Introduction: Pro or precursor forms of prostate-specific antigen (PSA) have emerged as potentially important diagnostic serum markers for prostate cancer detection. Immunoassays were developed to measure specific proPSA forms containing propeptides of 2, 4, and 7 amino acids [(-2)proPSA, (-4)proPSA, and (-7)proPSA, respectively].
Methods: Research-use dual monoclonal antibody immunoassays using europium-labeled detection monoclonal antibodies were developed for each form of proPSA. Sera from patients with prostate cancer or benign prostate disease containing 4–10 μg/L PSA were assayed and analyzed by area under the ROC curve (AUC) for specificity and sensitivity.
Results: The proPSA forms had quantification limits of 0.015–0.025 μg/L in serum, with cross-reactivities <1% with PSA. The sum of the proPSA forms divided by free PSA (percentage proPSA) had a higher AUC than did percentage of (-2)proPSA, free PSA, and complexed PSA with AUC (95% confidence intervals) of 0.69 (0.64–0.74), 0.64 (0.58–0.68), 0.63 (0.58–0.68), and 0.57 (0.51–0.62), respectively. The proPSA comprised a median of 33% of the free PSA in cancer and 25% in noncancer sera (P <0.0001). One-third (33%) of cancer samples had >40% proPSA, whereas only 8% of noncancer samples did (P <0.0001). In men with cancer and >25% free PSA, the (-2)proPSA had an AUC of 0.77 (0.66–0.86), with 90% sensitivity and 36% specificity at 0.04 μg/L.
Conclusions: The percentage of proPSA gave better cancer detection in the 4–10 μg/L range than did percentage of free PSA and complexed PSA. (-2)proPSA significantly discriminated cancer in men whose serum had >25% free PSA, for whom there is currently no good marker for cancer detection.
The measurement of serum prostate-specific antigen (PSA)1 is widely used for the screening and early detection of prostate cancer (PCa) (1)(2)(3). Serum total PSA (TPSA) values >4 μg/L are significantly associated with an increased risk of PCa. However, the measurement of PSA suffers from a lack of specificity because various benign prostatic conditions can also led to increased serum PSA (4). The immunologically measurable PSA in serum is present in the noncomplexed ∼33-kDA form, called “free” PSA (FPSA), and in a “complex” (cPSA), primarily with the serum protease inhibitor α1-antichymotrypsin (ACT) (5)(6)(7). TPSA is equal to FPSA plus cPSA (8)(9). Typically, 70–90% of the PSA in serum is cPSA, with the remainder being FPSA. In the blood, FPSA and cPSA have different half-lives and elimination kinetics (10)(11)(12). The proportion, or ratio, of FPSA to TPSA in serum has been demonstrated to significantly improve the discrimination of PCa from benign prostatic hyperplasia (BPH), particularly in patients with TPSA concentrations in the 4–10 μg/L range (13)(14). A higher percentage of free (noncomplexed) PSA in the serum is correlated with a lower risk of PCa, whereas FPSA values <10% are more highly associated with cancer.
The basis of the increased percentage of FPSA in benign disease provides clues for further optimization of PSA testing for cancer. Because PSA is a chymotrypsin-like serine protease, it could be assumed that the FPSA in serum was enzymatically inactive or it would otherwise have formed a complex with protease inhibitors in the serum, such as ACT (15). BPH and PCa tend to be localized within different regions of the prostate gland (16). BPH is characterized by hyperplasia of the inner portion, or “transition zone”, of the prostate, which contains increased concentrations of inactive (internally cleaved) PSA (17). This finding is consistent with the increased percentage of FPSA found in non-cancer serum. Cancers typically develop within the outer portion, or “peripheral zone”, of the prostate gland. The transition and peripheral zone tissues from the same prostate also have two forms of free PSA, called “benign” PSA (BPSA) and proPSA (18)(19). The amount of proPSA was increased in prostate tumor, whereas BPSA was increased in nodular BPH transition zone tissue compared with its concentration in peripheral zone tissue. BPSA and proPSA represent distinct forms of FPSA that are more purely disease-associated than PSA, FPSA, or cPSA. In prostate tissues, virtually all of the PSA is present as free, uncomplexed PSA (19)(20). Thus the enzymatically active PSA from prostate tissues would be expected to complex with serum protease inhibitors on contact with the serum, whereas the inactive forms of PSA, such as BPSA and proPSA, remain as FPSA. Research immunoassays have been developed to measure BPSA and proPSA forms in serum (Beckman Coulter, Inc.; for research use only; not for diagnostic procedures). These additional forms of PSA have increased the complexity of data analysis (21), but have also expanded the possibilities for use of PSA as a PCa detection marker.
ProPSA has been found as the native proPSA form containing a 7-amino-acid pro-peptide leader [(-7)proPSA] (22) as well as forms with truncated pro-peptide leaders. Truncated proPSA forms consist primarily of proPSA with a 5-amino-acid [(-5)proPSA; containing leucine-isoleucine-leucine-serine-arginine attached to the NH2 terminus], 4-amino-acid [(-4)proPSA; containing isoleucine-leucine-serine-arginine attached to the NH2 terminus], or 2-amino-acid [(-2)proPSA; containing serine-arginine attached to the NH2 terminus] pro-peptide leader (23)(24). The more truncated proPSA forms are increasingly resistant to activation by human kallikrein 2 (hK2) and trypsin, and the (-2)proPSA cannot be activated (24). The (-2)proPSA form has received the most attention because it was the primary form found in tumor extracts and shows higher immunostaining in prostate tumor than benign tissue (25). In serum studies of men with PCa, proPSA has been shown to improve the specificity for cancer detection (26)(27)(28). Both proPSA [sum of (-7)-, (-5)-, (-4)-, and (-2)proPSA forms] and the individual (-2)proPSA form appear to provide similar cancer detection properties, although the sum of the proPSA forms may be slightly superior in the 4–10 μg/L PSA range, whereas (-2)proPSA may have superior performance for cancer detection in the 2–4 μg/L PSA range (27). This report describes the analytical performance of three proPSA assays for (-5,-7)proPSA [combined (-5)proPSA and (-7)proPSA], (-4)proPSA, and (-2)proPSA; we also extend the findings for the clinical utility of proPSA forms.
Materials and Methods
Retrospective serum samples containing 4–10 μg/L PSA were selected from men enrolled in PCa screening studies at Washington University (St. Louis, Mo) between 1995 and 2001 who had undergone prostate biopsy. There were 380 serum samples in this study: 238 from men with biopsy-confirmed cancer and 142 from biopsy-negative men. The median age of the donors was 66 years. All serum samples were collected under an Internal Review Board-approved protocol that passed Health Insurance Portability and Accountability Act (HIPAA) compliance. Informed consent was obtained from all men. Blood was processed within 3 h of draw, and serum was stored in portions at −70 °C, with no freezing/thawing before the proPSA assay. Samples with >10 μg/L PSA were identified retrospectively from men with biopsy-confirmed PCa.
The development of monoclonal antibodies (mAbs) to proPSA forms has been described previously (24). This includes three different mAbs to proPSA forms containing pro-peptide-leader sequences of 7, 5, 4, and 2 amino acids, designated (-5,-7)proPSA [combined (-5)proPSA and (-7)proPSA], (-4)proPSA, and (-2)proPSA, respectively. Samples were blinded until after proPSA assays were performed. The immunoassay format was a dual monoclonal sandwich assay in a microtiter plate under standard conditions that uses a biotinylated capture anti-PSA mAb and europium-labeled proPSA-specific mAbs for detection by a Victor 1420 multilabel counter (Perkin-Elmer). The term “proPSA” without prefix indicates the combined sum of (-2)proPSA, (-4)proPSA, and (-5,-7)proPSA. The hydrophobic interaction chromatography buffers and high-performance hydrophobic interaction chromatography (HIC-HPLC) elution conditions have been described previously (24).
The immunoassay protocol was as follows: 50 μL of biotinylated anti-PSA antibody PSM 773 at 5 mg/L in Tandem® PSA zero cal diluent was added to a streptavidin-coated microplate and allowed to react at room temperature for 1 h with shaking. The plate was then washed five times with Tandem E wash, and 50 μL of Tandem PSA zero cal diluent (Beckman Coulter) was then added to the plate, followed by 50 μL of the serum or antigen to be tested. The mixture was allowed to react at room temperature for 2 h as above. The plate was washed five times with Tandem E wash, after which 100 μL of europium-labeled detection mAb, prepared according to manufacturer’s instructions (Perkin-Elmer) for the intended measurement of each form of proPSA, was added to the plate. For (-2)proPSA, this was PS2X373; for (-4)proPSA, it was PS2V 411; and for (-5,-7)proPSA, it was PS2P309. The mixture was allowed to react at room temperature for 1 h as above. The plate was then washed five times with Tandem E wash and read with the Victor instrument.
ProPSA forms were secreted into and purified from the medium of mammalian AV12 cells as described previously (29). The individual standards or calibrators of the proPSA forms were quantified by FPSA assay. The homogeneity of the proPSA forms was established by HIC-HPLC, as seen in Fig. 1⇓ , and N-terminal sequencing as described previously (24). The term minimum detectable concentration (MDC) is defined as the mean plus 2 SD above the mean for 20 individual female serum samples.
The concentration of PSA in serum was determined by Hybritech Tandem PSA and Tandem free PSA assays (Beckman Coulter). The values for cPSA were calculated by subtracting the FPSA from TPSA, a procedure shown to give cPSA with diagnostic performance indistinguishable from cPSA measured with the Bayer assay (9). Results were analyzed using ROC curves for the sensitivity, specificity, and ROC area under the curve (AUC). The χ2 test for comparison of proportions was used to determine significance of specificity values. All statistical and ROC analyses were performed with MedCalc software, Ver. 7.1 (http://www.medcalc.be/).
Three immunoassays were developed to measure serum concentrations of native and truncated forms of proPSA (proPSA). The individual PSA and proPSA forms were isolated and purified by HIC-HPLC. Fig. 1⇓ shows a chromatographic overlay of the individual proPSA forms individually purified by HIC-HPLC. The first of these assays measured the native proPSA with the full 7-amino-acid pro-peptide leader [(-7)proPSA] together with the slightly truncated (-5)proPSA containing a 5-amino-acid pro-peptide leader. These forms also coeluted in HIC-HPLC (Fig. 1⇓ ). This assay was designated (-5,-7)proPSA. The second assay measured (-4)proPSA, which contains a 4-amino-acid pro-peptide leader, and the third assay measured the most truncated form of proPSA, (-2)proPSA, which contains only a 2-amino-acid pro-peptide leader and is typically found in significant quantities in serum. All assays had <0.2% cross-reactivity with PSA and <2% cross-reactivity with one another (Table 1⇓ ). The analytical MDC, representing the lowest concentrations that could be quantified in buffer, was <0.01 μg/L for all assays. The biological MDCs for the quantification limit in serum were 0.025 μg/L for the (-5,-7)proPSA and (-4)proPSA assays and 0.015 μg/L for (-2)proPSA. The assays were linear up to 10 μg/L and remained linear on >100-fold serial dilution (r = 0.99). In the latter case, the linearity on dilution was not affected regardless of whether calibrators were diluted in buffer or whether serum samples with initially high proPSA concentrations (>1 μg/L) were serially diluted into serum from women down to concentrations near the MDC.
The proPSA assays were stable over time. The CV for the triplicate wells of a single sample in each plate was typically <5%. Table 2⇓ shows the intraday and interday comparison of assay CV for the different proPSA assays. The intraday results report the mean CV from plate to plate of four plates during a single day of testing, whereas more rigorous interday values indicate the variation (CV) among 28 plates run on 7 different days over 6 weeks. High and low serum pools containing proPSA were tested in each experiment to evaluate proPSA stability and reproducibility in serum. The CV of the measured fluorescence units at each point in the calibration curve is also shown. In all of these studies, none of the CV values exceeded 20% for any assay component, and most values were closer to 10% throughout the range of analyte concentrations. No significant changes or trends in the absolute concentrations of any of the proPSA forms added to the frozen serum calibrators were observed over ∼6 weeks.
Shown in Fig. 2⇓ are the distributions of (-2)proPSA, (-4)proPSA, (-5,-7)proPSA, and the sum proPSA as a percentage of FPSA for the cancer (Fig. 2A⇓ ) and benign (Fig. 2B⇓ ) cohorts. The serum values for the sum proPSA in Fig. 2⇓ were sorted and plotted from lowest to highest. Although the total range of percentage of proPSA was similar between the benign and cancer sera (5–60%), the proportion of the sample population with higher or lower percentages of proPSA was significantly different between the two groups. The x axis shows the percentage of the samples in which the sum proPSA represents 0–20%, 20–30%, 30–40%, and >40% of the FPSA. Only 14% of the cancer samples had <20% proPSA, whereas 36% of the benign samples had <20% (P <0.0001). In contrast, 33% of the cancer samples had >40% proPSA, whereas only 8% of the benign samples were in this range (P <0.0001). Overall, the median percentage of proPSA was 25% in benign samples compared with 33% in the cancer sera (P <0.0001; Table 3⇓ ). The individual proPSA forms [(-2)proPSA, (-4)proPSA, and (-5,-7)proPSA] in Fig. 2⇓ showed similarly ascending trends that were comparable to the rising sum proPSA. The relative proportions of the percentages of (-2)proPSA, (-4)proPSA, and (-5,-7)proPSA to the total percentage of proPSA were similar between the cancer and benign samples: 11%, 35% and 53%, respectively (Table 3⇓ ). Within this cohort, 12% of the (-2)proPSA values were below the MDC of 0.015 μg/L: 7% of the cancer samples and 16% of the benign samples. The percentage of proPSA in the FPSA increased gradually with increasing TPSA (Table 4⇓ ). Serum with >100 μg/L PSA had a median of 50% proPSA compared with 33% in the 4–10 μg/L range. The (-2)proPSA form showed the largest relative increase of the proPSA forms in cancer serum with increasing TPSA.
The ROC curves for 238 benign and 142 cancer sera in the TPSA range of 4–10 μg/L PSA are shown in Fig. 3⇓ . The percentage of proPSA (sum proPSA:FPSA) gave the highest AUC of 0.69, compared with (-2)proPSA (0.64), percentage of FPSA (0.63), and cPSA (0.57). The AUC for the percentage of proPSA was significantly higher than that for cPSA only and TPSA (P <0.05). However, the specificities at 95%, 90%, and 85% sensitivity for percentage of proPSA were 19% (95% confidence interval, 10–31%), 31% (20–44%), and 40% (28–54%), respectively, all of which were significantly greater than the specificity values observed in the other forms of PSA (Table 5⇓ ).
The proPSA forms enhanced the detection of PCa in samples with high and low percentages of FPSA. In this study, the 238 benign samples had a median of 19% FPSA, whereas the 142 cancer sera had a median of 15.5% (P <0.0001). Fig. 4⇓ shows ROC curves for the 61 benign and 19 cancer sera with FPSA >25%. The percentage of (-2)proPSA gave the highest AUC of 0.77 (95% confidence interval, 0.66–0.86) compared with the percentages of proPSA [0.69 (0.58–0.79)], FPSA [0.53 (0.41–0.64)], and cPSA [0.50 (0.39–0.62)]. The AUC for the percentage of (-2)proPSA was significantly higher than all PSA forms except for the percentage of proPSA (Table 5⇓ ). Using percentage of (-2)proPSA at a cutoff of 2.5%, we found that 90% of the cancers would have been detected with 36% of the men spared biopsies (sensitivity, 90%; specificity, 36%). The specificity for percentage of (-2)proPSA was significantly higher than the other forms of PSA except for the percentage of proPSA (P <0.0001). The concentration (μg/L) of (-2)proPSA [AUC, 0.75 (95% confidence interval, 0.68–0.84)] in the serum also provided equivalent value to the percentage of (-2)proPSA. At a cutoff of 0.040 μg/L (-2)proPSA, 17 of 19 cancers (90%) were detected with 36% specificity (percentage of men spared biopsies). The specificity for (-2)proPSA was also significantly higher than that of the percentage of FPSA or cPSA (P <0.0001). The specificities were not different at 95% sensitivity. Almost identical results for sensitivity and specificity were obtained with samples containing >22% FPSA.
Identification of the cancer samples was also improved in samples with <15% FPSA. Shown in Fig. 5⇓ are the ROC curves for the 66 benign and 65 cancer sera with 4–10 μg/L PSA and FPSA <15%. The AUC was 0.70 (95% confidence interval, 0.62–0.78) for percentage proPSA, 0.67 (0.58–0.75) for percentage (-2)proPSA, 0.59 (0.50–0.68) for percentage free PSA, and 0.54 (0.45–0.63) for cPSA. The AUC for percentage proPSA was significantly higher than the AUC for cPSA and TPSA (Table 5⇓ ). However, the specificity at 95% sensitivity was significantly higher than that for all PSA forms except percentage of (-2)proPSA and was higher than all PSA forms at 90% and 85% sensitivity (Table 5⇓ ).
This study describes the development of highly sensitive and specific immunoassays for intact and truncated forms of proPSA. Analysis of benign and cancer sera demonstrated that the percentage of proPSA can improve the detection of cancer in the 4–10 μg/L PSA range (Fig. 3⇓ ) and that it may have unique value in further differentiating cancer in those men with >25% FPSA (Fig. 4⇓ and Table 5⇓ ). Men with known high percentages of FPSA have already been selected as a population with a low probability of cancer, and there are currently no methods to provide additional diagnostic value in this population. This becomes a greater problem when considering repeat biopsies in initially biopsy-negative men with increased TPSA. Both the percentage of (-2)proPSA [ratio of (-2)proPSA to FPSA] and the serum concentration (in μg/L) of (-2)proPSA were similarly effective in identifying cancer in samples with a high percentage of FPSA. The (-2)proPSA detected 90% of the cancers (17 of 19) while sparing 36% of men from unnecessary biopsies. The overall percentage of (-2)proPSA in these men did not account for a significant proportion of the increased FPSA because proPSA still represented only ∼5% of the TPSA in the cancer serum (Table 3⇓ ). However, the presence of the (-2)proPSA form appears to be a sensitive indicator of cancer. The median serum concentration of (-2)proPSA in the serum of men with cancer and >25% FPSA was 0.073 μg/L, the highest concentration observed in any of the analyses (Table 3⇓ ). This observation clearly demonstrates the contrasting characteristic of proPSA, which is that it is strongly associated with PCa although it is a form of FPSA. The median value (0.031 μg/L) for (-2)proPSA in the total population of men with cancer was significantly lower than that (0.073 μg/L) for men with cancer and >25% free PSA (P <0.0001). The percentage of proPSA provided a similar improvement in discriminating additional cancer in men with <15% FPSA (Table 3⇓ and Fig. 5⇓ ).
This study demonstrates that serum proPSA is clearly cancer-associated and can be used to augment current PSA assays to improve cancer detection. The results in Table 3⇓ suggest that additional algorithms may be useful in evaluating the nonlinear relationships among the different PSA forms and within different patient populations. Logistic regression has shown that proPSA adds significant diagnostic value (28). Similar results were obtained in this study. In particular, backward logistic regression showed that the (-2)proPSA form was the only variable to significantly predict the presence of cancer (P = 0.001) in men with FPSA >25%. In most circumstances the percentage of proPSA in FPSA provides the best clinical value; therefore, FPSA must also be known. As an absolute concentration in the serum, proPSA was not significantly higher in cancer than in benign conditions, but it was highly significant as a ratio of FPSA (Table 3⇓ ). The ratio of proPSA to FPSA may provide a more stable indicator because studies suggest that internal ratios of different PSA forms are subject to less biological variability than PSA components measured individually (30). Additional stability studies are needed, but it is likely that internal ratios of FPSA forms would be more stable than fluctuations in the percentage of FPSA as a whole. When we examined relatively high FPSA (>25%) or low FPSA (<15%), the AUCROC value for percentage of FPSA decreased significantly compared with the entire range, for which it is intended to be used (Table 3⇓ ). It is in these outside ranges that the percentage of proPSA was particularly useful in further identifying cancer.
The proPSA assays described here had cross-reactivities <0.2% with PSA (Table 1⇓ ). This aspect of the study was limited by the inability to demonstrate the total absence of proPSA forms in the purified PSA. The front half of the chromatographic PSA peak from seminal plasma PSA was collected to minimize contamination with proPSA forms, particularly (-2)proPSA, which elutes nearest to PSA (Fig. 1⇓ ). Minor concentrations of proPSA have been measured in seminal plasma PSA (S. Mikolajczyk, unpublished results); therefore, it is not possible to determine whether 0.2% is attributable to cross-reactivity or contamination. Similar apparent cross-reactivities with PSA were initially observed in the immunoassays for hK2 until recombinant preparations of PSA were produced that were known to contain no hK2. In that case, the apparent hK2 assay cross-reactivity of 0.5% with PSA decreased to <0.001% when recombinant PSA was used (31). In the current study, all natural and recombinant sources of PSA may contain minor concentrations of proPSA forms. The current values for cross-reactivity (1.5% to <0.2%) seen throughout the different preparations in Table 1⇓ represent upper limits because of the almost inevitable minor contamination with other forms of proPSA that occurs during conventional purification.
In conclusion, these studies indicate that it is important to measure all proPSA forms in the serum and that the sum of the proPSA forms provides the best clinical value in the 4–10 μg/L range. Individually, the (-5,7)proPSA and (-4)proPSA forms did not appear to offer additional value and had generally lower specificities than the sum of the proPSA forms. We therefore have developed a pan-proPSA assay for use as a single assay in future experiments (data not shown). The (-2) form of proPSA had clinical value similar to the percentage of proPSA in most cases, but it appeared to offer unique diagnostic value in samples with high percentages of FPSA. The (-2)proPSA form has also been shown to be the best marker in samples with TPSA <4 μg/L (26)(27). Recent clinical studies suggested a potential role for proPSA forms in detecting more aggressive forms of PCa (32).
These studies were funded by Beckman Coulter, Inc. All authors except Dr. Catalona are employees of Beckman Coulter, Inc. We thank Christine Knott for excellent technical support.
↵1 Nonstandard abbreviations: PSA, prostate-specific antigen; PCa, prostate cancer; TPSA, total PSA; FPSA, free PSA; cPSA, complexed PSA; ACT, α1-antichymotrypsin; BPH, benign prostatic hyperplasia; BPSA, benign PSA; hK2, human kallikrein 2; mAb, monoclonal antibody; HIC-HPLC, high-performance hydrophobic interaction chromatography; MDC, minimum detectable concentration; and AUC, area under the curve.
- © 2004 The American Association for Clinical Chemistry