation.
For preliminary characterization of the antigen and Western blot studies (see below), the antigen was purified using the Protein G Immunoprecipitation Kit (Boehringer Mannheim, Indianapolis, IN) according to the manufacturer's instructions. Briefly, the crude extract was incubated with MAbs 107-1A4 or MOPC-21 at 1 μg/mL, and the antigen-antibody complexes were bound to separate aliquots of immobilized protein G. After washing, bound proteins were eluted from the resins with SDS-PAGE sample buffer by boiling for 5 min and then separated by SDS-PAGE using a 4% to 20% gradient gel (BioRad, Hercules, CA). Protein was visualized with SYPRO Orange, and reactivity with MAbs PM2J004.5 or MOPC-21 was detected by Western blots as described below.
Preparative scale immunoabsorption.
The LNCaP extract was mixed with 0.25 mL of MOPC 21-agarose and rocked gently for 2 hr at 4°C. The resin suspension was microfuged as above for 30 sec. The resin pellet was washed exhaustively with extraction buffer and then with PBS. The resin pellet was dispersed in 0.2 mL PBS containing SDS sample buffer (BioRad) with reducing agent and then boiled for 5 min. The fraction of the extract that did not bind immobilized MOPC-21, was mixed with 0.25 mL of MAb 107-1A4-agarose and rocked gently in the cold for 16 hr. The resin was washed and then treated with SDS-sample buffer as described above. Aliquots of the MOPC-21 agarose-bound and the MAb 107-1A4- agarose-bound fractions were subjected to SDS-PAGE on 4% to 20% gradient gels (BioRad). Protein bands were detected by staining with Coomassie blue under standard conditions. The dominant band observed in the MAb 107-1A4-bound fraction was excised with a razor blade. To control for peptides generated from background proteins and trypsin autolysis, a zone of approximately the same size as the stained band was prepared from an unstained region of the lane containing MAb 107-1A4-bound specimen.
In-gel digestion of 107-1A4 antigen and analysis by SELDI
In-gel tryptic digestions of the excised 107-1A4 antigen band and the control zone were based on the procedure used in the UCSF Mass Spectrometry Facility (http://donateno.ucsf.edu/MSF/ingel.html). Each gel slice was placed in a 0.5 mL microfuge tube that had several holes punched in the bottom with a 24 G needle. Minced gel pieces were obtained by subsequent centrifugation at 14,000 rpm and collected into a 1.5 mL microcentrifuge tube. The resulting gel pieces were destained and dehydrated with two washes (each on the agitator for 10 min) using enough 25 mM ammonium bicarbonate, 50% acetonitrile (v/v) to immerse the gel particles fully. The destained gel particles were dried in a Speed-vac for approximately 30 min and then rehydrated with 5 μL of 25 mM ammonium bicarbonate containing 0.1 μg/μL sequencing grade trypsin (Promega, Madison, WI). Digestion was allowed to proceed overnight at 37°C.
Digestion solution (1 to 5 μL) was applied to a normal phase SELDI ProteinChip array (Ciphergen Biosystems). Since the digest was performed using MALDI-type conditions, the normal washing step was bypassed. Under other conditions in which buffers not compatible with subsequent mass analysis (e.g., urea) are present, the ProteinChip array is washed to remove interfering substances. After drying, α-cyano-hydroxy cinnamic acid (1.0 mg/mL) in 50% acetonitrile (v/v), 0.5% trifluoroacetic acid (v/v) was addedm and the ProteinChip array was analyzed by time-of-flight mass spectrometry in a Ciphergen Protein Biology System II (PBS II). The spectra contained an average of 84 laser shots collected in positive ion mode. Angiotensin I [Glu 1] fibrinopeptide B, endothelin, adrenocorticotropic hormone (ACTH; 1 to 24), and β-endorphin (61 to 91; Ciphergen Biosystems) were used for internal calibration.
LNCaP cell lysates were generated as described above. Immunoprecipitations using MAbs 107-1A4, MOPC-21, PM2M440 and PM2J004.5 were performed.. Samples of immunoaffinity-purified antigen were subjected to SDS-PAGE under both reducing and non-reducing conditions as described above and transferred to nitrocellulose membranes.11 The membranes were blocked overnight at 4°C with 2% nonfat dry milk, washed with TTBS (0.05% Tween 20 in 0.15 M NaCl, 0.02 M Tris-HCl, pH 8.0) and probed with 2 μg/mL solutions of PM2J-004.5 or MOPC-2 in 2% BSA/PBS. After washing with TTBS, the membrane-bound MAbs were detected with goat anti-mouse IgG (H+L)-HRP (1 μg/mL) in 2% BSA/PBS and visualized using an enhanced chemiluminescence kit (Amersham, Arlington Heights, IL).
Microtiter wells were coated with LNCaP cells and fixed with formalin as described previously.2 Solutions (100 μL/well) containing biotinylated MAb 107-1A4 or biotinylated MOPC-21 at 2 μg/mL with and without competing, unlabeled MAbs at 50 μg/mL in 2% BSA/PBS were added to microtiter wells, which had been coated with LNCaP cells.2 After incubation for 1 hr at 37°C, the plates were washed with TTBS, and 50 μL/well of ABC-alkaline phosphatase (Vectastain; Vector, Burlingame, CA), prepared according to the manufacturer's instructions, was added. After a 15 min incubation at room temperature, the plates were washed, and bound conjugate was detected with para-nitrophenylphosphate. The reaction was stopped by the addition of 50 μL/well of 1 M NaOH, and the absorbance was measured on a microplate reader at 410 nm.
Molecular weight of MAb 107-1A4 antigen
Small amounts of antigen were prepared by immunoprecipitation from LNCaP extracts by incubation with free MAb 107-1A4 followed by Protein G-agarose to bind the complex and any remaining free antibody. The results of SDS-PAGE under reducing conditions on bound fractions derived from extracts of LNCaP cells are shown in Figure 1. Lane 1 contained the Mark 12 protein ladder (Novex, San Diego, CA). Bands corresponding to molecular weights of approximately 100, 53 and 28 kDa were observed in lane 2, which contained proteins eluted from MAb 107-1A4-Protein G immunoprecipitate. Lane 3 ( MAb 107-1A4 alone) and lane 4 (immunoprecipitate with MOPC-21) had only the 53 and 28 kDa bands, which are presumed to be the antibody heavy and light chains, respectively. Thus, the antigen recognized by MAb 107-1A4 appeared to have a molecular weight of approximately 100 kDa. In data not shown, this band was not visualized by Western blots with MAb 107-1A4 as probe, confirming the earlier report that the MAb 107-1A4 epitope is conformational.
Identification of the MAb 107-1A4 antigen as PSMA
Monoclonal antibodies, which had been coupled to agarose, were used to prepare antigen for tryptic digestion and subsequent SELDI analysis. LNCaP cell extracts were first absorbed with MOPC-21-agarose to remove non-specifically bound proteins, and this was followed by treatment with MAb 107-1A4-agarose to extract the target antigen. In data not shown, only a few weak bands were observed when the control lanes containing the MOPC 21-bound fraction were stained with Coomassie blue. In contrast, the lanes containing the MAb 107-1A4-bound fraction exhibited the dominant band corresponding to a molecular weight of approximately 100 kDa as was seen in the analytical preparations (Fig. 1), in addition to the weakly staining bands observed in the MOPC 21-bound fraction. This 100 kDa band and a control region lacking protein as determined by Coomassie staining were excised, digested with trypsin and analyzed as described below.
The 100 kDa protein was enzymatically cleaved with trypsin by in-gel proteolysis. The resulting peptides were applied to a normal-phase SELDI ProteinChip array and analyzed by LDI-TOF MS using a Ciphergen Biosystems PBS II (Fig. 2a). A protein-free gel digest was run in parallel to identify signals associated with autolysis of the trypsin and possible proteolytic fragments of contaminant proteins, such as human cytoskeletal proteins (Fig. 2b). A total of 19 unique peptides (denoted by asterisks in Fig. 2a) were used for a ProFound search of human protein sequences between 0 and 150 kDa (http://prowl.rockefeller.edu/cgi-bin/ ProFound). Two separate searches were performed to allow for potential modification of cysteine residues by acrylamide. In both cases, the two closest matches of the top ten candidates identified the 107-1A4 antigen as PSMA. The probability of the match to PSMA was 0.5, with the next closest match showing a probability that was several orders of magnitude lower. Thirteen of the 19 proteolytic fragments used for the database search were matched to the sequence of PSMA with a mass error less than 0.072% and an average mass accuracy of 0.029% ± 0.021 (Fig. 2, Table I). The observed mass assignment error is consistent with the expected performance of this analytical system. A coverage map of PSMA is shown in Figure 3. The 13 fragments matched to PSMA cover 27% of the primary sequence of the protein.
Table I. Residues of PSMA Identified Following In-Gel Tryptic Digestion
Observed mass in daltons after internal calibration as described in materials and methods.
Average chemical masses, in daltons, for tryptic fragments.
Difference, in daltons, between observed and calculated masses.
Percent mass error.
Six fragments unique to the 107 antigen band did not match the primary sequence of PSMA. The potential identity of these fragments is discussed in the text. The peak in the mass spectrum at 946.90 Da may be derived from residues 208–215, 565–571 or both regions of PSMA as determined within the mass accuracy. Similarly, the peak at 1184.03 Da may represent residues 182–190, 547–545 or both.
Confirmation of PSMA as the 107-1A4 antigen by Western blot analysis
Since the structural analysis of the MAb 107-1A4 antigen by SELDI strongly suggested that the 100 kDa protein was, in fact, PSMA, Western blots were performed using an antibody known to be specific for the intracellular domain of PSMA (PM2J004.5) as probe (Fig. 4). The crude LNCaP extract (lane 4) and the immunoprecipitate formed with Mb 107-1A4 (lane 6) revealed equivalent bands at approximately 100 kDa, as well as lower molecular weight components. The same general pattern was seen when immunoprecipitates formed using MAb PM2J004.5 (lane 3) and anti-PSMA MAb, PM2M440 (lane 2), which is specific for the extracellular domain of PSMA, were probed with PM2J004.5. In contrast, the 100 kDa band was not seen on the LNCaP extracts, which had been immunoprecipitated with MOPC-21 (lane 1) or in the MAb 107-1A4 immunoprecipitate probed with the control MAb, MOPC-21 (lane 7). The dark bands at 50 to 55 kDa in all samples subjected to immunoprecipitation (lanes 1, 2, 3, 6 and 7) and in the MAb 107-1A4 control (lane 5) represent the mouse monoclonal IgG heavy chain.
Other anti-PSMA MAbs do not compete with 107-1A4
As seen previously,2 biotinylated MAb 107-1A4 specifically bound to formalin-fixed LNCaP, since the signal observed with streptavidin alkaline phosphatase as conjugate was significantly higher than either that of control wells not exposed to biotinylated MAb or that in wells incubated with biotinylated MOPC-21 (Table II). Incubation of LNCaP cells with biotinylated-MAb 107-1A4 at 2 μg/mL in the presence of unlabeled Mab 107-1A4 at 50 g/mL yielded a signal that was indistinguishable from that of the control wells. None of the five anti-PSMA MAbs, which were provided by Hybritech, gave signals that were significantly lower than that resulting from incubation with biotinylated MAb 107-1A4 alone.
Table II. Competition of the Binding of Biotinylated Mab 107-1A4 to LNCaP Cells by Various Monoclonal Antibodies1
Competition of the binding of biotinylated MAb 107-1A4 to LNCaP cells by various monoclonal antibodies. Biotinylated 107-1A4 (B107) or biotinylated MOPC-21 (BMPOC) at 2 μg/mL was incubated in microtiter wells containing LNCaP cells either alone or in the presence of 50 μg/mL of the antibodies cited. The binding of biotinylated MAb was detected with an avidin-biotin complex kit as described in the text. N, number of replicates; A, mean A410 nm; SD, standard deviation.
Blk (no B107)
B107 + PM2E-343.4
B107 + 107
B107 + PEQ-226.5
B107 + PM2J-004.5
B107 + PM2M-440
B107 + PM2M-474
Previously, we reported that 107-1A4 restrictively reacts with prostate epithelial membranes. We concluded that the target antigen was distinct from PSA, PAP and PSP-94, since it failed to react with these purified proteins.2 By comparing the reactivity of 107-1A4 with those of other MAbs on tissues and cell lines, we eliminated several other known antigens as potential targets. The distribution of the 107-1A4 antigen in prostate tissue was similar to that reported for PSMA by immunohistochemical staining; however, there was an apparent difference in the reported regulation of PSMA by androgens and that we observed for the 107-1A4 antigen, in both cell lines and xenografts. Unfortunately, hormone regulation studies are difficult to perform and at that time, we did not have access to an anti-PSMA MAb to serve as positive control. Thus, our original conclusion that PSMA was not the antigen recognized by 107-1A4 was incorrect.
Clearly, additional approaches to determine the identity of the 107-1A4 antigen were needed. Although several strategies exist for protein identification, they each have their limitations. Microsequencing analysis, for example, can be a laborious process that requires relatively large quantities of protein. In addition, most mammalian proteins, including PSMA, have blocked N-termini and are thus not appropriate candidates for direct sequencing. Moreover, without fragmentation of the molecule, only the first 20 or so amino acids are likely to be identified in a protein the size of PSMA.
A second approach for identifying an unknown protein is to screen an expression library made from cells that highly express the antigen. There are at least three shortcomings of this route. First, the expressed protein from the corresponding cDNA library may not fold properly and thus, an antibody like 107-1A4 that recognizes a conformational epitope will not bind to the antigen-expressing clone. Second, if the antigen is large and the antibody recognizes an epitope in the amino-terminal region of the antigen, the chance of getting a clone containing the epitope is relatively low. Finally, if the mRNA of the antigen is of low abundance, it may not be properly represented in the limited number of clones available for screening.
More recently, methods utilizing mass spectrometry have proved to be powerful approaches for protein identification and complement well the techniques described above.12 In this report we demonstrate the application of SELDI technology for peptide mass mapping followed by database searching to identify the 107-1A4 antigen. On the basis of the peptide mass fingerprint derived from the SELDI spectrum, the protein was identified as PSMA. Thirteen of 19 peptides matched to PSMA covering 27% of the primary sequence (Fig. 3).
Although the structural analysis presented here is quite compelling, we observed both unidentified fragments and incomplete coverage of the PSMA molecule (i.e., undersampling) by SELDI. These observations are typical of this type of technology and were expected. There may be several explanations for the presence of unidentified fragments. For example, these peptides may correlate to digest products containing post-translational modifications. Consistent with this idea, examination of the primary sequence of PSMA reveals 10 potential N-linked glycosylation sites, and previous studies have demonstrated that as much as 20% to 25% of the mass of PSMA comes from carbohydrate.13 All of the consensus sequences for N-glycosylation reside outside the regions of PSMA covered by the identified digest products. Additional studies could be performed using N-glycanase to deglysoylate the protein and monitor whether the masses of the unmatched peptides subsequently shift to match the predicted mass of other tryptic digest products. It should be noted that unambiguous identification of these fragments would require partial sequencing or tandem mass spectrometric (MS/MS) analysis.
The undersampling of the primary structure may also be explained in part by the inaccessibility of trypsin to certain cleavage sites as a result of post-translational modifications to the protein. As discussed above, a significant portion of the mass of PSMA comes from carbohydrate and 10 potential N-linked glycosylation sites reside within the regions of PSMA not covered by the digest products.13 In addition, several other factors have been reported to contribute to incomplete mass maps including the amount of protein available as well as the staining procedure.14 Moreover, analyte ion yields can be significantly affected by a variety of external factors including matrix solution conditions, matrix crystal morphology, the intrinsic properties of the peptides and the presence of peptide mixtures (in this case trypsin autolysis products).15–18 A more rigorous study aimed at addressing these issues could be employed in an effort to improve the coverage map obtained.
Since the mass spectrometric strategy tentatively identified the 107-1A4 antigen protein as PSMA, we have confirmed this assignment by Western blots using well-characterized anti-PSMA Mabs on antigen that had been immunoprecipitated with 107-1A4. Although the 107-1A4 antigen was clearly shown to be PSMA by SELDI and Western blot studies, the other anti-PSMA MAbs available to us did not compete with biotinylated 107-1A4 for binding to LNCaP cells, suggesting that the 107-1A4 epitope may be unique.
In this report we have demonstrated that PSMA is the target antigen for MAb 107-1A4 using SELDI peptide mass mapping of in-gel tryptic digests of immunoaffinity-purified antigen and confirmed this identification using MAbs of known specificity. The example presented here clearly demonstrates the effectiveness of SELDI for rapid protein identification. The utility of this technology has also been demonstrated for a number of other applications including the analysis of post-translational modifications, protein-protein interactions and immunoaffinity assays. For example, Cardone et al.19 have used SELDI to analyze the phosphorylation of caspase 9 (Casp9) by the kinase Akt. Initial SELDI experiments analyzing V8 proteolytic digests of recombinant caspase 9 reveal one peptide that was shifted in mass by 80 daltons, an observation consistent with the presence of a single Akt phosphorylation site on Casp9.
More recently, Hinshelwood et al.20 have described the use of SELDI affinity mass spectrometry to study the interaction between a recombinant von Willebrand factor type A domain (vWF-A) in complement factor B with C3. Briefly, C3 is covalently attached to the surface of a ProteinChip array and used to specifically capture the vWF-A domain. The complex is then digested with trypsin and washed to remove unbound fragments. Mass analysis of the remaining fragments identified two peptides involved in the binding of the vWF-A domain to C3.
Finally, a similar approach has been taken in experiments aimed at investigating the processing of the amyloid precursor protein in Alzheimer's disease. In this example, a polyclonal antibody is covalently immobilized on the ProteinChip surface and used to capture and purify multiple immunoreactive amyloid beta fragments from small volumes of unfractionated biological samples.21 The assay has been utilized by investigators using cell culture models and brain tissue homogenates to study the mechanisms of amyloid beta peptide generation.22–24
We suggest that SELDI and related mass spectrometric approaches will become commonly used to complement conventional protein techniques as the need to identify low amounts of proteins rapidly increases.
The authors thank Mr. Scott Weinberger of Ciphergen Biosystems for his assistance in reviewing this manuscript. The authors also thank Ms. Lisha Brown for her help in generating the 107-1A4 antibody.