Selection of proteotypic peptides

The first step in the development of a an PRM-MS based analysis is the selection of a subset of peptides to be used as quantitative representatives ‘Proteotypic peptides’ for each candidate protein [29, 30]. The identification of optimal proteotypic peptides representing the targeted proteins is very important for the accurate quantification of the target proteins using the targeted proteomics approach. ‘Proteotypic peptides’ must be sequence specific, detectable and the most responsive peptides. In this study, proteotypic peptides were selected using both computational and experimental strategies. The enhanced signature peptide predictor [29] was used to predict the high responding peptides for the cTnI protein. As a result of the analysis, seven peptides with the highest enhanced signature peptide predictor prediction factor were selected. These peptides are as follows, with the prediction factor from highest to lowest; NIDALSGMEGR, NITEIADLTQK, ISADAMMQALLGAR, TLLLQIAK, MADGSDAAR, EPRPAPAPIR and AYATEPHAK. The computational approach was also confirmed experimentally. Fully tryptic peptides of cTnI were screened using bottom-up proteomics. After trypsin digestion, the cTnI peptide solution was analyzed using with a Nano UPLC-Q-Exactive HF-X Orbitrap mass spectrometer with full MS/dd-MS2 (TopN) mode. Raw files were processed by Proteome Discoverer 2.5 (Thermo) using the Sequest algorithm. Advanced parameters of the Sequest algorithm were set as 10 ppm precursor mass tolerance, 0.02 Da fragment mass tolerance, static modifications of carbamidomethyl, dynamic modifications of acetyl and oxidation. Responses of tryptic cTnI peptides screened for PRM analysis are given in Table 1. Trypsin digestion hydrolyses the primary sequence of the target protein into specific peptide ending in C-terminal lysine or arginine residues. Both abundance and the number of peptide-spectrum match (PSMs) in the table aim to represent the response of the corresponding peptide, providing important insights into its relative abundance and confidence of identification. To assess potential phosphorylation sites of cTnI, we utilized the NetPhos-3.1 software, a generic phosphorylation site prediction tool for eukaryotic proteins. The results from NetPhos-3.1 indicated that threonine residues at positions 51, 124, and 129 were not predicted to be phosphorylated in cTnI. Based on this information, we selected two signature peptides, NITEIADLTQK and TLLLQIAK, for quantification. The primary selection criteria for these peptides were their high peptide-response and the fact that they have no known post-translational modifications, including phosphorylation [23, 31]. ISADAMMQALLGAR and NIDALSGMEGR gave high signal responses but were not selected due to their possible post-translational modifications. NITEIADLTQK and TLLLQIAK were chosen as surrogate peptides for the quantification of cTnI using ID-LC–MS/MS.

Table 1 Responses of tryptic cTnI peptides screened for PRM analysis and assay developmentImmunoaffinity enrichment

In serum proteomics analysis, the presence of high-abundance proteins can significantly suppress the signal of low-abundance proteins, making it extremely difficult to analyze proteins with concentrations, such as cTnI in the low μg/L range [32]. In this study, immunoaffinity enrichment at the protein level was performed using a protein-based internal standard (NIST SRM 2921) to cope with the complexity and dynamic range of the serum to quantify cTnI by LC–MS/MS. This internal standard underwent the same sample processing as the endogenous cTnI, allowing us to assess potential problems such as incomplete protein extraction, incomplete proteolysis, or material loss during the sample preparation steps.

By using magnetic beads smaller than 1 μm in diameter, a higher capture efficiency can be achieved due to the higher surface area-to-volume ratio [13]. Therefore, the magnetic nanoparticles with particle diameters of 130 nm and the surface functionalities COOH for the covalent binding of the antibody was selected for immobilization of anti-cTnI antibody.

The ID-LC–MS/MS method presented by Schneck et al. was used to directly quantify and measure the amount of antibody bound to magnetic nanoparticles [13]. Figure 2A–D illustrates the methodology used to quantifying the antibody bound to magnetic particles using the ID-LC–MS/MS technique. Briefly, matrix-matched external calibrants were prepared by combining the antibody and internal standard in a manner that mimicked the sample matrix. The heavy isotope labeled synthetic DLPSPIER peptides were purchased as internal standards because of their uniqueness to the constant region of the monoclonal antibody (mAb), robust MS/MS signal intensity and absence of amino acid residues prone to chemical modification, such as cysteine or methionine residues. These calibrants were used to establish a calibration curve for accurate quantification. Following antibody immobilization, the antibody-magnetic particle conjugates were extensively washed to remove non-specifically adsorbed antibodies. The immobilized antibodies were then subjected to in situ trypsin digestion together with the magnetic particles. Internal standard was added to the digested samples, which were then analyzed directly by LC–MS/MS. Quantification was performed by monitoring a specific transition for DLSPIER peptides. To determine the maximum loading capacity of the magnetic nanoparticles, different antibody concentrations were used while keeping the amount of nanoparticles constant. In this study, approximately 100, 200, and 400 μg of antibody were added to the immobilization solutions containing 1 mg of beads. It was observed that the maximum surface loading of antibodies onto the magnetic nanoparticles was achieved at a ratio of approximately 100 μg antibody per mg of nanoparticles. In addition, further increases in antibody loading resulted in a decreased in efficiency. When immobilizing of antibodies on the surface of nanoparticles by physical adsorption, it is crucial to select the optimal antibody concentration for each specific case, which is not necessarily the highest concentration available [33]. This suggests that at high antibody concentrations, a portion of the nanoparticle surface that is not accessible for antibody binding due to steric hindrance. As shown in Fig. 2C, the highest amount of immobilized antibody was achieved by conjugating 1 mg of nanoparticles with 100 μg of antibody (59.2 ± 5.7 μg/mg). To calculate the amount required for cTnI enrichment from 1 mL of serum using the synthesized nanoparticle (NP)-antibody conjugate, enrichment was performed using 5, 10, 20, and 30 μL of NP-antibody conjugate. As shown in Fig. 2, the amount of peptide measured reaches saturation after the use of 10 μL of NP-antibody conjugate in serum cTnI enrichment. Therefore, the use of 10 μL of conjugate appears to be sufficient to capture all of the cTnI in 1 mL of serum for the analysis. The remaining part of the study was performed using these optimized parameters.

Fig. 2

A Extracted ion chromatograms (EIC) for both DLSPIER and IS-DLSPIER peptides. B The Calibration Curve, C bar graph depicting the amount of immobilized anti-cTnI antibody per milligram of nanoparticles. The standard deviation error bars indicate the variability between duplicate preparations of the conjugates. D Bar graph depicting the amount of NP-antibody conjugate required for 1 mL serum enrichment

Quantification of cTnI in human serum

Monitoring more than one proteotypic peptide as a quantifier increases confidence in the accuracy of protein analysis in matrix-based assays. Considering the wide range of cTnI modifications and the variability of cTnI isoforms among different patients’ plasma, the measurement of multiple peptides provides greater confidence and accuracy [34]. Therefore, the proteotypic peptides NITEIADLTQK and TLLLQIAK were selected for cTnI quantification and simultaneously monitored by PRM. The collision energies for each peptide were optimized experimentally optimized by PRM monitoring. The optimized parameters were utilized to observe the fragmentation transitions of the two cTnI peptides (NITEIADLTQK and TLLLQIAK) and a single antibody peptide (DLPSPIER) in both labeled and unlabeled forms. Table 2 summarizes the PRM transitions of selected peptides for cTnI measurement by the nanoparticle enrichment method using the intact protein internal standard (NIST SRM 2921).

Table 2 PRM transitions, amino acid sequence and retention time for the detection of the peptides

The PRM MS/MS spectrum of the selected cTnI proteolytic peptides and the corresponding extracted ion chromatograms are shown in Fig. 3. The peak areas of the selected proteotypic peptides were used to establish a calibration curve based on IDMS and quantitative analysis.

Fig. 3

PRM analysis of tryptic digests of cTnI spiked serum. A MS/MS spectrum of targeted precursor ion TLLLQIAK2+ (selected product ions y5+ and y6+ for quantification in red); B MS/MS spectrum of targeted precursor ion NITEIADLTQK2+ (selected product ions y6+ and y9+ for quantification in blue); C EICs obtained when quantifying cTnI-free human serum spiked with a cTnI at LOQ level (1.8 µg/L) and SIL-cTnI at 15 µg/L showing coelution of the TLLLQIAK peptide and its isotopically labeled counterpart; D EICs obtained when quantifying cTnI-free human serum spiked with a cTnI at LOQ level (1.8 µg/L) and SIL-cTnI at 15 µg/L showing coelution of the NITEIADLTQK peptide and its isotopically labeled counterpart. Precursor ions were isolated using an isolation window of 0.6 m/z

This study utilized NIST SRM 2921, a protein calibrant, along with its labeled internal standard, to establish a calibration curve. Endogenous cTnI from patient serum samples and SIL-cTnI were isolated by immunoaffinity enrichment strategy and subsequently analyzed by ID-MS. There are different forms of cTnI in blood such as triple complex (cTnC-cTnT-cTnI), binary complex (cTnI-cTnC), free form, proteolyzed form, heparin bound form, phosphorylated form, oxidized and reduced forms. Furthermore, cTnI can be degraded both in vivo and in vitro [35]. Due to this variability and heterogeneity, there is no perfect matrix-matched protein calibrant for LC–MS quantification strategy. Nevertheless, SRM 2921 has been purified from human heart tissue and because of its heterogeneity is currently the best choice for purpose [26].

Method validationLinear range

The method was validated using the NIST SRM 2921 as calibrant. To assess the linearity of the method response, the integrated peak area ratios of cTnI peptide transitions (unlabeled to labeled) were plotted against the mass ratios of SRM 2921. The quantities used for the plot ranged from 0.7 to 24 μg/L, with specific concentrations of 0.7, 1.2, 3.5, 6, 12 and 24 μg/L. The peak areas were automatically integrated by computer using the Quan Browser software program (Thermo Scientific). The method response was found to be linear for between 0.7 μg/L and 23.3 μg/L of cTnI with a regression coefficient of 0.996 and 0.992 for the peptides TLLLQIAK and NITEIADLTQK respectively according to the equations presented in Table 3. The LOD was calculated as defined by ICH [28] using the formula LOD = 3 × Sa/b, where b is the slope of the calibration curve and Sa is the standard deviation of the intercept. The LOD values were found to be 0.6 ng and 1.6 ng for cTnI using the peptides TLLLQIAK and NITEIADLTQK for quantification, respectively. The LOQ was also determined according to the defined criteria, LOQ = 10 × Sa/b [28], and was found to be 1.8 ng and 4.8 ng for cTnI using the peptides TLLLQIAK and NITEIADLTQK for quantification, respectively.

Table 3 Linear range of method response by surrogate peptides TLLLQIAK and NITEIADLTQKAccuracy

The accuracy is defined under precision and trueness of the test results. For precision of the measurement system, the repeatability and intermediate precision of the method were evaluated by intra-day (analysis of QC solutions in replicates of three in the same day) and inter-day (in replicates of three on four different days) assay variance, respectively. At the QC concentration levels evaluated, the repeatability (intra-run) and intermediate precision (inter-run) of the acceptance standards should be less than 10% (percent RSD). RSDRepeatability (1) and RSDintermediate precision (2) were calculated using the formulas given below. For QCs, the values obtained for repeatability and intermediate precision were less than 10% RSD (Table 4). These values met the acceptance requirements, indicating that the current method has sufficient precision.

$$RSD_ = \frac } }} }} \times 100,$$

(1)

$$RSD_ = \frac - MS_ } }} }} \times 100.$$

(2)

Table 4 The precision results of different cTnI concentrations

The trueness assessment was performed with the recovery evaluated using NIST SRM 2921. Three QC levels 2.0 (low), 5.0 (medium), and 10.0 (high) μg/L were assessed for each peptide. The obtained recovery data for TLLLQIAK was 104.2%, 97.3% and 94.8% for 2, 5 and 10 μg/L cTnI levels respectively. The obtained recovery data for NITEIADLTQK was 104.5%, 99.9% and 106.1% for 2, 5 and 10 μg/L cTnI levels respectively. Accordingly, % recovery values were calculated and presented in Table 5. Those values met the recovery acceptability requirements, indicating that the current approach was adequate in terms of trueness.

Table 5 The trueness results of different cTnI concentrationsCarryover

In the present method, a solution of acetonitrile/water (80/20; v/v) with 0.1% formic acid was used to rinse the syringe and injection port. This wash procedure was performed several times before and after each injection. Under these washing conditions, the signal observed at the retention time of each peptide (area below the peak) was less than 1% compared to that found in the LOD after injection of a blank sample.

Evaluation of measurement uncertainty

The uncertainty of the method was evaluated according to EURACHEM/CITAC Guide CG 4 (third edition) entitled “Quantifying Uncertainty in Analytical Measurement” [24, 25]. Uncertainty sources of arising from operations steps such as pipetting, dilution, balances and volumetric equipment are covered by the reproducibility, intermediate precision and recovery uncertainties. The sources of uncertainty identified for the present method are the following defined parameters: balance, the repeatability standard deviation sr (3), the contribution of the grouping factor to the total variation (sbetween) (4) the uncertainty of the calibration curve (5). The formula of the standard combined uncertainty (6) given below. The expanded uncertainty has been calculated considering a coverage factor of 2 for a confidence of approximately 95%. The breakdown of the uncertainty budget is presented in Table 6.

$$s_ = \sqrt - MS_ }} }}} ,$$

(4)

$$u_ = \frac \times \sqrt }} + \frac }} + \frac - x_ } \right)^ }} - x_ } \right)^ }}} ,$$

(5)

$$u_ = k\sqrt }^}} + \frac}^}} + }^ + }^} .$$

(6)

Table 6 Breakdown of the uncertainty budgetApplication to clinical samples

To evaluate the results of the developed reference method, the candidate reference method was applied to patient serum samples. Although we have almost mimicked the patient sample by adding the human cTn complex to human cTnI free serum, it is still important to quantify the developed method on clinical samples considering variety of cTnI modification and unique cTn profile of individuals. Unfortunately, in all samples collected (n = 25), the levels of cTnI were found to be below the LOQ of our method. From all the collected patient samples, the four samples which cTnI values are above the detection limit were selected and analyzed by the developed ID-LC–MS/MS method.

Table 7 presents the results obtained from both the ID-LC–MS/MS method and the Siemens Atellica® Solution immunoassay. The recovery and uncertainty was not calculated because the cTnI values of the selected samples, as detected by the immunoassay, were below the quantification limit of the developed ID-MS method. Based on the results of the validation study, we predicted that these samples could be detected but might not provide fully accurate quantitative results. The experimental results confirmed our prediction. However, the fact that the developed methodology performed similarly on the patient samples demonstrates the reliability of the method. This is further supported by the high recovery values observed for the QC samples prepared at concentrations of 2, 5, and 12 µg/L, as shown in Table 5.

Table 7 Application to clinical samples

In future developments, it is important to improve the developed method by achieving a lower LOQ. Additionally, a larger number of samples should be utilized to assess the applicability of the method to clinical samples.