BACKGROUND: Biomarker validation remains one of the most challenging constraints to the development of new diagnostic assays. To facilitate biomarker validation, we previously developed a chromatography-free stable isotope standards and capture by antipeptide antibodies (SISCAPA)-MALDI assay allowing rapid, high-throughput quantification of protein analytes in large sample sets. Here we applied this assay to the measurement of a surrogate proteotypic peptide from protein C inhibitor (PCI) in sera from patients with prostate cancer.
METHODS: A 2-plex SISCAPA-MALDI assay for quantification of proteotypic peptides from PCI and soluble transferrin receptor (sTfR) was used to measure these peptides in 159 trypsin-digested sera collected from 51 patients with prostate cancer. These patients had been treated with radiation with or without neoadjuvant androgen deprivation.
RESULTS: Patients who experienced biochemical recurrence of prostate cancer showed decreased serum concentrations of the PCI peptide analyte within 18 months of treatment. The PCI peptide concentrations remained increased in the sera of patients who did not experience cancer recurrence. Prostate-specific antigen concentrations had no predictive value during the same time period.
CONCLUSIONS: The high-throughput, liquid chromatography–free SISCAPA-MALDI assay is capable of rapid quantification of proteotypic PCI and sTfR peptide analytes in complex serum samples. Decreased serum concentrations of the PCI peptide were found to be related to recurrence of prostate cancer in patients treated with radiation with or without hormone therapy. However, a larger cohort of patients will be required for unequivocal validation of the PCI peptide as a biomarker for clinical use.
Since the discovery of an immunoglobulin light-chain as a diagnostic indicator for multiple myeloma by Henry Bence Jones in 1847 (1), more than 1200 proteins have been claimed as potential cancer biomarkers (2). However, validation of candidate biomarkers has lagged far behind their discovery, presenting a roadblock to the development of new diagnostic assays. Addressing the dilemma of biomarker validation for clinical use has been a purported goal in the field of proteomics over the past decade, yet in our informal literature surveys we could find only 3 publications in which specific proteins were measured by use of selected reaction-monitoring assays in more than 1000 patient samples (3–5).
To overcome the challenges inherent in using complex samples for selected reaction-monitoring assays (6), an immuno–mass spectrometry (MS)5 assay known as stable isotope standards and capture by antipeptide antibodies (SISCAPA) was developed (7) in which a high-affinity antipeptide antibody was used to enrich peptides of interest from trypsin-digested human plasma or sera before MS analysis. Recently we reported a modified SISCAPA approach that significantly reduced background peptides, eliminating the need for upstream liquid chromatography (LC) (8–9). This technological improvement allowed us to combine the SISCAPA work flow with high-throughput MALDI-TOF/MS (8) and a solid-phase extraction (SPE)-tandem MS (MS/MS) technology known as RapidFire (9). However, measurement of biomarkers in large numbers of clinical specimens such as sera or plasma has not been tested using these LC-free platforms.
In the work reported here, we developed a high-throughput SISCAPA-MALDI-TOF method and used it to measure a selected tryptic peptide of human protein C inhibitor (PCI) in sera obtained in a longitudinal study of 51 patients with prostate cancer. PCI belongs to the superfamily of serine protease inhibitors and is best known for its role in inhibiting activated protein C, an anticoagulant that is also involved in regulating antiapoptotic and antiinflammatory pathways (10). In a biomarker discovery effort it was reported that serum concentrations of an N-terminal peptide fragment of PCI were lower in prostate cancer patients who had a biochemical recurrence of the disease than in patients who received the same treatment (radical prostatectomy) but who did not have a recurrence (11). The mechanism by which a decline in serum concentrations of this N-terminal PCI peptide might be related to prostate cancer recurrence is not understood.
We chose to examine the potential predictive value of a selected PCI tryptic peptide in prostate cancer patients who received either external beam radiation or brachytherapy (with or without hormone therapy), because prostate-specific antigen (PSA), the gold standard biomarker for monitoring prostate cancer, is ineffective in predicting recurrence at an early stage in these patients (12).
Materials and Methods
PEPTIDES, ANTIBODIES, AND SISCAPA PROTOCOLS
Our general methods for synthesis and quantification of peptides, derivation and selection of high-affinity rabbit monoclonal antibodies, trypsin digestion of human plasma, and SISCAPA work flow have been described previously (13), whereas detailed methods used specifically in the current manuscript are described in the Supplemental Methods in the Data Supplement that accompanies the online version of this report at http://www.clinchem.org/content/vol59/issue10.
HUMAN PLASMA AND SERA
Normal human plasma samples were purchased from BioReclamation. PCI-deficient human plasma was purchased from Affinity Biologicals. Sera from prostate cancer patients and controls were obtained from the BC Cancer Agency (BCCA) in Victoria, British Columbia, after ethics approval was obtained from the Research Ethics Board of the University of British Columbia and the BCCA (protocol numbers H06-03666 and H03-60026) and from the Human Research Ethics Board of the University of Victoria (ethics protocol number 11-151). These samples were mainly collected between 2004 and 2008 from patients with prostate cancer who received external beam radiation with neoadjuvant androgen deprivation or brachytherapy with or without neoadjuvant androgen deprivation. Sera were collected from a total of 51 patients before any treatment (pretreatment). For most patients, longitudinal samples were also collected after treatment. The number of posttreatment samples collected (from 2 to 7) varied between individuals owing to a lack of staff for blood collection or the unwillingness of patients to donate blood on every visit. Importantly, the outcomes of treatment were not known at the time of sample collection.
Of the 51 patients, 34 had donated samples both before treatment and within the first 18 months after treatment. Long-term follow-up data were available for all 34 of these patients. Of the remaining 17 patients, 2 passed away due to secondary cancers (#6 and #12), 3 had known outcomes at the time of sample collection (#4, #16, and #17), 3 did not have long-term follow-up data (#1, #31, and #38), and 9 patients (#10, #15, #21, #22, #36, #37, #40, #45, and #48) were missing either pretreatment samples or posttreatment samples within the first 18 months of treatment (see online Supplemental Table 1).
We used a combination curve, essentially as previously described (8), to examine the linearity and imprecision of PCI peptide measurement by SISCAPA-MALDI. Synthetic light and heavy SIS PCI peptides were used to create 2 peptide solutions (using 0.1% formic acid in water), 1 with a light/heavy (L:H) ratio of 10:1 and the other with an L:H ratio of 1:10. These solutions represented the 2 ends of an 11-point curve over a 100-fold dynamic range. The remaining points on the curve were generated by mixing equal volumes of consecutive solutions; for example, when equal volumes of the 10:1 and 1:10 solutions were mixed, a third solution was generated with an L:H peptide ratio of 1:1 representing the center of the curve. With the use of this strategy, the total peptide concentration (i.e., L + H) in all samples remained constant at 660 fmol/μL. To test the assay in a human plasma matrix, 2 μL of each sample was pipetted into 10 μL of trypsin-digested PCI-deficient plasma, and the SISCAPA capture procedure was performed. Peptide solutions were also directly spotted onto a MALDI target plate without antibody capture. Four replicates of this experiment were conducted to obtain CVs corresponding to each point on the combination curve.
LIMITS OF DETECTION AND QUANTIFICATION AND COMPARISON TO REFERENCE METHOD
To determine the limit of detection (LOD) and limit of quantification (LOQ), 2 calibration curves were generated in digested, pooled plasma. The first calibration curve, called a forward curve, was generated by spiking constant amounts of the heavy peptide (500 fmol/well) and varying amounts of the synthetic light peptide to generate a 12-point curve, with the light peptide being titrated from 1000 fmol to 0 fmol (2-fold dilutions; no spike in the 12th sample). The forward curve revealed the endogenous concentration of the analyte in the sample being analyzed. We generated the second curve, called a reverse curve, by spiking constant concentrations of the light peptide (500 fmol/well) and 2-fold dilutions of the heavy peptide (from 1000 fmol to 0.5 fmol). The reverse curve was used to determine the LOD and LOQ of the assay with respect to the endogenous concentration (14). The LOD was defined as the lowest spiked concentration of the SIS peptide that was identifiable in all replicates in the experiment. The LOQ was defined as the lowest concentration of the analyte that was identifiable in all replicates with a CV of <20%. To compare our SISCAPA-MALDI assay to an LC-MS/MS reference method, we performed a replicate of the forward/reverse curve experiment using an Agilent LC-6490 triple quadrupole mass spectrometer.
IMPRECISION AND ROBUSTNESS STUDIES
We measured the PCI and soluble transferrin receptor (sTfR) analytes in a serum sample from one individual to establish the robustness and imprecision of these assays. Ten microliters of the sample were separately digested 6 times/day over 5 days (n = 30). The CV was calculated for intraassay and interday variability and the mean analyte concentrations were compared using one-way ANOVA.
Intraassay (well-to-well) imprecision for the PCI assay was additionally tested by digesting 10 different human plasma samples and measuring the concentrations of the PCI analyte for each individual sample over 10 SISCAPA capture replicates (n = 100). Interassay (day-to-day) variation of the PCI assay was further tested by measuring the PCI analyte in triplicate SISCAPA captures in digested sera from 25 patients with prostate cancer on 2 occasions, 12 months apart.
PEPTIDE RECOVERY STUDIES
The recovery of PCI peptide from plasma digests was investigated by spiking 250 fmol/well and 50 fmol/well of the synthetic light peptide into PCI-deficient trypsin-digested plasma, followed by peptide enrichment using SISCAPA. The heavy peptide was spiked into the elution plate at 500 fmol/well before MS analysis. The experimental ratio of L:H peptides compared to the theoretical ratio was used to calculate percent recovery.
MEASURING PCI IN PATIENTS WITH PROSTATE CANCER
This study was conducted in a double-blind format. Sera collected at the BCCA were randomized and coded by a third party before they were presented for analysis. Three SISCAPA capture replicates were performed to measure PCI and sTfR peptide concentrations in all patients' sera. Patients in this study are still being monitored for biochemical recurrence of cancer. The patient outcome data in this report are up to date as of November 2012. To determine biochemical recurrence of prostate cancer, the BCCA uses the Phoenix definition (15). Based on this definition, a patient is considered to have a biochemical recurrence if he experiences a spike of >2.0 ng/mL in PSA concentrations compared to the nadir concentration, defined as the lowest PSA measurement obtained for the patient after treatment.
After SISCAPA measurement of the PCI analyte in 159 randomized serum samples, the key used for the double-blind study was revealed to allow assembly of the data into pretreatment and posttreatment sets for 2 groups of patients with recurrent and nonrecurrent prostate cancer. If multiple posttreatment samples were drawn for an individual, the mean PCI concentration in all samples was used for analysis.
Repeated measures ANOVA at 95% CI was used to assess the changes in PCI concentrations as a function of time (MedCalc software version 12.7.0). The PCI concentration in posttreatment samples in both groups was used to evaluate the ability of this analyte to predict biochemical recurrence of prostate cancer. Given the varying sample sizes between the 2 groups, we used the Matthews correlation coefficient (MCC), which is regarded as a balanced measure even when sample sizes vary, to appraise PCI as a predictor of biochemical recurrence of cancer. The formula for MCC calculation is: MCC = [(TP × TN) − (FP × FN)]/square root[(TP + FP) (TP + TN) (FP + FN) (TN + FN)], where TP is true positive, FP is false positive, TN is true negative, and FN is false negative. The same analysis was performed using the PSA data collected within the first 18 months after treatment to provide a point of reference for the performance of the PCI analyte. We also generated an ROC curve for both PCI and PSA using XLSTAT (Addinsoft).
LINEARITY OF L:H PEPTIDE RATIOS IN DIGESTED HUMAN PLASMA
The experimental L:H peptide ratios from the combination curve were plotted against theoretical values (Fig. 1, A and B). The linear fit for the curve had a slope of 1.267 (0.001) and a y intercept of −0.112 (0.002) at the 95% CI (R2 = 0.999). The deviation of the slope from the ideal slope of 1.0 is thought to be due to imprecisions in the concentration of the SIS peptide.
Four technical replicates of this experiment were performed to examine the percentage CV for each data point across the 100-fold dynamic range. As observed previously with different peptides (8), when the peptide solutions were directly spotted onto the target, the CVs ranged from 2.1% to 8.5%, with the best CVs toward the center of the dynamic range (i.e., close to L:H ratio of 1:1). Importantly, a similar trend was observed when the PCI peptide analyte was spiked into digested plasma and subjected to antibody capture (Fig. 1C). The CVs here ranged from 3.7% to 10.9%. We also tested linearity by mixing PCI-deficient plasma with pooled plasma containing endogenous analyte to form a 1:1 mixture. An R2 value of 0.982 was observed (see online Supplemental Fig. 1).
DETERMINING THE LOD AND LOQ
The LOD for the PCI peptide was 4.4 (0.13) fmol, approximately 40-fold lower than the endogenous concentration of the analyte in 10 μL of pooled plasma. Consistent with the results reported above, the CV at the endogenous concentration was 3.8%. The imprecision of measurement increased below 15 fmol, where the CV fluctuated (Fig. 2A). The LOQ for the PCI analyte was 20.2 (2.2) fmol, at which point the CV was 10.8%. Based on the reverse curve it was determined that the assay was linear from 4 to 1000 fmol of the analyte, with an R2 = 0.999 [y = 0.872 (±0.001)x − 0.974 (±0.200)]. To determine if the correct endogenous range for this analyte was measured using SISCAPA-MALDI, we repeated replicates of the forward and reverse curves using LC-MS/MS with a triple quadrupole mass spectrometer (Fig. 2B). The same endogenous concentration of the PCI peptide analyte (160 fmol in 10 μL of pooled plasma) was measured with both platforms.
IMPRECISION OF THE SISCAPA-MALDI ASSAY FOR THE PCI PEPTIDE
The mean values for the PCI analyte on 5 consecutive days were 69.7 (5.5), 65.4 (8.0), 69.5 (9.5), 74.7 (6.1), and 72.5 (5.3) pmol/mL. A one-way ANOVA revealed a P value of 0.24, indicating that there were no significant differences (α = 0.05) in the mean PCI concentrations measured on separate days. The mean intraassay CV was 10% including the digestion step and 3.2% excluding the digestion step. The interassay CV was 5% for the PCI analyte. Interassay imprecision was also tested by measuring (in triplicate) the PCI-specific peptide in digested sera from 25 patients with prostate cancer at 2 time points, 12 months apart. An R2 of 0.965 was observed between the 2 measurements (Fig. 3). The linear fit had a slope of 1.162 (0.003) and a y intercept of −1.056 (0.057) at the 95% CI. Similar results were obtained for the sTfR analyte and when pooled human plasma was used. Freezing and thawing (a minimum of 6 cycles) and storage at room temperature (or at 4 °C) for short periods of time (<24 h) did not affect the measurement of the PCI or sTfR analytes. The effect of storage for longer periods of time was not tested. Addition of bilirubin and hemoglobin, which are known to interfere with some immunoassays, had no impact on the SISCAPA-MALDI assay when added at high concentrations (10–40 μg/mL).
RECOVERY OF THE SURROGATE PCI PEPTIDE
To determine the analytical recovery of the PCI analyte from digested plasma, light peptide was spiked into the PCI-deficient digest at concentrations of 250 fmol and 50 fmol. The SIS was spiked into the elution buffer at 500 fmol/well. The theoretical ratio of L:H peptide was compared to the experimentally recovered amount of the light peptide to calculate recovery. The mean recovery was 81.7% (1.7%) for both spiked concentrations (see online Supplemental Fig. 2), which was consistent with that reported for other SISCAPA assays when an antibody with subnanomolar affinity was used for peptide capture (16).
MEASURING THE PCI PEPTIDE IN PATIENTS WITH PROSTATE CANCER
In a 2-plex format, the PCI and sTfR peptide analytes were measured in 159 serum digests from 51 patients with prostate cancer and in 10 sera from aged-matched, cancer-free controls. Tryptic digestion was monitored using the digestion control peptide. The tryptic release of this peptide and its abundance relative to the SIS peptide were monitored in all samples. We observed a CV of 2.3% based on the ratio of the digestion control peptide to the SIS peptide, suggesting that the digestion step was uniform.
Because the patient samples were collected over a protracted period (months to years), the sTfR analyte was measured as a control to determine whether sample collection and storage caused adverse effects. sTfR was chosen because the serum concentrations of this protein are roughly the same as those of PCI (17–18). Triplicate samples were run and the mean CV was determined to be 4.7% for the PCI peptide and 7.6% for the sTfR peptide. The sequence identities of the 2 peptides were confirmed by MS/MS (see online Supplemental Fig. 3).
Quantitative measurements of the PCI and sTfR peptides were performed using a 4800 MALDI TOF/TOF instrument (Applied Biosystems) and a microflexTM LT (Bruker Daltonics). An R2 of 0.971 for the PCI peptide [linear fit slope of 1.092 (0.001) and y intercept of 3.201 (0.021) at 95% CI] and 0.878 for the sTfR peptide [linear fit slope of 1.046 (0.002) and y intercept of 3.267 (0.017) at 95% CI] were observed when comparing the 2 instruments (Fig. 4). The mean CVs on the microflex instrument were 5.7% for the PCI peptide and 8.2% for the sTfR analyte. The variation in the concentrations of the sTfR analyte was 8-fold lower than the variation of the PCI peptide (7.3 vs 59.4 pmol/mL, Fig. 5A), suggesting that the variation in the concentrations of the PCI peptide was not due to differences in sample collection, storage, or processing.
We analyzed the PCI peptide data for 34 of the patients, 9 of whom showed a biochemical recurrence of their cancers within 48–60 months of their radiation treatment. Most interesting was the statistically significant decline (P < 0.05) observed in the concentrations of the PCI analyte in the first 18 months after treatment in sera from patients who had a biochemical recurrence of their cancers (Fig. 5B). The PSA values for these patients' samples, measured in the same sera at exactly the same time points, were also available (determined at the BCCA). ROC analysis was performed on the data for both the PCI peptide and PSA: an area under the curve (AUC) of 0.774 was obtained for the PCI peptide compared to an AUC of 0.506 for PSA (Fig. 6A). Using a cutoff of 12.77 pmol/mL for PCI peptide measurements, the diagnostic sensitivity of the assay was calculated to be 0.778 and the diagnostic specificity 0.846 (Fig. 6B). At the cutoff value of 4.20 ng/mL, the diagnostic sensitivity and specificity of the PSA assay were 0.222 and 0.923, respectively, values consistent with the diagnostic sensitivity and specificity of PSA reported in the literature (19). With the use of the same cutoff values, the MCC was calculated for both analytes to evaluate their predictive value, taking into account the varying sample sizes between the 2 groups: PCI had an MCC value of +0.6, whereas the same measure for PSA resulted in a value of +0.2.
Validating protein biomarkers for potential clinical application is a challenge because prospective biomarkers must be measured in large numbers of samples. To eliminate this bottleneck, we developed the LC-free high-throughput SISCAPA® assays using both MALDI-TOF (8) and triple quadrupole MS/MS (9) platforms that allowed very short sample analysis times (<10 s/sample). At this level of throughput, biomarker validation can be temporally and fiscally feasible.
In this study we demonstrated that a SISCAPA-MALDI assay could be used to quantify peptide analytes in samples with an acceptable imprecision. It is important to note that imprecision of peptide measurements by MALDI is related to ion counts (8), which in turn are associated with ion suppression effects, especially at low concentrations of analytes. The low CVs obtained using the SISCAPA work flow described in this manuscript demonstrate that any ion suppression effects have been minimized by reducing the nonspecific background peptides. We also demonstrated that relatively inexpensive MALDI-TOF instruments (for example the user-friendly Bruker microflex LT mass spectrometer) could be used for peptide measurements by SISCAPA. However, MS/MS and thus unequivocal identification of the analyte is not currently supported by the microflex instrument, which may pose a challenge for clinical implementation. In the absence of MS/MS, multiple peptides from the same protein may offer one alternative to further ensure the identity of the protein target under study.
We applied the SISCAPA work flow to evaluate the proteotypic PCI peptide as an early predictor of biochemical recurrence in prostate cancer patients who receive radiation therapy. The patients who eventually developed biochemical recurrence were compared to other patients who received the same treatment but who did not show biochemical recurrence of their cancers, and thus the study was internally controlled. We observed a significant posttreatment decline in concentration of the PCI peptide analyte in patients who eventually experienced cancer recurrence. Statistical analysis of the data demonstrated that, in this cohort of patients, the ability of the PCI analyte to predict prostate cancer recurrence at an early stage was superior to that of the current clinical biomarker, PSA. However, the utility of the PCI assay needs to be further tested in a significantly larger cohort of patients before any clinical implications can be drawn.
This study did not follow a randomized controlled trial design in which baseline equality is unequivocally assured (20). For validation of the PCI peptide as a predictor of cancer recurrence, such a randomized controlled trial design could be employed. Nevertheless, precise quantification of PCI peptide in the large number of serum samples required for clinical validation is clearly attainable using the SISCAPA-MALDI approach presented here.
We thank A. Camenzind for helping us with the Applied Biosystems-MDS SCIEX 4800 MALDI-TOF/TOF instrument. We thank N. Nesslinger and N. Chima at the BC Cancer Agency for their help in organizing the clinical data. Last but certainly not least, we are grateful to all the patients whose blood donation made this study possible.
↵5 Nonstandard abbreviations:
- mass spectrometry;
- stable isotope standards and capture by anti-peptide antibodies;
- liquid chromatography;
- solid phase extraction;
- tandem MS;
- protein C inhibitor;
- prostate-specific antigen;
- the BC Cancer Agency;
- soluble transferrin receptor;
- limit of detection;
- limit of quantification;
- Matthews correlation coefficient;
- area under the curve.
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.
Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:
Employment or Leadership: M. Razavi, University of Victoria; N.L. Anderson, SISCAPA Assay Technologies and Clinical Chemistry, AACC; T. Pearson, University of Victoria.
Consultant or Advisory Role: M. Razavi, SISCAPA Assay Technologies; T. Pearson, SISCAPA Assay Technologies.
Stock Ownership: N.L. Anderson, SISCAPA Assay Technologies; T. Pearson, SISCAPA Assay Technologies.
Honoraria: None declared.
Research Funding: Agilent Technologies, Genome Canada, Genome BC, and SISCAPA Assay Technologies (which supplied the Kingfisher magnetic particle processor and SISCAPA antibodies); J. Lum, the Canadian Institutes for Health Research; T. Pearson, NSERC funding (Canada), and the Canadian Institutes for Health Research (grant MOP 81267).
Expert Testimony: None declared.
Patents: M. Razavi, provisional patent application number 61720386; N.L. Anderson, US 7,632,686 and provisional patent application number 61720386; T. Pearson, provisional patent application number 61720386.
Other: N.L. Anderson, Agilent Technologies.
Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.
- Received for publication November 29, 2012.
- Accepted for publication July 1, 2013.
- © 2013 The American Association for Clinical Chemistry