A novel multiplex assay combining autoantibodies plus PSA has potential implications for classification of prostate cancer from non-malignant cases

The study focused on 6 PCAA, namely NY-ESO-1, SSX-2,4, XAGE-1b, AMACR, p90, and LEDGF. All had been reported by multiple groups with data on gene expression and autoAb presence in prostate cancer patients. As previously described [18], prediction and screening of peptide epitopes was conducted using classic ELISA. Peptides were considered positive based on recognition by serum samples from prostate cancer patients (n > 50) but not healthy donors (n > 20). Then, peptide-reacting serum samples were verified for recognition of the full-length or a truncated recombinant protein using Western blot. Only after such a procedure, a validated peptide was conjugated onto seroMAP microbeads for multiplex measurement. All peptides involved in this study were synthesized at Genscript Inc. (Piscataway, NJ) and GeneMedicine, Inc. (San Antonio, TX). Historical serum samples from cancer and HD as described previously [18] were used for identification of peptide epitopes, which were independent of those used in the subsequent study comparing A+PSA index with PSA. In the case of identifying peptide epitopes from shared cancer/testis antigen XAGE-1b and SSX2,4, serum samples from NSCLC were used. This choice was made based on higher frequency of seropositive subjects in NSCLC and the fact that peptides from shared antigens identified using one type of cancer patients can be equally well recognized by prostate cancer patients [18].

All serum samples were collected under institutional review board-approved protocols from UCLA (IRB#06-03-044) and collaborating hospitals, and stored at -20°C until use. Serum samples from normal healthy subjects were collected at the time of blood donation in subjects routinely screened to exclude the presence of concomitant disease such as cancer according to standard blood bank policies. Serum samples from biopsy-confirmed prostate cancer patients were collected at the time of biopsy and prior to surgery. Patients with BPH and/or prostatitis, specified as non-cancer or BPH/prostatitis patients throughout this manuscript, were those with clinical signs and symptoms, for instance, characteristic lower urinary tract symptoms, International Prostate Symptom Scores, urinary leukocytes, and so on. These patients were subsequently underwent a routine fine needle prostate biopsy with at least 6-12 samples taken showing no evidence of prostate cancer. Table 1 shows the demographic and clinical characteristics of the subjects involved in the comparison of A+PSA with PSA alone. Additional file 1 illustrates the distribution of their total PSA values, which were measured using a standard ELISA approach according to the manufacturer's recommendations at the time of diagnosis.

Since this was a pilot study, cohort size and relevant parameters such as age, racial and ethnical background were not sufficient to match samples according to potential clinical co-founders. However, all samples themselves were handled and stored according to the same conditions prior to assay; and normalization with samples from HD was conducted when needed in order to minimize experiment-to-experiment variations.

Conjugation of peptide epitopes defined in this study onto seroMAP beads was conducted according to the manufacturer's recommendations (Luminex Corporation, Austin, TX). In the final configuration of the A+PSA assay, seroMAP microbeads region 001 were conjugated with the NY-ESO-1 peptide epitope as previously reported [18], region 010 with the XAGE-1b epitope (amino acid 1-25), region 020 with the SSX2,4 epitope (amino acid 110-139), region 030 with the AMACR epitope (amino acid 251-281), region 040 with the p90 autoantigen epitope (amino acid 796-827), region 050 with a control peptide from β-galactosidase, and region 060 with the LEDGF epitope (amino acid 448-468). A 96-well filter bottom plate (Millipore, Billerica, MA) was pre-washed followed by addition of blocking buffer and incubation for 1 hour at room temperature. About 50 μl of serum samples pre-diluted at 1 to 10, 1 to 20, and 1 to 50 were mixed with an equal volume of the above-conjugated seroMAP microbeads at 5000 beads/region, and were added to each well. After one hour of incubation, plates were washed 3 times, followed by addition of 100 μl PE-labeled detection Ab (Ab against human IgG and Ab against human total PSA) to each well. After 30 min, plates were washed 3 times, and added 100 μl blocking buffer into each well. The plate was read by Bioplex-200 (Bio-Rad Laboratories, Hercules, CA) to obtain the mean florescent intensity (MFI) for each seroMAP region.

In addition to measuring autoAb, seroMAP microbeads region 100 were conjugated with a monoclonal Ab against human PSA (Biocon, Inc. Rockville, MD) to quantify total PSA levels. seroMAP-based PSA quantification was compared with standard ELISA-based PSA assays (American Qualex) and also made compatible with the measurement of the above-mentioned 6 autoAb to constitute the A+PSA assay.

AutoAb present in patients' serum samples were previously measured using a standardized ELISA approach [18]. In brief, 1 μg of a synthetic peptide was diluted in 5 ml phosphate buffered saline (PBS) and adsorbed onto a 96-well MaxiSorp plate (Nunc, Denmark) overnight at room temperature. Control plates were coated with bovine serum albumin (BSA) at 15 μg/plate or about 150 ng/well. Plates were blocked with 5% Fetal Bovine Serum in PBST (PBS plus 0.05% Tween-20) for at least 2 hours, washed with PBST, and loaded with 100 μl of serum samples diluted at 1:25, 1:125, and 1:625 with PBST containing 5% Fetal Bovine Serum. After a 2-hour incubation at room temperature, plates were washed, and loaded with secondary antibodies (goat anti-human immunoglobulin conjugated with horseradish peroxidase, Sigma Co., St. Louis, MO) diluted with 5% Fetal Bovine Serum in PBST. Plates were developed after a one-hour incubation, and absorbance at 450 nm was read by using an ELISA reader. Signal to noise ratio for ELISA-based approaches was defined as the OD against a target epitope/average OD from at least 8 HD. Signal to noise ratio for seroMAP-based approaches was defined as the specific MFI ratio against a target peptide/average specific MFI ratio against the same peptide from at least 8 HD, where the specific MFI ratio is defined as the MFI against a PCAA peptide/MFI against a control peptide.

To better predict prostate cancer, it is necessary to create an index integrating both autoAb against the 6 above-described PCAA and the patient's PSA status. For total PSA and autoAb against each peptide epitope, an index value was calculated based on the mean MFI ratio, which is defined as the florescent intensity against a specific peptide/florescent intensity against a control peptide. The Kolmogorov-Smirnov test was used in the 6 autoAb markers to determine if the histograms between prostate cancer and BPH and/or prostatitis differ significantly. The A+PSA index was defined as the probability of being prostate cancer, which was obtained by combining the six above referenced epitope indices with the PSA index using the logistic regression method

where each Ni represented the average MFI values for an autoAb from three dilutions, N _PSA was the average MFI for PSA from three dilutions, and a0,...a7 were estimated regression coefficients of the logistic regression model. In the logistic regression model, the binary dependent variable is 1 for a patient with prostate cancer and 0 for a patient with nonmalignant conditions, for example, BPH and/or prostatitis. The receiver operating characteristic (ROC) curve was used to compare the diagnostic power between PSA alone and the combined A+PSA index for distinguishing prostate cancer from BPH and prostatitis in all subjects and subjects with 4-10 ng/ml PSA. Area Under the Curve (AUC) from Receiver operating characteristic (ROC) analysis was calculated from the logistic regression model. To avoid a potential overfitting issue in modeling and the testing within the same data set, the bootstrap method [19] was applied to construct 95% confidence intervals for the AUCs and test their difference. Values of P < 0.05 were considered statistically significant.

Similar to NY-ESO-1, XAGE-1b and SSX-2,4 are cancer/testis antigens shared among cancers of the prostate, lung, breast and others [20, 21]. To identify dominant B cell epitopes from XAGE-1b, computer-aided algorithms were applied to predict the peptide epitopes [18]. Two candidate peptides were screened by ELISA (Figure 1A) with serum samples from cancer patients. Three of 48 cancer patients were tested positive reacting with XAGE-1b peptides based on previously described criterion [18]. Two of the 3 seropositive patients reacted only with XAGE:1-25 peptide; while the other reacted with both XAGE:1-25 and XAGE:57-81 peptides. Western blot confirmed that sera recognizing the XAGE:1-25 peptide reacted with the full-length XAGE-1b protein from a transfected 293 cell line (Figure 1B).

Similarly, a SSX-2,4 peptide epitope was identified and confirmed with Western blot that the serum reacting with SSX-2,4:110-139 was able to recognize the full-length recombinant protein (Additional file 2). In addition, candidate peptides from AMACR, p90 autoantigen, and LEDGF were screened using serum samples from prostate cancer patients and control samples from HD (data not shown). Verification of the AMACR and LEDGF peptide epitopes by Western blot is also shown in Additional file 2.

Following the identification and confirmation of peptide epitopes from the above-mentioned PCAA, each peptide was conjugated onto seroMAP microbeads with a specific region number (Materials and Methods). The ease of conjugating peptides over purified recombinant proteins onto seroMAP microspheres allowed multiplex detection of autoAb against the above-described peptide epitopes from XAGE-1b, SSX2,4, AMACR, p90 autoantigen, LEDGF, and NY-ESO-1 [18]. Specific MFI ratios, defined as the ratio of the MFI against a target peptide to the MFI against a control peptide, were compared with those from at least 8 HD, which was defined as the relevant signal-to-noise ratios for seroMAP-based approaches. Similarly, signal-to-noise ratios for ELISA-based approaches were determined (Materials and Methods). The seroMAP-based approach showed significantly improved signal-to-noise ratios over ELISA-based approach in measuring autoAb against a prototype NY-ESO-1:1-40 epitope among 4 randomly selected seropositive prostate cancer patients with 8 HD as controls (Figure 2A). Similarly, improved signal-to-noise ratios against the XAGE-1b epitope were observed using the seroMAP-based approach over ELISA.

To develop a multiplex assay that measures total PSA and autoAb in a single reaction, conventional PSA tests were first converted from ELISA to seroMAP-based approaches. PE-conjugated secondary Ab against human IgG and PSA were mixed in the multiplex assay to accommodate staining of autoAb and PSA binding to distinct seroMAP regions, allowing simultaneous quantification of autoAb plus PSA in one reaction (termed the A+PSA assay).

To ensure that the multiplex A+PSA assay did not interfere with the quantification of individual autoAb, autoAb against the prototype NY-ESO-1:1-40 epitope using the multiplex A+PSA assay were compared with those measured using seroMAP-based singular assays. It was found that autoAb against NY-ESO-1:1-40 measured by these two assays correlated markedly well among 40 randomly selected subjects (correlation coefficient was 0.98, Figure 2B). Similarly, purified PSA standards (n = 4) determined by the seroMAP-based A+PSA assay produced a trendline with a correlation coefficient of 0.98 with that obtained from a commercial ELISA kit (data not shown). For clinical samples, the correlation coefficient of PSA values obtained by ELISA (Figure 2C, x-axis) and the seroMAP-based A+PSA multiplex assay (y-axis) was 0.89 over a wide dynamic range from 0.1 to 60 ng/ml in 376 randomly selected subjects. Thus, the A+PSA assay format did not appear to produce interference by quantifying autoAb and PSA simultaneously in one reaction. In other words, the A+PSA assay is as specific as measuring individual autoAb and total PSA separately while providing the simplicity and cost-effectiveness of a multiplex assay that requires less sample and handling time of quantifying 6 or more autoAb and PSA simultaneously.

Pre-surgery serum samples from biopsy-confirmed prostate cancer patients (n = 131), BPH/prostatitis patients (n = 121) and healthy donors (n = 124), which were independent of the samples used in the epitope discovery phase, were subjected to determining total PSA and autoAb against the 6 defined PCAA epitopes. Histograms of the density or frequency against all 6 PCAA are depicted in Figure 3. Patients with non-malignant conditions had a narrower distribution of specific MFI ratios; meanwhile prostate cancer patients exhibited a much broader range of specific MFI ratios from 1 to nearly 300 for autoAb against NY-ESO-1. Histograms of other PCAA are shown in Figure 3B-F. The Kolmogorov-Smirnov tests for each of the 6 autoAb between prostate cancer and BPH/prostatitis groups were performed resulting in highly statistically significant differences between the two groups except for autoAb against AMACR. All four p-values of LEDGF, p90 autoantigen, SSX-2, 4 and XAGE-1b were less than 0.001 and the p-values of NY-ESO-1 and AMACR autoAb were 0.029 and 0.134, respectively.

A combined A+PSA index was created as the predicted probability of prostate cancer based on a logistic regression model. The classifications were made to prostate cancer if the probability was > = 0.5 and no cancer if the probability was <0.5. PSA alone and A+PSA at three different dilutions were compared with the mean and maximal dilution for sensitivity, specificity, accuracy and area under the curve (AUC). Mean values were selected as the optimal method to obtain sensitivity and specificity, and the receiver operating characteristic (ROC) curve was used to compare the diagnostic power between PSA alone and the combined A+PSA index for distinguishing prostate cancer from BPH and prostatitis. While the addition of any individual autoAb marker barely improved PSA test (data not shown), the addition of all 6 autoAb markers to PSA increased the assay sensitivity (success rate of predicting cancer), specificity (success rate of predicting non-cancer) and prediction accuracy (Table 2). AUC was also increased substantially from 0.66 for PSA alone to 0.91 for A+PSA. Figure 4 shows the ROC curves comparing the diagnostic power between PSA alone and the combined A+PSA index for distinguishing prostate cancer from nonmalignant BPH and/or prostatitis cases commonly seen in the clinic. The 95% bootstrap confidence interval in PSA alone was [0.59, 0.73], whereas the interval of A+PSA including the 6 autoAb was [0.88, 0.95]. A significant difference of AUC between A+PSA and PSA alone was observed (P < 0.001). This pilot study indicated potential benefits of the A+PSA assay in differentiating prostate cancer from non-malignant conditions commonly seen in the clinic.