Strategies to verify equimolar peptide release in mass spectrometry-based protein quantification exemplified for apolipoprotein(a)

Lipoprotein(a) enrichment from serum samples

All solvents were liquid chromatography (LC) MS grade and the reagents used were of the highest available purity. Biotinylation of anti-apo(a) antibodies (Merck or Abcam) was performed using 100 µL of antibody solution (60 µg/100 µL). Streptavidin-coated magnetic beads were used for immobilization of anti-apo(a) antibody (Cat. Nr. 65801, DynaBeads™, Streptavidin Trial Kit, Invitrogen™, ThermoFisher, New York, NY). Two different serum samples (200 µL each) were used for enrichment, namely sample L8 ([apo(a)]=263 nmol/L, 17 kringles) and sample L3 ([apo(a)]=17 nmol/L, 29 kringles) (see Supplemental Table 1).

Protein digestion and N-glycan release

Captured Lp(a) was denatured and disulfide bonds were reduced with 5 mM tris(2-carboxyethyl) phosphine in ammoniumbicarbonate buffer at 56 °C for 30 min. The reduced cysteine residues were alkylated with 5 mM iodoacetamide at room temperature in the dark for 30 min. For optimal digestion, samples were incubated with 0.4 µg of Lys-C for 1 h followed by with 0.4 µg trypsin at 37 °C, pH=8 for 3 h in a total volume of 35 µL [29], [30], [31]. For non-optimal digestion, samples were incubated with 0.4 µg of trypsin at 20 °C, pH=8 for 15 min. In both cases proteolysis was quenched with 5 μL 40 % (v/v) methanol/4 % (v/v) formic acid in milliQ water to obtain pH=3. Enzymatic N-glycan release was performed by adding 0.4 mU PNGase F (Roche) and overnight incubation at 37 °C. Prior to LC-MS/MS analysis each protein digest was purified using Oasis HLB 1cc Vac Cartridge (Waters, 30 mg Sorbent per Cartridge, 30 µm). The solid-phase extraction was performed with two washes (0.1 % formic acid (FA) aqueous solution) and an elution with water/acetonitrile/FA 70/30/0.1 [v/v/v]). After lyophilization the digest was redissolved in 30 μL of 0.1 % FA aqueous solution.

LC–MS/MS analysis

Protein digests were analyzed (after SPE) by online C18 nanoHPLC MS/MS with a system consisting of an Easy-nanoLC 1,200 gradient HPLC system (Thermo Fisher Scientific) and an Orbitrap Fusion™ Lumos™ Tribrid™ mass spectrometer (Thermo). Samples (2 μL) were injected onto a homemade precolumn (100 μm × 15 mm; Reprosil-Pur C18-AQ 1.9 μm; Dr Maisch) and eluted on a homemade analytical nano-HPLC column (30 cm × 75 μm; Reprosil-Pur C18-AQ 1.9 μm). The gradient was run from 2 % to 40 % solvent B (water/acetonitrile/FA 20/80/0.1 [v/v/v]) in 30 min. The nanoHPLC column was drawn to a tip of 5 μm, which acted as the electrospray needle of the MS source. The Lumos mass spectrometer was operated in data-dependent MS/MS mode for a cycle time of 3 s, with an HCD normalized collision energy of 32 % and recording of the MS2 spectrum in the Orbitrap. In the master scan (MS1), the resolving power was 120,000, the scan range was m/z 400 to 1,500, at an automatic gain control target of 400,000 at maximum fill time of 50 ms. Dynamic exclusion was set after n=1 with exclusion duration of 10s. Charge states 1 to 5 were included. For MS2, precursors were isolated with the quadrupole with an isolation width of 1.2 Da. The MS2 scan resolution was 30,000 with an automatic gain control in standard mode at a maximum fill time of 60 ms.

LC-MS/MS data analysis

MS data were imported in XCalibur Qual Browser (version 2.2; Thermo Fisher Scientific) for manual interpretation and annotation of the spectra. For database searches, XCalibur raw files were converted into peak lists using Proteome Discoverer 2.4.0.305 (Thermo Fisher Scientific) and searched against the human proteome using the Mascot search algorithm (version 2.2.07; Matrix Science). The MS tolerance was set at 10 ppm, and the MS/MS tolerance was set at 20 milli mass units. Trypsin was selected as enzyme (allowing two missed cleavages), and carbamidomethylation was set as fixed modification on cysteine residues and oxidation (methionine) as a variable modification. A target false discovery rate of 0.01 at the peptide level (based on a target-decoy analysis) was used.

LC–MRM-MS analysis

Apolipoprotein(a) measurements were performed within the standard operating procedure for sample preparation and tryptic digestion of serum apolipoprotein panel (including apoA-I and B-100) as published previously [30]. Multiple rection monitoring (MRM) MS measurements were performed on a 1,290 ultraperformance liquid chromatography system coupled to a 6495A triple–quadrupole mass spectrometer equipped with electrospray ionization (Agilent Technologies, Santa Clara, CA) operating in positive ion mode. Peptide separations were performed on a Zorbax SB-C18 column 2.1 × 50 mm, 1.8 µm (Agilent Technologies, Santa Clara, CA) and the column oven was set at 70 °C. Mobile phase A consisted of 5 % (v/v) methanol and 0.05 % (v/v) formic acid in water and mobile phase B of 95 % (v/v) methanol and 0.05 % (v/v) formic acid in water. The LC-system was operated at constant flow of 0.3 mL/min with a non-linear gradient profile over 10 min followed by a isocratic recondition step of 1 min prior to system re-equilibration. The collision energy (CE) was optimized per transition to obtain a maximum signal intensity.

SIL-peptides, SIL-isotopologues and MRM transitions

All MRM transitions with regard to quantification of apo(a), apoB-100 and apo-AI have been reported previously [29]. Four SIL peptide isotopologues (“SIL-1”, “SIL-2”, “SIL-3” and “SIL-4”) for each different quantifier peptides (LFLEP, GISST and TPENY) were synthesized and added into nine different in-house standards L3-L11 (containing various apo(a)-concentrations with various numbers of kringle repeats as indicated in Supplemental Table 1) to a concentration of 10, 20, 40 and 100 nmol/L, respectively. In a similar manner these isotopologues were added to eight different routine serum samples that previously demonstrated >15 % differences between immunoturbidimetric assay (ITA) and MS-based apo(a) quantification. Transitions of SIL-peptides (“SIL-1”, “SIL-2”, “SIL-3” and “SIL-4”) are summarized in Supplemental Figure S1. In order to setup transitions for additional peptides, namely apo(a)-peptides that contain a missed cleavage, Expasy and Protein Prospector were used (https://web.expasy.org/peptide_mass/) (https://prospector.ucsf.edu/prospector/mshome.htm).

For verification of previously reported post-translational modifications the UniProt database was used for P08519 (https://www.uniprot.org).

General MS-based strategies to verify equimolar release of quantifier peptides

Detailed knowledge on the digestion characteristics of a protein into proteotypic peptides is crucial when aiming for accurate protein quantifications. Comprehensive databases such as NextProt and ProteinAtlas provide a valuable source of previously observed digestion products of a specific protein, nevertheless due to lab-to-lab variations CLSI-C64 recommends “experimental verification” on site [8]. A workflow to address digestion outcome in an exploratory survey manner is proposed in Figure 1.

Figure 1:

MS-based strategies to collect evidence for equimolar proteolysis. Survey peptide identifications include mapping fully cleaved peptides as well as miscleaved peptides in samples that are subjected to either short or optimal digestion times. In addition, the occurrence of ragged peptides with unexpected termini is verified. Targeted peptide quantifications includes the use of SIL-peptides for the quantifier and qualifier peptides and monitor specific miscleaved peptides.

In a survey all proteolytic peptides are identified in two types of protein digests, namely one obtained using non-optimal conditions (for example short digestion times) and one from optimal conditions as defined in the RMP-conditions [30]. The first sample contains partially digested proteins and is suitable for detailed mapping of areas in the protein backbone that are not fully cleaved or unexpectedly proteolyzed. To this end, an untargeted data-dependent or data-independent MS-analysis of the complex digest is performed on a high resolution MS system to identify ragged and miscleaved peptides. In the same survey analysis regular (expected) proteotypic peptides are identified and compared to previously reported peptides in single/multiple reaction monitoring (SRM/MRM) databases (in silico) [32], [33], [34]. Next, peptide candidates are selected for further evaluation of their performance using a targeted liquid chromatography (LC) MRM-MS set-up [9]. From the combined results from databases and survey analysis a set of candidate peptides is selected and monitored in digestion time courses to differentiate between slow and quickly forming peptides and determine peptide stability after release from the corresponding protein backbone [8], 9], 18], 35]. Furthermore the digestion efficiency and reproducibility are used to select the most optimal surrogate peptide(s). Although thus obtained results are highly informative, it is noted that a digestion plateau does not warrant full conversion of the protein into peptides. Commonly, interpeptide disagreement points towards issues on digestion efficiency provided that two or more quantifier peptides from one protein are available, but excellent agreement does not necessarily imply equimolar conversion. The absence of ragged and miscleaved peptides suggests equimolar release of corresponding target peptides, provided it is demonstrated that such species potentially can be detected in a specific MRM set-up. To this end specific miscleaved peptides are synthesized and used to develop and optimize MRM transitions. For specific purposes these transitions can be included in a targeted strategy to detect potential protein digestion anomalies. In addition, stable isotope-labeled (SIL) peptides (with heavy R or K incorporated) are used for normalization purposes [36], 37].

Survey evaluation of apo(a) proteolysis

The commercially available immunoassay-based Lp(a) tests exhibit large between-test result variations, which impedes meta-analysis of epidemiological research and pharmaceutical trials and stresses the importance of standardization with a new RMP [27]. It is postulated that one Lp(a) particle (present in the circulation) contains one apo(a) and one apoB100 molecule [38], 39]. Apo(a) contains various N-and O-glycosylation sites and is furthermore characterized by a size-polymorphism due to inheritance of different apo(a) alleles with varying numbers of K-IV₂ (kringle) repeats (Figure 2 exemplifies six repeats) [22], 40], 41]. The total number of N-glycosylation sites partly depends on kringle repeats (grey arrows in Figure 2), resulting in an overall molecular weight of apo(a) ranging from 250 to 800 kDa. Early studies of apo(a) glycopeptides have revealed that O-glycosylation is clustered in the kringle K-IV₂ linker domains [40], 42]. At that time it was already noted that “the factors influencing apo(a) proteolysis are uncertain” and that O-glycosylation could play a role in restricting proteolytic cleavages of the interkringle linkers, stressing the importance of a careful assessment of digestion efficiency [40].

Figure 2:

Schematic of apo(a) with varying numbers of K-IV₂ repeats and inherent size-polymorphisms. The multiple N-glycosylation sites are indicated with grey arrows, of which the site-occupancy and structure analysis have been overviewed [41]. The three quantifier peptides used for LC-MRM-MS are depicted with their abbreviated sequences. Three different scenarios are illustrated that start with a first cleavage at three different positions in the intact protein. In an optimal situation all scenarios result in the same end-point with release of three target peptides in an equimolar manner. In practice the scenarios may result in a non-equimolar outcome due to loss of one of the partially digested protein (chemical or physical instability) on the way to the end-points.

In Figure 2 three apo(a) quantifier peptides “TPENYPNAGLTR”, “GISSTTVTGR” and “LFLEPTQADIALLK” are further abbreviated as TPENY, GISST and LFLEP [29]. The selection of these peptides for quantification purposes is based on previous studies in our Leiden Apolipoprotein Reference Laboratory that included evaluation of the digestion time courses [9], 30]. Alternatively, apolipoproteins can be measured by MS in their intact form (top-down approach), but the quantitative performance of this approach is still in its infancy [43]. Aiming to collect additional evidence on the completeness of apo(a) digestion three different scenarios are plotted with digestion trajectories that all lead to these three quantifiers (i.e. theoretical end-products of the digestion). It is hypothesized that these scenarios may exhibit different digestion efficiencies due to formation of various large “intermediate polypeptides” that can complex with other intermediate structures (proteins) that are present in a complex serum digest. On top of this, apo(a) digestion scenarios inherently differ due to the earlier explained size-polymorphisms. The outcomes were further studied by performing a survey analysis on eight different serum digests that were obtained after Lp(a)-enrichment using two different anti-Lp(a) antibodies of two different serum samples (“L8” and “L3”) to yield samples L8M, L8A, L3M and L3A (Supplemental Table 1, Lp(a)-enrichment was necessary for in-depth identification of all apo(a) peptides). These samples were subjected to short or optimal digestion conditions. In order to study potential steric hindrance from the presence of multiple N-glycans a third set of samples was analyzed after N-deglycosylation. The survey analysis was used to focus on three quantifier peptides TPENY, GISST and LFLEP. In a previous study the digestion kinetics of these proteotypic peptides of apo(a) were reported and it was found that all three peptides readily reach a plateau [30]. Nevertheless, the appearance of a plateau does not necessarily mean that all protein is converted into peptide. Therefore, all apo(a)-peptides were evaluated in samples that were subjected to a short digestion, including those that contained a miscleavage. More than 95 % of peptide intensities (derived from peptide spectral matches) corresponded to “regular” apo(a) proteotypic peptides, and only two miscleaved peptides were found in proximity of the quantifier peptides (Supplemental Figure S2). MS/MS-data unequivocally identified these miscleaved peptides in the sequences of quantifier peptides LFLEP or GISST (Figure 3). Although the fully cleaved quantifying peptides indeed are the dominant species this also demonstrates that digestion at this part of the protein is not complete after 15 min.

Figure 3:

Identification of two apo(a) quantifier peptides that contain a miscleavage followed by targeted peptide quantification of synthetically prepared surrogate peptides. In a Survey two apo(a) peptides were detected when applying a short proteolysis (15 min), namely “LFLEP” (in yellow), and “GISST” (in blue), as well as two miscleaved peptides, namely SYRGISSTTVTGR and LFLEPTQADIALLKLSRPAVITDK. Quantifier peptide TPENY was identified solely as end-product upon both short and optimal digestion (not shown). The miscleaved surrogate peptides were used for optimization of transition collision energies (left and right corner) in LC-MRM-MS. Thus developed transitions can be included in a targeted peptide quantification workflow to identify proteolysis anomalies.

The miscleaved peptides SYRGISSTTVTGR and LFLEPTQADIALLKLSRPAVITDK were not observed in any of the samples that were digested according to the candidate RMP digestion protocol. Their absence suggests that digestion had been “completed” at that position in the protein. From this we conclude that, although the digestion of the measurand apo(a) may not be complete for the full part of the protein sequence, the conversion into the quantifying peptides is equimolar. The sequence coverage results from the survey analysis of apo(a) are summarized in Supplemental Figure S2. Finally, the presence of non-tryptic peptides (both on N- and C-termini) in the survey data was evaluated. Such peptides would interfere with accurate quantification of the corresponding protein in case their sequence would overlap with quantifying peptides. As an example all non-tryptic peptides (some with multiple peptide spectrum matches (PSMs)) that were detected in sample L8A are summarized in Supplemental Table 2. Importantly, just one “odd peptide” from apo(a) was detected, however this peptide “E₂₀QSHVVQDCYHGDGQSYR₃₇” is explained by expected protein processing (AA 1–19 is the signal peptide). This demonstrates that “odd peptides” could be detected and identified in the database search. The data from the other apolipoproteins also demonstrate that if odd cleavages would occur these indeed are detected.

Targeted evaluation of apo(a) proteolysis

In quantitative protein MS workflows a SIL-peptide is added for each quantifier (and qualifier) peptide to correct for ionization differences and ion suppression [8], 12]. The use of SIL peptides enables a careful evaluation of various sample preparation conditions and the “behavior” of different samples (such as lipemic vs. regular serum), as well as the acquisition of a precise digestion time course. Furthermore, for the example of apo(a), two miscleaved peptides that were found in the survey analysis were synthesized that contain quantifier peptide sequences with one elongation, namely SYRGISSTTVTGR and LFLEPTQADIALLKLSRPAVITDK. These peptides were used to develop and optimize MRM-transitions that were used to verify the presence of miscleaved quantifier peptides. It is noted that a SIL-peptide does not correct for sample-to-sample variations with regard to protein extraction or digestion efficiency. Intuitively, the optimal internal standard (IS) would be a properly folded (equivalent) full-length isotope-labeled protein of interest that would exhibit the same biological interactions as the native protein in matrix. Such an IS could control for all potential sources of error in the assay, especially in the case where proteins are first enriched from the matrix before quantification by MS. Unfortunately, such standards are often not available and as an alternative winged or concatenated peptides have been proposed, albeit that are not identical to the protein of interest [44]. In the current study we have evaluated the application of multiple SIL-peptides, namely isotopologues with various molecular masses, to gain a qualitative insight into proteolysis with regard to linearity and potential concentration biases [36], 37]. Here it is emphasized that these isotopologues are only used for a sub-set of samples and that their application does not replace the traditional external calibration curves. Four SIL peptide isotopologues (“SIL-1”, “SIL-2”, “SIL-3” and “SIL-4”) of three different quantifier peptides (LFLEP, GISST and TPENY) were added into nine different in-house standards L3-L11 (containing various apo(a)-concentrations with various numbers of kringle repeats as indicated in Supplemental Table 1) to a concentration of 10, 20, 40 and 100 nmol/L, respectively. Regression analysis (R²-values >0.99) demonstrated that MRM-derived peak areas were linear at the applied concentration range, indicating that the isotopologues are perfectly suited for normalization of signal intensities (Supplemental Figure S3). Within one sample the three slopes of the three quantifier peptides differed from each other, explained from the fact that all peptides exhibit different MS response factors. When comparing samples L3-L11 it was observed that the slopes of a certain quantifier were sample-specific and it was hypothesized that the addition of SIL isotopologues into other samples could be used to pinpoint specific aberrances. To further test this hypothesis we spiked the four isotopologues of three different apo(a)-quantifier peptides into serum samples that previously demonstrated >15 % differences between immunoturbidometric assay (ITA) and MS-based apo(a) quantification. It is noted that serum Lp(a) levels are routinely quantified using commercially available immunoassays such as the ITA and we have compared Lp(a)-ITA and MS-based apo(a) quantifications in hundreds of serum samples and reported that the majority of clinical specimens were in a good agreement, whereas for some individual specimens marked discordances (>15 %) were observed [27]. Interestingly, excellent linearities of the different isotopologues were observed in the specimens that differed in Lp(a)-ITA and MS-based apo(a) concentrations (Figure 4). It is clear that the slopes differed from sample-to sample as was observed for samples L3-L11, but so far no specific correlations are seen with regard to ITA-lower or ITA-higher apo(a) concentrations when comparing to MS.

Figure 4:

Application of four isotopologues of three different apo(a)-quantifier peptides that were spiked into serum samples that previously demonstrated >15 % differences between the ITA- and MS-based apo(a) quantification (see Supplemental Table 1). For clarity reasons the MRM-MS peak areas that are used as an intensity measure are not depicted on the y-axis. On the left-hand side the ITA-quantity was higher than MS, on the right-hand side the MS-quantity was higher than ITA.