Quantification of serum apolipoproteins A-I and B-100 in clinical samples using an automated SISCAPA–MALDI-TOF-MS workflow

Methods 81 (2015) 74–85

Contents lists available at ScienceDirect

Methods journal homepage: www.elsevier.com/locate/ymeth

Quantification of serum apolipoproteins A-I and B-100 in clinical samples using an automated SISCAPA–MALDI-TOF-MS workflow Irene van den Broek a,⇑, Jan Nouta a, Morteza Razavi d, Richard Yip d, Marco R. Bladergroen b, Fred P.H.T.M. Romijn a, Nico P.M. Smit a, Oliver Drews c, Rainer Paape c, Detlev Suckau c, André M. Deelder b, Yuri E.M. van der Burgt b, Terry W. Pearson d, N. Leigh Anderson d, Christa M. Cobbaert a a

Department of Clinical Chemistry and Laboratory Medicine, Leiden University Medical Center (LUMC), Albinusdreef 2, 2333 ZA Leiden, The Netherlands Center for Proteomics and Metabolomics, Leiden University Medical Center (LUMC), Albinusdreef 2, 2333 ZA Leiden, The Netherlands c Bruker Daltonics GmbH, Fahrenheitstraße 4, 28359 Bremen, Germany d SISCAPA Assay Technologies Inc., Box 53309, Washington, DC 20009, USA b

a r t i c l e

i n f o

Article history: Received 29 November 2014 Received in revised form 1 March 2015 Accepted 2 March 2015 Available online 9 March 2015 Keywords: Clinical proteomics Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) Stable-isotope standards and capture by anti-peptide antibodies (SISCAPA) Cardiovascular disease risk assessment Analytical method validation

a b s t r a c t A fully automated workflow was developed and validated for simultaneous quantification of the cardiovascular disease risk markers apolipoproteins A-I (apoA-I) and B-100 (apoB-100) in clinical sera. By coupling of stable-isotope standards and capture by anti-peptide antibodies (SISCAPA) for enrichment of proteotypic peptides from serum digests to matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) MS detection, the standardized platform enabled rapid, liquid chromatography-free quantification at a relatively high throughput of 96 samples in 12 h. The average imprecision in normo- and triglyceridemic serum pools was 3.8% for apoA-I and 4.2% for apoB-100 (4 replicates over 5 days). If stored properly, the MALDI target containing enriched apoA-1 and apoB-100 peptides could be re-analyzed without any effect on bias or imprecision for at least 7 days after initial analysis. Validation of the workflow revealed excellent linearity for daily calibration with external, serum-based calibrators (R2 of 0.984 for apoA-I and 0.976 for apoB-100 as average over five days), and absence of matrix effects or interference from triglycerides, protein content, hemolysates, or bilirubins. Quantification of apoA-I in 93 normo- and hypertriglyceridemic clinical sera showed good agreement with immunoturbidimetric analysis (slope = 1.01, R2 = 0.95, mean bias = 4.0%). Measurement of apoB-100 in the same clinical sera using both methods, however, revealed several outliers in SISCAPA–MALDI-TOF-MS measurements, possibly as a result of the lower MALDI-TOF-MS signal intensity (slope = 1.09, R2 = 0.91, mean bias = 2.0%). The combination of analytical performance, rapid cycle time and automation potential validate the SISCAPA–MALDI-TOF-MS platform as a valuable approach for standardized and high-throughput quantification of apoA-I and apoB-100 in large sample cohorts. Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction The potential of mass spectrometry (MS) for the standardization of protein assays in the clinical laboratory has been acknowledged since the first publication in 1996 from Barr and co-workers, demonstrating the absolute quantification of apolipoprotein A-I (apoA-I) based on MS detection of an enzymatically derived, ⇑ Corresponding author at: Department of Clinical Chemistry and Laboratory Medicine, Leiden University Medical Center (LUMC), P.O. Box 9600, 2300 RC Leiden, The Netherlands. Tel.: +31 71 526 6257. E-mail address: [email protected] (I. van den Broek). http://dx.doi.org/10.1016/j.ymeth.2015.03.001 1046-2023/Ó 2015 Elsevier Inc. All rights reserved.

specific, signature peptide relative to a stable-isotope labelled (SIL) peptide analog [1]. Since then, MS has been increasingly recognized as an alternative or complementary technique to overcome many of the interference issues associated with current, immunoassay-based, clinical protein assays [2–4]. In addition, the multiplexing capacity of MS allows simultaneous quantification of multiple proteins in the same assay and can provide more-detailed information on a patient’s health status than the often single-analyte immunoassays. Since the introduction of MS-based (bottom-up) protein assays, serum apolipoproteins have been particularly attractive target molecules in various single- or multiplex quantitative approaches

I. van den Broek et al. / Methods 81 (2015) 74–85

[5–12]. As important mediators of lipoprotein (and cholesterol) metabolism, apolipoproteins play a crucial role in the processes that lead to cardiovascular disease (CVD), the number one cause of death worldwide (www.who.int). Whereas apoA-I reflects the number of anti-atherogenic high-density lipoprotein (HDL) particles, apolipoprotein B-100 (apoB-100) reflects the total number of atherogenic particles, including low-, intermediate low-, and very low-density lipoproteins (LDL, IDL, and VLDL, respectively). The ratio of apoB-100 to apoA-I is a well-established marker for CVD risk and has been proposed to improve current CVD risk assessment based on LDL and HDL cholesterol (LDLc and HDLc) [13]. Despite separate, well-standardized immunoassays for apoA-I and apoB-100, the apoB-100/apoA-I ratio has not been adopted for general CVD risk assessment. MS-based multiplexed apolipoprotein quantification may allow a new approach to apolipoprotein measurement in clinical practice. Several clinical applications of MS-based bottom-up proteomics for the simultaneous quantification of apoA-I and apoB-100 in serum or plasma have been described [5–8]. These methods involve manual sample processing (i.e., denaturation, reduction, alkylation, and digestion) before manual or on-line desalting followed by liquid chromatographic (LC) separation and MS/MS detection in selected reaction monitoring (SRM) mode. Although the time-consuming digestion step (typically performed overnight) limits sample turnaround time, the major bottleneck for robust and high-throughput quantification by the current MS-based approaches is manual processing, which is laborious and a main source for bias and imprecision [14]. In addition, the LC–MS/MS runtimes (between 7 and 18 min per sample in the references listed above) can seriously limit sample throughput for large-scale analyses. To streamline the analytical workflow for adoption of MS-based protein assays into routine clinical testing, we have developed a platform for automated sample preparation in 96-well format that combines trypsin digestion with selective peptide enrichment and eliminates the need for LCseparation. Stable-isotope standards and capture by anti-peptide antibodies (SISCAPA) is an approach that specifically enriches proteotypic signature peptides derived after enzymatic digestion of the plasma or serum sample [15]. As a consequence, the peptide analytes are of such high purity that they can be quantified by MS without prior LC separation. Examples of LC-free quantification of SISCAPAenriched peptides include ultrafast, SRM-MS analysis using the RapidFire™ system that combines solid-phase extraction and electrospray ionization (ESI) [16], and time-of-flight (TOF)-MS using matrix-assisted laser desorption/ionization (MALDI) [17]. Although MALDI-TOF-MS has not been widely used for absolute quantification of protein or peptide analytes, stable isotope labelled peptide internal standards can be used to normalize the MALDI-TOF-MS response and have shown to provide highprecision peptide measurements [18,19]. In addition, MALDITOF-MS instruments are easily operated and routinely used in many clinical (microbiology) laboratories for the diagnosis of infectious diseases [20], offering a widely available platform that can be used for protein measurement. Moreover, other (semi-) automated quantitative methods have been reported that couple immunoaffinity extraction of the target peptide or protein to MALDI-TOF-MS in so-called immuno-MALDI (iMALDI) [21,22] or mass spectrometric immunoassays (MSIA) [23,24]. In the present study, we have investigated the suitability of an automated SISCAPA–MALDI-TOF-MS workflow for multiplexed quantification of apoA-I and apoB-100 in routine clinical chemistry. Besides automation and further streamlining of the analytical workflow, our work has placed specific focus on demonstrated and well-controlled analytical quality; all essential for adoption of MSbased protein assays in routine clinical testing.

75

2. Materials and methods 2.1. Serum samples Five normotriglyceridemic sera (NTG1 (LOT 2009.0361), NTG2 (LOT 2009.0362), NTG3 (LOT 2009.0363), NTG4 (LOT 2013.2061), and NTG5 (LOT 2013.2062)) and three hypertriglyceridemic sera (HTG1 (LOT 2013.2063), HTG2 (LOT 2013.2064), and HTG3 (LOT 2013.2065)) were produced in a certified production laboratory (MCA labs, Winterswijk, The Netherlands) according to clinical laboratory and standards institute (CLSI) guideline C37-A [25], and obtained from the Dutch Foundation for Quality Assessment in Medical Laboratories (SKML, Nijmegen, The Netherlands). Upon arrival, all serum samples were thawed once, dispensed into single-use 50 lL portions and stored at 80 °C until further use. The five normotriglyceridemic sera (NTG) were value-assigned for apoA-I and apoB-100 by the Northwest Lipid Metabolism and Diabetes Research Laboratories (Seattle, WA, US) [26,27] and used for external calibration. The three hypertriglyceridemic sera (HTG) were equally mixed to provide an HTGpool. In addition, an NTGpool was prepared from left-over sera from ten random patients with triglyceride and total cholesterol concentrations below 2 and 5 mmol/L, respectively. For method comparison, left-over patient sera with cholesterol levels < 10 mmol/L were collected with either normal (<2.3 mmol/L; n = 49) or high (2.3–20 mmol/L; n = 44) triglyceride concentrations. All left-over sera were collected within 24 h of blood collection and were fully-anonymized. Aliquots of 60 lL were prepared in 0.5 mL polypropylene microtubes (Sarstedt, Nümbrecht, Germany) and stored at 80 °C until analysis.

2.2. Preparation of samples and reagents All solvents and reagents were of LC–MS grade or the highest analytical grade available, unless otherwise specified. Stable isotope-labelled (SIL) peptides ATEHLSTLSEK (apoA-I) and FPEVDVLTK (apoB-100) were synthesized by New England Peptides Inc. (Gardner, MA) with 13C615N2 stable-isotope labelled lysine (K). Selection of these two peptides was based on earlier literature for SRM-based protein quantification [12]. The SIL-peptides were obtained as lyophilized powder with a certificate for >95% purity and peptide content, as determined by HPLC and amino acid analysis, respectively. The SIL-peptides were reconstituted in 30% (v/v) acetonitrile (ACN; Biosolve, Valkenswaard, The Netherlands) with 0.1% (v/v) formic acid (Sigma Aldrich, Zwijndrecht, The Netherlands) at a concentration of 30 lmol/L, and stored at 80 °C. On each measurement day, the SIL-peptide stock solutions were combined and diluted with 0.2 mol/L Tris (Trizma preset crystals pH 8.1, Sigma–Aldrich) to obtain a working solution of 8 lmol/L for ATEHLSTLSEK and 0.2 lmol/L for FPEVDVLTK. High-affinity rabbit monoclonal antibodies specific for the selected signature peptides ATEHLSTLSEK and FPEVDVLTK were provided by SISCAPA Assay Technologies (SAT, Washington DC) at a concentration of 0.25 and 0.50 g/L, respectively. The methodology for production, purification, and selection of monoclonal antibodies has been previously described [28]. The anti-peptide antibodies were aliquoted in 25 lg portions and stored at 20 °C. DynabeadsÒ Protein G-coated magnetic beads (2.8 lm, Novex-Life Technologies, Bleiswijk, The Netherlands) were washed four times by vigorous mixing with 0.03% (w/v) 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS; Sigma Aldrich) in phosphate-buffered saline (PBS; Braun, Melsungen, Germany), and subsequently incubated with the

76

I. van den Broek et al. / Methods 81 (2015) 74–85

anti-peptide antibodies (150 lg beads and 2  0.5 lg antibody per sample) for 1 h at room temperature under gentle rotation. The antibody-bead complex was washed once with 2 mL 0.03% CHAPS in PBS to remove unbound antibody and thereafter suspended in the same solution and stored at 4 °C for a maximum of one day. One day before performing the SISCAPA–MALDI workflow, a 96well polypropylene round-bottom microplate (Greiner Bio-One, Alphen a/d Rijn, The Netherlands) was prepared as digestion plate. Therefore, 17 lL of a mixture of 9 mol/L urea (Sigma–Aldrich) and 0.05 mol/L tris(2-carboxyethyl)phosphine (TCEP; Thermo Scientific, Rockford, IL) in 0.2 mol/L Tris was added to each well before overnight evaporation in a dry incubator at 37 °C. The MALDI matrix solution was freshly prepared on each measurement day by mixing 50 lL 10% (v/v) trifluoroacetic acid (TFA; Sigma Aldrich), 50 lL 100 mmol/L ammonium phosphate (Sigma Aldrich), and 4788 lL ACN/H2O/TFA (90/10/0.1 v/v/v) with 112 lL of a saturated a-cyano-4-hydroxycinnamic acid (HCCA; Bruker Daltonics GmbH, Bremen, Germany) solution in ACN/H2O/TFA (90/10/0.1 v/v/v). Serum samples were allowed to thaw at room temperature, homogenized, and placed in a Hamilton MicrolabÒSTAR liquid handling workstation (Hamilton, Bonaduz, Switzerland), equipped with eight individually controlled 300 lL pipetting channels. Ten lL amounts of serum were transferred individually from the 60 lL aliquots to pre-defined locations in a 96-well PCR plate (polypropylene, Greiner Bio-One), hereafter named sample plate. The sample plate was designed to have the calibrators, patient sera, and samples for method validation evenly distributed among all columns and rows. 2.3. Digestion procedure A Hamilton MicrolabÒSTAR and a Hamilton MicrolabÒSTARplus, connected via a Hamilton MicrolabÒ SWAP, were used for fullyautomated sample preparation (Fig. 1). The Hamilton MicrolabÒSTAR (left robot) was equipped with eight 300 lL pipetting channels, whereas the Hamilton MicrolabÒSTARplus (right robot)

was equipped with eight 1000 lL pipetting channels and a 300 lL 96 channel pipetting head. A detailed description of the general set-up of the robotic system has been reported previously [29]. The automated procedure started on the right robot by addition of 30 lL Tris (pH 8.1) to the sample plate with the 96-channel pipetting head. After homogenisation, 10 lL of the diluted serum were transferred to the digestion plate, and the digestion plate was mixed for 5 min at 1500 RPM to provide a final concentration of 9 mol/L urea, 0.05 mol/L TCEP, and 0.2 mol/L Tris. The digestion plate was then transferred to the left robot, where the plate was covered by a dark, heat conducting, aluminium lid, followed by incubation for 30 min at 25 °C. After denaturation and reduction, 10 lL 25 mmol/L iodoacetamide (Sigma–Aldrich) were added, and the plate was covered before 10 min incubation at 25 °C in the dark. After alkylation, the urea concentration was diluted to 1 mol/L by addition of 105 lL Tris and 10 lL SIL-peptide working solution, followed by addition of 10 lL 0.925 mg/mL L-(tosylamido 2-phenyl) ethyl chloromethyl ketone (TPCK) treated trypsin (code TRTPCK, Worthington Biochemical Corp., Lakewood, NJ) in 10 mmol/L HCl (Merck, Darmstadt, Germany) from a thermostatically cooled storage rack, which was kept at 4 °C. Trypsin digestion was performed at 37 °C for 3 h with an approximate trypsin-toprotein ratio of 1:20 (w/w), before the reaction was quenched by addition of 10 lL 0.075 mmol/L tosyl-L-lysine chloromethyl ketone (TLCK; Fluka Biochemica, Buchs, Switzerland) in 10 mmol/L HCl. 2.4. SISCAPA procedure The digestion plate, containing the peptide digest and the SILpeptide internal standards, was back-transferred to the right robot. Here, the magnetic bead-antibody complexes, stored at 4 °C in a specially designed storage rack, were re-suspended, and 10 lL were added to the digestion plate to provide 0.5 lg of each antibody per well. The digestion plate was placed on the shaker for 1 h at 1500 RPM at 25 °C to allow capture of the endogenous and SILpeptides from apoA-I and apoB-100. After incubation, the digestion plate was placed on an in-house developed magnetic plate with 96

Fig. 1. Overview of the Hamilton robotic liquid handling platform and the SISCAPA–MALDI-TOF-MS workflow for quantification of apoA-I and apoB-100. Digestion of the serum samples was performed on the left robot in 96-well format (A). After digestion, the plate was transferred to the right robot using a Hamilton MicrolabÒ SWAP (B). Protein G magnetic beads coated with anti-peptide antibodies were added from a specifically designed cooled storage rack (C). After SISCAPA enrichment, the elution solvent was directly spotted on the MALDI target plate (D) and analyzed by MALDI-TOF-MS (E).

I. van den Broek et al. / Methods 81 (2015) 74–85

individual magnets positioned at the bottom of each well. The remaining digestion solution was discarded and the bead-antibody-peptide complexes were washed three times with 150 lL 0.03% (w/v) CHAPS in 25 mmol/L ammonium phosphate. The fourth wash was performed with 150 lL of an equal (v/v) mixture of ACN and 25 mmol/L ammonium phosphate. During all washes, the bead-antibody-peptide solution was vigorously shaken for 1 min at 1500 RPM. Finally, peptides were eluted by addition of 25 lL 10% (v/v) ACN in 0.1% (v/v) formic acid and 5 min shaking at 1500 RPM, before transfer to a 96 well PCR elution plate. 2.5. MALDI-TOF-MS From each well of the elution plate, 2 lL eluate were spotted in quadruplicate on a MTP 384 AnchorChip 800 lm MALDI target plate (Bruker Daltonics). The droplets were allowed to dry for at least 1 h at room temperature. The HCCA matrix solution was stored on-deck in a special container with automated sliding lids. Just before use, the matrix solution was transferred to an open 384-well polypropylene PCR plate (Greiner Bio-one), named as matrix plate, and 0.5 lL was spotted in quadruplicate on top of the dried droplets by the 96 channel pipetting head. An ultrafleXtreme (Bruker Daltonics) MALDI-TOF instrument was used for MS analysis. Data acquisition in the 500–3500 Da mass range was performed in the positive-ion reflector mode with a laser frequency of 2000 Hz at 200 shots/raster until a total of 20,000 shots was acquired. FlexAnalysis version 3.4 software (Bruker Daltonics) was used to assign peaks and to determine the area underneath the complete isotopic envelope with the SNAP algorithm. Data analysis for quantitative purposes was performed with the MALDI ImmunoTyper Software (Bruker Daltonics), using the area of the complete isotopic envelope from endogenous ATEHLSTLSEK (m/z 1215) and the corresponding SILpeptide (m/z 1223) for quantification of apoA-I, and the area of the complete isotopic envelope from endogenous (m/z 1047) and SIL (m/z 1055) FPEVDVLTK for quantification of apoB-100. Before each experiment, the MALDI target plate was washed by a standard washing procedure, consisting of a short wash with 80% 2-propanol followed by incubation in 50% ACN (for at least 1 h) and a final rinse with methanol. The same MALDI target plate was used for all analyses. 2.6. Calibration and quantification The average relative response of the four replicate spots was used for each individual sample. Outliers as a result of irreproducible spots were automatically detected and excluded, based on the median absolute deviation (MAD). Data-analysis for quantification of apoA-I and apoB-100 was performed separately, because the MALDI ImmunoTyper Software allows targeted dataextraction from only one m/z-value for the endogenous peptide and one m/z-value for the SIL-peptide. The ‘‘calibration’’ routine was used to assign the known protein concentrations to the location of the calibration samples on the target plate, import the flexAnalysis data, and construct the calibration line based on linear regression analysis without weighting. The ‘‘quantitation’’ option was used to import the unknown sample list to assign the sample names to the location on the target plate and calculate the protein concentration based on the linear regression equation provided by the calibration results. 2.7. Method comparison with immunoturbidimetric analysis Serum apoA-I and apoB-100 test results produced by the automated SISCAPA–MALDI-TOF-MS workflow were compared to test results derived from clinical immunoturbidimetric analysis (ITA)

77

on a Cobas Integra 800 (Roche Diagnostics, Mannheim, Germ any), according to CLSI guideline EP09-A3 [30]. Both procedures were calibrated with the same, value-assigned sera (NTG1, NTG2, NTG3, NTG4, and NTG5) to guarantee traceability to WHO-IFCC reference standards SP1-01 for apoA-I and SP3-07 for apoB-100. Quantification by SISCAPA–MALDI-TOF-MS was performed by duplicate sample preparations of all patient sera (49 normotriglyceridemic (NTG) and 44 hypertriglyceridemic (HTG)), divided over five days. Quantification with ITA was performed on a single day by duplicate measurement of aliquots from the same clinical sera, as described previously [14]. The same five value-assigned NTG sera were used for daily calibration of SISCAPA–MALDI-TOF-MS and ITA quantification. EP-evaluator software, release 10 (Data Innovations LLC, South Burlington, VT) was used for Deming regression analysis to provide slopes, intercepts, and determination coefficients (R2). 2.8. Method validation Method validation of the automated SISCAPA–MALDI-TOF-MS workflow for quantification of serum apoA-I and apoB-100 was based on guidance documents from the CLSI [31,32] and European medicines agency (EMA) [33], and on best-practice recommendations for validation of MS assays for protein biomarkers [34]. When necessary, the validation guidelines were adapted to fit to the application of SISCAPA (e.g., selectivity) or MALDITOF-MS (e.g., carry-over and re-analysis). 2.8.1. Imprecision (EP-15) Within-run (wr) and between-day (dd) imprecision were determined according to CLSI guideline EP15-A2 [31]. Therefore, the NTGpool and HTGpool were individually prepared in quadruplicate (n = 4) on five different days, and apoA-I and apoB-100 concentrations were calculated based on daily calibration with value-assigned sera. The within-run imprecision (% CVwr) was expressed as the average standard deviation (sd) relative to the grand mean qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi _ over five days (% CVwr ¼ ððsd2 þ sd2 þ sd2 þ sd2 þ sd2 Þ=5Þ= (X) 1 2 3 4 5 X_  100%). The between-day imprecision (% CVdd) was expressed qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 as % CVdd ¼ ðB2  sdwr =nÞ=X_  100%, with B as the standard deviation of the daily mean, and n = 4 for the number of replicate sample preparations on one day. The total imprecision was calcuqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi lated by the sum of squares (% CVtot ¼ ð% CV2wr þ % CV2dd Þ). 2.8.2. Matrix-effects and interference testing A five-point mixing scheme was used to evaluate the presence of matrix effects on the quantification of apoA-I and apoB-100 in serum with increasing levels of triglycerides, or increasing ratios of apoB100 to apoA-I. The evaluation included a serum pool (Pool 1) with relatively low apoB-100/apoA-I ratios (average 0.43) and triglyceride levels (average 1.3 mmol/L), a serum pool (Pool 2) with relatively high apoB-100/apoA-I ratios (average 1.7) and triglyceride levels (average 9.0 mmol/L), and 3:1, 1:1, and 1:3 (v/v) admixtures of both pools. All sera were prepared in triplicate, and the experiment was performed twice, once on each of two different days. The recovery of apoA-I and apoB-100 was calculated by comparison of the calculated concentrations in the admixtures with the expected concentrations based on linear extrapolation of the average concentrations in Pool 1 and Pool 2. The same samples were also used to determine the linearity in relative MS response between the five-points, and, as a consequence, the presence of matrix-effects (e.g., from triglycerides) on the relative MS response. The effects of common clinical interferences on the quantification of apoA-I and apoB-100 by the automated

78

I. van den Broek et al. / Methods 81 (2015) 74–85

SISCAPA–MALDI-TOF-MS workflow were evaluated with the use of an ASSURANCE™ Interference test kit (Sun Diagnostics, New Gloucester, ME), following the manufacturer’s protocol based on CLSI guideline EP7-A [32]. The concentrations of the five highly concentrated potential interferents in the kit, as well as their concentrations in NTGpool, were determined by routine clinical measurements. Each of the five potential interferents was diluted 20-fold with NTGpool, except for the highly-concentrated protein solution (from albumin and gamma globulins) that was diluted twofold with NTGpool, to provide (1) lipemic serum with a total concentration of 12.6 mmol/L triglycerides (2) haemolytic serum with a total concentration of 0.16 mmol/L hemolysate, (3) hyperproteinemic serum with a total protein concentration of 150 g/L, (4) icteric serum with a total concentration of 0.18 mmol/L conjugated bilirubin, and (5) icteric serum with a total concentration of 267 lmol/L unconjugated bilirubin. The influence of the clinical interferences was expressed as the % deviation of the apoA-I and apoB-100 concentration determined in the sera spiked with interferents (adjusted for the dilution factor) relative to the average apoA-I and apoB-100 concentration determined in NTGpool during the five days of imprecision evaluation _ A bias within ±15% was defined as accept(i.e., the grand mean or X). able [34]. All five spiked sera were prepared in triplicate, and the experiment was performed twice, once on each of two different days. 2.8.3. Selectivity and carry-over The selective capture of the apoA-I and apoB-100 peptides by the anti-peptide antibodies was evaluated by an identical sample preparation of all 93 clinical sera using ‘‘blank’’ Protein G magnetic beads, i.e., not coated with anti-peptide antibodies. In addition, the potential for carry-over after the standard washing procedure was evaluated by spotting of blank elution- and matrix solution on the cleaned MALDI-target plate (from day-3 and day-5) and re-analysis by MALDI-TOF-MS. In addition, the MALDI-target plate from day-1 was re-analysed after cleaning and application of blank matrixsolution only. The effect of non-selectivity or carry-over on the quantitative results was evaluated by comparison of the obtained peak intensity, signal-to-noise (S/N) ratio, and quality factor of any detectable signal around the m/z-values of the endogenous or SIL-peptides from apoA-I or apoB-100 with the average or minimal peak parameters from the normally measured clinical samples. Furthermore, the MALDI-TOF-MS spectra were inspected for nonspecific signals around the appropriate m/z-values of the endogenous and SIL-peptides from apoA-I and apoB-100, followed by an accurate mass analysis at ultrahigh resolution using a Bruker 15 tesla solariX XR™ Fourier transform ion cyclotron resonance (FTICR) mass spectrometer equipped with a CombiSource (in MALDI-mode) and a ParaCell™ (Bruker Daltonics). Because of the large variation in absolute MALDI signal intensity, the typical acceptance criteria for selectivity and carry-over of bioanalytical LC–MS/MS methods (i.e., the response should be less than 20% of the lower limit of quantification for the analyte and less than 5% of the internal standard) [33] were slightly adapted for SISCAPA– MALDI-TOF-MS application. Selectivity and carry-over were accepted when the S/N ratio of any detectable peak (S/N > 3) was below 20% of the minimum S/N ratio for 99% of all measured clinical samples, representing a typical clinical measurement range with regard to a 1% probability for low-signal outliers. 2.8.4. Re-analysis The MALDI target plates from days-3, -4, and -5 were re-analyzed after, respectively, one, three, or seven days storage under dark conditions at room temperature. The relative responses after re-analysis of NTGpool and HTGpool were compared with the relative

responses after initial measurement of the same spots. In addition, the apoA-I and apoB-100 concentrations in the re-analyzed NTGpool and HTGpool were calculated based on calibration with re-analyzed calibration samples from the same plate. The % bias after re-analy_ whereas sis was calculated by comparison to the grand mean (X), the % CV of the re-analyzed samples was compared to the % CV for the initial measurements. 3. Results 3.1. Optimization of MALDI-TOF-MS 3.1.1. Selectivity of MALDI-TOF-MS spectra During method development, a selective wash with 50% ACN appeared crucial for removal of background interference in the isotopic profile of endogenous ATEHLSTLSEK (at m/z-value 1216). Moreover, although the addition of a detergent, i.e., 0.03% CHAPS, aided the homogeneous resuspension of the magnetic beads, remaining CHAPS suppressed the analyte signals during MALDITOF measurements, and the fourth (final) wash was, as a compromise, performed without detergent. A representative MS spectrum of a clinical sample obtained by the SISCAPA–MALDI-TOF-MS workflow is shown in Fig. 2A, showing peaks for endogenous and SIL-peptides from apoB-100 and apoA-I at m/z-values 1047, 1055, 1215, and 1223, respectively. The intense signal at m/z-value 568.2 results from HCCA-matrix. Other signals are observed at m/z-values 945.5, 1312.6, 1623.8, 1467.9, and 1924.9. Based on accurate mass measurements on an FTICR-system, the signals at m/z-values 1623.8 and 1467.9 were assigned as tryptic peptides from serum albumin (SwissProt P02768) that were non-specifically bound to the magnetic beads (i.e., peptides DVFLGMFLYEYAR and RHPDYSVVLLLR at m/z-values 1623.7824 ± 0.9 ppm and 1467.839 ± 0.7 ppm, respectively). Such albumin signals are in agreement with a previous report [18]. The identity of the signals at m/z-values at 945.5, 1312.6, and 1925.1 (m/z-values in the MALDI–FTICR-analysis at 1312.5845 ± 0.7 ppm and 1924.9547 ± 0.9 ppm) is not known. Interestingly, the peak at m/z 945.5 did not appear in the spectra obtained from sera prepared without anti-peptide antibodies (Fig. 2B), suggesting specific or non-specific binding to one or both antibodies rather than non-specific binding to the magnetic beads. Nonetheless, none of these common peaks interfered with the m/z-values of the target peptides from apoA-I and apoB-100. In addition, no interfering peaks were observed for samples prepared according to the same protocol without anti-peptide antibodies coated to the magnetic beads (an example is shown in Fig. 2B). 3.1.2. Quality and reproducibility of MALDI-TOF-MS spectra The presence of 10% ACN in the elution solvent and a spotting volume of 2 lL provided optimal reproducibility and quality of the spots. The acquisition of 200 shots per raster spot, scanning across a sample area until the maximum number of shots was reached, did not show any depletion of the analyte signal in either the raster spot or the sample area and was, therefore, applied to provide optimal signal intensity and S/N ratio for the analyte peaks. As an example, after measurement of quadruplicate spots from duplicate sample preparations of 93 clinical sera (n = 680 for apoA-I and 646 for apoB-100), the median S/N ratio [inter quartile range, IQR] for endogenous ATEHSLTLSEK, SILATEHSLTLSEK, endogenous FPEVDVLTK, and SIL-FPEVDVLTK were 375 [234–677], 290 [180–507], 267 [132–611], and 170 [88–362], respectively. In addition, the median quality factor (a parameter provided by the flexAnalysis software to describe the quality of a picked peak as judged by the intensity and the consistency with

79

Intens. [a.u.]

I. van den Broek et al. / Methods 81 (2015) 74–85

0.0 x10 4

568.284

1215.688

A

1.5 1047.637 1.0 1312.654

Intens. [a.u.]

0.5

945.490

0.0

1467.919

568.216

B

1467.918

4000 3000 1312.651

2000 1000 0

600

800

0.0 x10 4 1.25

1000

1200

FPEVDVLTK 1047.637

1400 0.0 x10 4

1600

1800

1215.688

m/z

ATEHLSTLSEK

1.5

1.00 1.0

0.75 1055.650

0.50

0.5

0.25 0.00

0.0

400

400

300

300

200

200

100

100

0

1045

1050

1055

1223.703

1060 m/z

0 1210

1215

1220

1225

m/z

Fig. 2. Representative MALDI-TOF-MS spectra obtained after automated sample preparation of a selected clinical sample with (A) and without (B) antibodies against ATEHLSTLSEK (apoA-I) and FPEVDVLTK (apoB-100). The inlets show the same spectra zoomed in around the m/z for endogenous (unlabelled) and SIL-FPEVDVLTK (left, in green) and ATEHSTLSEK (right, in blue). The peaks at m/z 568.2 correspond to an HCCA-matrix cluster, whereas m/z-values 1623.8 and 1467.9 correspond to non-specifically bound tryptic peptides from serum albumin (DVFLGMFLYEYAR and RHPDYSVVLLLR, respectively), see text for more detail.

its known isotopic patterns) was 10907 [IQR 8030–23536] for endogenous ATEHLSTLSEK, 5272 [IQR 4109–10174] for SILATEHLSTLSEK, 17345 [10678–38163] for endogenous FPEVDVL TK, and 6567 [IQR 4555–12607] for SIL-FPEVDVLTK, which indicates an intense signal and a very good match of the experimental with the expected isotopic pattern. It is furthermore worth mentioning that the amount of SILpeptide was specifically balanced to approximate a 1:1 ratio in endogenous-to-SIL-peptide responses within the normal serum reference ranges of apoA-I (1–2 g/L) and apoB-100 (0.5–1.5 g/L). The approximate 1:1 ratio between endogenous and SIL-peptide is critical for optimal measurement precision, particularly in the SISCAPA–MALDI-TOF workflow where the total amount of captured peptides is limited by the amount of antibodies, and where ionization efficiency depends on the ratio and interplay between all matrix components. As a result, the total amount of added SIL-peptides was determined at 80 pmol for SIL-ATEHLSTLSEK and 2 pmol for FPEVDVLTK.

3.2. Method validation 3.2.1. Daily calibration The average results for daily linear regression analysis after duplicate sample preparations of the five, value-assigned sera are shown in Table 1. It should be noted that the MALDI ImmunoTyper Software only allowed linear regression based on the average y-value (i.e., endogenous-to-SIL peptide ratio) from samples or replicates with the same concentration (x-value). The average of the duplicate prepared calibrators was, therefore, used as a single value, and outliers as a result of irreproducible spots were excluded manually by applying the MAD algorithm on the four quadruplicate results from a single sample preparation. In addition, the two calibration samples with an identical apoA-I concentration (NTG2 and NTG3, see Table 1) were ‘‘artificially’’ differentiated by defining a 0.000001 g/L concentration difference. Good correlation was observed between the average response ratio of the calibration samples and the value-assigned protein

80

I. van den Broek et al. / Methods 81 (2015) 74–85

Table 1 Average results for linearity and recovery in the daily calibration samples.

2

R Slope Intercept

ATEHLSTLSEK (apoA-I)

FPEVDVLTK (apoB-100)

0.984 ± 0.010 1.28 ± 0.31 0.36 ± 0.36

0.976 ± 0.033 1.97 ± 0.52 0.55 ± 0.41

Calibration sample

Target valuea (g/L)

Proteinb recovery (%)

Peptidec recovery (%)

Target valuea (g/L)

Proteinb recovery (%)

Peptidec recovery (%)

NTG1 NTG2 NTG3 NTG4 NTG5

1.15 1.60 1.60 2.15 1.34

101.3 ± 3.4 98.1 ± 2.4 104.5 ± 2.7 99.3 ± 3.4 98.9 ± 2.0

96.9 ± 6.5 101.0 ± 8.6 109.0 ± 10.0 108.1 ± 14.4 98.3 ± 4.9

0.76 0.80 1.29 0.69 1.39

101.7 ± 3.5 102.0 ± 2.7 105.0 ± 7.3 99.4 ± 3.1 96.1 ± 8.4

52.8 ± 4.3 54.5 ± 3.0 68.0 ± 12.5 47.7 ± 2.6 62.1 ± 10.2

a

Value-assigned and traceable to WHO-IFCC standards for apoA-I and apoB-100. Average of all daily protein recoveries ± standard deviation (n = 10), expressed as ((peak areaA/peak areaS)  intercept)/(slope * TV) * 100%. Average of the daily peptide recovery, expressed as ((peak area/peak areaS) * lmolS * MW)/(V(lL) * TV) * 100%. ‘A’ represents the analyte and ‘S’ the internal standard; ‘lmols’ is the added amount of internal standard (80 * 106 lmol for SIL-ATEHLSTLSEK and 2 * 106 lmol for SILFPEVDVTLK); ‘V’ is the volume of serum (2.5 lL); ‘MW’ is the molecular weight of apoA-I (30,779 Da) or apoB-100 (515,605 Da); ‘TV’ is the value-assigned target value in g/L. b

c

concentrations on each day of analysis (R2 between 0.967 and 0.992 for apoA-I and between 0.919 and 0.999 for apoB-100). For both proteins, a negative intercept with the y-axis was observed, whereas the variation in the slope was 24 % CV for apoA-I and 26 % CV for apoB-100 (n = 5 days). The variation in the slopes can be a result of either small differences in the amount of added SIL-peptide or the amount of formed endogenous peptides on the five different days. However, because the same IS-solution is added to the calibrators and patient sera, which are prepared under identical digestion and sample preparation conditions, daily calibration is likely to normalize for between-day variability in patient sera. A more detailed evaluation of within-run and between-day imprecision based on either endogenous-to-SIL peptide ratios or protein concentrations after daily calibration is described in Section 3.2.2. The % recovery of apoA-I and apoB-100 in the sera used for external calibration was calculated for each individual sample preparation by back-calculation of the measured concentration on the daily linear regression line. Comparison to the value-assigned target values for both apolipoproteins revealed protein recoveries between 94.8% and 108.3% for apoA-I and between 82.3% and 118.3% for apoB-100 (Table 1). The absolute peptide recovery could be additionally determined by comparison of the recovered concentration of native peptide (based on the relative response and the added amount of SIL-peptide) with the (molar) target values for apoA-I or apoB-100 in the five sera used for external calibration (Table 1). Because the SIL-peptides were added concurrent with digestion, a compensation for native peptide losses during digestion, SISCAPA, and spotting is expected. The average peptide recovery of 102.6% for ATEHLSTLSEK (apoA-I), therefore, suggests that this peptide is completely released after 3 h trypsin digestion. The average peptide recovery of 57.0% for FPEVDVLTK (apoB100), on the other hand, indicates that FPEVDVLTK is not fully-released during 3 h digestion, which is in agreement with the observed slow formation of this peptide under various denaturation conditions [5,6,35,36]. A similar conclusion could be drawn from the observed slopes of the daily calibration curve (Table 1); the average slope of 1.3 for apoA-I is just above the theoretically expected slope of 1.0 (based on a 81 pmol increase in endogenous peptide per 80 pmol of added SIL-peptide for complete digestion of 1 g/L apoA-I), whereas the average slope of 2.0 for apoB-100 was lower than theoretically expected (based on a 5 pmol increase in endogenous peptide per 2 pmol of added SIL-peptide for complete digestion of 1 g/L apoB-100). It should, nonetheless, for both examples be emphasized that the peptide recovery is only an approximation because the true (exact) content of the SIL-peptides in the working solution includes uncertainties from amino acid analysis as well as from resuspension of the lyophilized powder.

However, because quantification of apoA-I and apoB-100 is based on external calibration with value-assigned native sera, trueness of the test results relies on between-sample consistency in digestion recovery [36] and the absolute peptide recovery is only evaluated to identify potential points for further improvement of the assay (i.e., improved digestion). In this respect, the good linear correlation between relative MS response and apoA-I or apoB-100 concentrations (Table 1) indicates that the peptide recovery is consistent among all native, value-assigned, sera, allowing absolute protein quantification that is independent of digestion completeness. 3.2.2. Imprecision The within-run (n = 4), between-day (D = 5), and total imprecision for quantification of apoA-I and apoB-100 in NTGpool and HTGpool by SISCAPA–MALDI-TOF-MS are shown in Table 2. With an average total imprecision (% CV) of 3.8% and 4.2% for quantification of apoA-I and apoB-100, respectively, the automated SISCAPA–MALDI-TOF-MS workflow demonstrated excellent reproducibility within the generally accepted best-practice recommendations of 15% [34]. Moreover, the total workflow CV meets the

Table 2 The within-run (n = 4), between-day (5 days), and total imprecision of quantification of apoA-I and apoB-100 by automated SISCAPA–MALDI-TOF-MS in NTGpool and HTGpool. ATEHLSTLSEK (apoA-I)

FPEVDVLTK (apoB-100)

NTGpool

NTGpool

HTGpool

HTGpool

Validation of imprecision (4 replicates, 5 days) 1.28 1.45 X_ (g/L), n = 20

0.98

0.92

% CVwr % CVdd % CVtot

3.1 2.0 3.7

2.7 1.7 3.2

4.9 1.3 5.1

Re-analysis of day-3 after 24 h storage Mean (g/L), n = 4 1.28 % bias 0.1 % CVwr 3.5

1.47 +1.4 5.3

0.98 0.02 2.4

0.97 +5.0 6.5

Re-analysis of day-4 after 4 days storage Mean (g/L), n = 4 1.31 % bias +2.2 % CVwr 4.1.

1.48 +2.4 2.1

1.00 +1.9 4.5

0.90 2.5 4.2

1.47 +1.7 1.6

0.99 +1.1 3.6 4.1

0.94 +1.5 0.8

4.0 0.5 4.0

Re-analysis of day-5 after 7 days storage Mean (g/L), n = 4 1.28 % bias +0.2 % CVwr 4.3 Average re-analysis % CVwr 3.7

81

I. van den Broek et al. / Methods 81 (2015) 74–85

minimal requirements for imprecision of diagnostic tests based on intra-individual biological variation (i.e., <5.0% and <5.3% for apoA-I and apoB-100, respectively) [37]. Daily calibration with value-assigned serum-based calibrators appeared beneficial to normalize between-day variations in relative MS response, which had been noted by the relatively high %CV for the slope of the daily calibration curve. Even more illustratively, the average total imprecision in endogenous-to-SIL-peptide ratio was 7.3% and 10.8% for the peptides ATEHLSTLSEK (apoA-I) and FPEVDVLTK (apoB-100), respectively, with an average % CVdd of 5.7% for ATEHLSTLSEK and 8.5% for FPEVDVLTK. In other words, whereas the % CVdd contributed for 27% to 81% (CV2dd/CV2tot * 100%) to the total imprecision in response ratio, the contribution of % CVdd to the total imprecision in calculated concentrations after daily external calibration was 2–30%. 3.2.3. Matrix effects and interference testing Dilution of two serum pools, one with low triglyceride and apoB-100/apoA-I levels (Pool 1) and one with high triglyceride and apoB-100/apoA-I levels (Pool 2), provided excellent linearity between the five expected and measured concentration levels (R2 = 0.991 for apoA-I and R2 = 0.993 for apoB-100). In addition, comparison between the measured and expected concentrations showed a recovery between 94.8% and 97.9% for both proteins in all samples (Table 3). The good agreement between expected and measured concentrations suggest that the relative MS response obtained after SISCAPA–MALDI-TOF-MS is not affected by matrix effects resulting from apoA-I, apoB-100, or triglyceride concentrations. Moreover, the % bias for quantification of apoA-I and apoB-100 was within ±12% when NTGpool was spiked with highly-concentrated hemolysates, proteins, or bilirubins (Fig. 3). The accurate quantification of apoA-I and apoB-100 in the sera spiked with 116.5 g/L protein particularly demonstrated that the absolute

peptide recovery during trypsin digestion is not affected by a changed enzyme-to-protein ratio. Addition of 11.4 mmol/L triglycerides, nonetheless, resulted in a positive bias of 62% for quantification of apoB-100, which could be explained by the presence of apoB-100 contaminants from the VLDL- and chylomicron-derived triglycerides. The presence of apoB-100 in the concentrated triglyceride solution was confirmed by ITA measurement, revealing an apoB-100 concentration of 0.37 g/L after 20-fold dilution of the concentrated triglyceride solution. Moreover, comparison between ITA measurement of apoB-100 in NTGpool with and without spiked triglycerides showed the same positive bias (50%). Representative MALDI-TOF spectra after sample preparation of NTGpool with and without potentially interfering compounds are presented in Supplementary Fig. 1. As an example, although the spectra from the sera spiked with protein interferents showed relatively abundant signals for the peaks identified as nonspecifically bound peptides from human serum albumin (observed at m/z-values 1467.9 and 1623.8), the endogenous-to-SIL-peptide ratios in the serum spiked with proteins agreed with the expected ratio observed in the non-spiked serum (after correction for the twofold dilution factor, as shown in Supplementary Fig. 1D). 3.2.4. Selectivity and carry-over A typical MALDI-TOF-MS spectrum of a clinical serum sample after sample preparation without anti-peptide antibodies coated on the magnetic beads is shown in Fig. 2B. Only in a few samples, peaks with an S/N ratio above 3 were observed around the m/z-values of the endogenous or SIL-ATEHLSTLSEK (Fig. 4A and B), but all were below the acceptability limit and no effect on the quantitative results is expected. For SIL-FPEVDVLTK, on the other hand, several peaks with an S/N ratio above 3 were observed for the clinical samples that were prepared without anti-peptide antibodies. Moreover, all observed peaks had an S/N ratio above 20% of the minimum S/N ratio for 99% of the clinical sera (Fig. 4D). This might

Table 3 Recovery of apoA-I and apoB-100 in admixtures of sera with low and high triglyceride levels (n = 6 for each concentration level). Sample

FPEVDVLTK (apoB-100) Concentration (g/L)

Expected

Measured (n = 6)

% recovery

Expected

Measured (n = 6)

% recovery

– – 1.70 1.48 1.26

1.92 ± 0.12 1.03 ± 0.05 1.66 ± 0.09 1.41 ± 0.04 1.03 ± 0.05

– – 97.9 ± 5.3 95.4 ± 2.8 94.8 ± 5.6

– – 1.05 1.30 1.56

0.79 ± 0.05 1.81 ± 0.08 1.02 ± 0.03 1.24 ± 0.03 1.49 ± 0.06

– – 97.7 ± 2.8 95.3 ± 2.0 95.6 ± 4.0

apoB-100 concentraon (g/L)

1 2 1:Pool 2 = 3:1 (v/v) 1:Pool 2 = 1:1 (v/v) 1:Pool 2 = 1:3 (v/v)

apoA-I concentraon (g/L)

Pool Pool Pool Pool Pool

ATEHLSTLSEK (apoA-I) Concentration (g/L)

1.8 1.6 1.4 1.2 1.0 0.8

NTGpool

1

2

3

4

5

1.8 1.6 1.4 1.2 1.0 0.8

NTGpool

1

2

3

4

5

Fig. 3. A comparison between quantification of apoA-I (left) and apoB-100 (right) by SISCAPA–MALDI-TOF-MS in NTGpool and in NTGpool spiked with common interferents (1 = triglycerides; 2 = hemolysates; 3 = protein; 4 = conjugated bilirubin; 5 = unconjugated bilirubin). The concentration is presented as the average concentration (n = 20 for NTGpool and n = 6 for spiked sera), corrected for the dilution factor (⁄100/95 for 1,3,4,5, and ⁄2 for 2). The error bars indicate the 95% confidence interval. The dashed black and red lines indicate the average concentration in NTGpool (grand mean) and ±15% boundaries, respectively.

82

I. van den Broek et al. / Methods 81 (2015) 74–85

Fig. 4. The signal-to-noise (S/N) ratio for peaks from endogenous ATEHLSTLSEK (A), SIL-ATEHLSTLSEK (B), endogenous FPEVDVLTK (C), and SIL-FPEVDVLTK (D) for all SISCAPA–MALDI-TOF-MS measurements of 93 clinical sera (+). In addition, identified peaks (S/N > 3) are indicated that were observed after sample preparation of the same clinical sera without anti-peptide antibodies (, assessed on all days) or after re-analysis of the same spots after cleaning of the MALDI target plate ( , assessed on day-1, day3, and day-5). The red dashed line represents 20% of the minimum S/N ratio for 99% of the measured clinical sera and is set as the acceptance limit for selectivity and carryover.

indicate non-specific binding of either the SIL-peptide itself or a contaminant with an identical mass. On day five, nonetheless, no peaks were observed, suggesting that the observed contaminants are more likely a result of contaminations on the MALDI target plate than of specific contaminants from the SISCAPA enrichment. The possibility of carry-over from the MALDI target plate was further evaluated by re-analysis of the cleaned MALDI target plates. It was found that signals around the m/z-value of SILFPEVDVLTK with an S/N ratio above the acceptability limit (Fig. 4D) appeared for various spots from day-1 and day-3, and only for one spot from day-5. In addition, various peaks were observed for both endogenous and SIL-ATEHLSTLSEK after reanalysis of the cleaned MALDI target plate from day-3 and day-5 (Fig. 4A and B). It should be noted that the carry-over assessment on day-3 and day-5 included the application of blank elution solvent prior to spotting of the matrix, whereas the carry-over assessment on day-1 only included application of matrix on the washed MALDI target plate. The plates from day-3 and day-5 underwent the same washing and spotting procedure, except for differences in storage time before washing (i.e., the plates from day-3 and day-5 were stored for four and eight days, respectively). The higher frequency and intensity of the carry-over signal for the day-3 experiment compared to day-5 can, however, not be rationalized. The observed carry-over, despite an extensive washing procedure that consisted of consecutive washings with 80% 2-propanol, 1 h incubation with 50% ACN, and rinsing with methanol, nonetheless, indicate that a better understanding is needed on, for example, the effects of short and long storage of the spotted MALDI target plate or the difference between new versus older (more-intensively used) MALDI target plates. In addition, although an S/N ratio below the acceptability limit suggests only minor effect on the quantitative results, the intensity and clear detectability of the analyte peaks after washing of the MALDI target plate questions the

applicability of reusable MALDI target plates for clinical use. Clinical application of the presented SISCAPA–MALDI-TOF-MS workflow for quantification of apoA-I and apoB-100, therefore, requires improvement and adequate validation of the washing procedure for various conditions. Disposable MALDI-target plates might, on the other hand, be a more attractive alternative. 3.2.5. Re-analysis The deviation in average relative responses after re-analysis of spots from NTGpool and HTGpool after one, four, or seven days storage of the MALDI target plate was between 5.7 and 0.5%. In addition, the calculated apoA-I and apoB-100 concentrations in NTGpool and HTGpool were within ±2.4% and ±5.0% of the grand mean, respectively (Table 2). Moreover, the average % CVwr for the calculated apoA-I and apoB-100 concentrations in the four replicates after re-analysis (3.7% and 4.1%, respectively) was comparable to the % CVwr obtained with direct measurement (Table 2). The results indicate that spots can be re-analyzed by MALDI-TOF-MS without effect on bias or imprecision within, at least, seven days after completion of the sample preparation. For most accurate results, reanalysis should include the calibrators to correct for small shifts in response ratio. With the applied MALDI-TOF-MS settings, reanalysis of all 40 spots (quadruplicate spots from duplicate preparations of five calibration samples) with calibration samples takes no more than 15 min. 3.3. Method comparison between SISCAPA–MALDI-TOF-MS and immunoturbidimetry The scatter and Bland Altman difference plots for the 49 NTG and 44 HTG sera are illustrated in Fig. 5, and show good agreement between quantification of apoA-I with SISCAPA–MALDI-TOF-MS and ITA (Deming slope = 1.01 and R2 = 0.95). The three samples

83

I. van den Broek et al. / Methods 81 (2015) 74–85

apoA-I: Scaer plot

A

apoA-I: Bland Altman 0.4 Mean bias: 0.055 g/L or 4.01%

3

0.3 0.2 Bias (g/L)

SISCAPA-MALDI-TOF-MS (g/L)

2

R : 0.9549 Slope: 1.005 Intercept: 0.0493

2

0.1 -0.1 -0.2

1 NTG HTG

NTG HTG

-0.3

0

-0.4 0

1

2

3

0

ITA (g/L)

2

Mean bias: 0.021 g/L or 1.93% 0.4 0.2

2

Bias (g/L)

SISCAPA-MALDI-TOF-MS (g/L)

apoB-100: Bland Altman

0.0 -0.2

1

-0.4

NTG HTG 0

NTG HTG

-0.6 0

1

2

0

3

ITA (g/L)

2

3

Average ITA and SISCAPA-MALDI (g/L)

0.6

2

R : 0.9312 Slope: 0.994 Intercept: -0.020

2

1

apoB-100/apoA-I: Bland Altman

apoB-100/apoA-I: Scaer plot

C

Mean bias: -0.025 or -2.91%

0.4 0.2 Bias (..)

SISCAPA-MALDI-TOF-MS (..)

3

0.6

R : 0.9136 Slope: 1.090 Intercept: -0.0776

3

2

Average ITA and SISCAPA-MALDI (g/L)

apoB-100: Scaer plot

B

1

1

0.0 -0.2

NTG

-0.4

NTG HTG

HTG 0

-0.6 0

1

2

ITA (..)

0

1

2

Average ITA and SISCAPA-MALDI (..)

Fig. 5. Comparison between SISCAPA–MALDI-TOF-MS (y-axis) and immunoturbidimetric analysis (ITA, x-axis) for quantification of apoA-I (A), apoB-100 (B), and the ratio between apoB-100/apoA-I (C) in 49 normotriglyceridemic (in blue) and 44 hypertriglyceridemic (in red) patient samples. All individual results from duplicate measurements are plotted, except for three samples with apoA-I concentrations <0.5 g/L. In the scatter plots, the 1:1 and Deming regression line are represented by a black and red dashed line, respectively, with the total allowable error (TEa) as scatter plot bounds (i.e., 13.7% for apoA-I and 17.4% for apoB-100). Deming regression parameters are shown in the top-left corner. In the Bland Altman difference plots, the mean absolute bias is represented by a dashed black line and the values for the mean absolute and relative bias are shown in the top-left corner.

with a low apoA-I concentration (<0.5 g/L by ITA) showed a relatively high positive bias as a result of the negative intercept for the daily calibration line and were excluded from the method comparison. Only one sample had bias outside the TEa for both duplicates (Fig. 5A).

For quantification of apoB-100 with SISCAPA–MALDI-TOF-MS, however, agreement and correlation with ITA was lower (Deming slope = 1.09 and R2 = 0.91). Although the average imprecision for apoB-100 quantification in the pooled sera was as low as 4.2%, a relatively large deviation was observed between several duplicate

84

I. van den Broek et al. / Methods 81 (2015) 74–85

SISCAPA–MALDI-TOF-MS results of the clinical samples, including the sample with the highest apoB-100 concentration (Fig. 5B). The clinical sample with the lowest apoB-100 concentration was, on the other hand, particularly biased by the relatively large negative intercept of the daily calibration curve (1.2 on day-3). In addition, the apoB-100-to-apoA-I ratios obtained by the two different procedures are plotted in Fig. 5C, indicating a good agreement (Deming slope 0.994, R2 = 0.93) for the ratios ranging between 0.3 and 2.1. Because of the observed bias between the sera with apoA-I concentrations below 0.5 g/L, the resulting relatively high apoB/apoA-I ratios in these sera have been excluded from the comparison.

4. Conclusions A fully-automated, multiplexed SISCAPA–MALDI-TOF-MS workflow was developed and validated for simultaneous quantification of the cardiovascular disease risk markers apoA-I and apoB-100 in clinical sera. The robotic liquid handling platform allowed robust and standardized sample preparation, starting with 2.5 lL serum and including denaturation, reduction, alkylation, trypsin digestion, SISCAPA enrichment, and preparation of the MALDI target plate. The combination of highly-selective SISCAPA enrichment in 96-well format and rapid MALDI-TOF-MS data acquisition resulted in a total analysis time (from preparation of solutions and sampling of sera until completion of MALDI-TOF-MS analyses) of approximately 12 h for 96 samples. As a comparison, a previously developed antibody-free MS method for the same apolipoproteins requires LC run times of 20 min per sample, resulting in at least 32 h of LC–MS/MS analysis time for a full 96-well plate [8]. In addition to the speed of data acquisition (20 s per spot), MALDI-TOFMS offered targeted measurement of endogenous-to-SIL-peptide ratios with high precision (i.e., an average imprecision after a full EP-15 validation, including within-run and between-day imprecision of NTG and HTG pooled sera, below 5% for both apolipoproteins). Our results furthermore showed that apoA-I and apoB-100 can be re-analyzed without effect on bias or imprecision after at least seven days storage of the MALDI target plate. Besides rapid and precise measurements, MALDI-TOF-MS instruments are easily operated and, hence, facilitate adoption by routine analytical personnel, making the SISCAPA–MALDI-TOF-MS workflow particularly attractive for large-scale or routine clinical analyses. However, the sensitivity and selectivity of MALDI-TOF-MS is typically lower than LC-based SRM (MS/MS) approaches. As a comparison, the previously developed LC–MS/MS method demonstrated accurate quantification of apoA-I and apoB-100 after 1000-fold dilution of 0.5 lL serum without any prior enrichment [8]. MALDI-TOF-MS, on the other hand, requires a highly-selective sample enrichment procedure as shown in this study, and the costs of the anti-peptide antibodies might be a drawback for routine application, although this cost is offset to a great extent by the reduction in instrument time. In addition, despite the high-selectivity of the monoclonal anti-peptide antibodies, non-specific binding of other serum peptides occurred, which can either interfere with the analyte response or cause ion suppression of the MALDI signal of the target peptides. The observed carry-over in our study raises concerns about the applicability of reusable MALDI target plates in a clinical setting and demands the use of disposable target plates or improved and carefully validated washing procedures. Lastly, the linear dynamic range of MALDI-TOF-MS might complicate the application of SISCAPA–MALDI-TOF-MS to proteins with a wider concentration range than the 1-log dynamic range of the apolipoproteins, and the multiplexing capacity of SISCAPA– MALDI-TOF-MS may be limited by interferences in MALDI-TOFMS analysis. Nevertheless, the method will be very powerful for

quantitative, rapid, high throughput measurement of selected analytes. Validation of the SISCAPA–MALDI-TOF-MS workflow for quantification of apoA-I and apoB-100 demonstrated excellent linearity between apoA-I concentrations of 1–2 g/L and between apoB-100 concentrations of 0.8–1.8 g/L, which both reflect the relevant clinical reference ranges. Moreover, the absence of matrix effects or interference from triglycerides, protein content, hemolysates, or bilirubins demonstrated good selectivity of the SISCAPA–MALDITOF-MS workflow for apoA-I and apoB-100. The excellent agreement between quantification of apoA-I and apoB-100 in 93 clinical sera with either immunoturbidimetric analysis or SISCAPA– MALDI-TOF-MS clearly demonstrates the good performance within the relevant measurement range. The variation observed between duplicate SISCAPA–MALDI-TOF-MS measurements of apoB-100 in several clinical sera, however, is not acceptable for clinical use and requires refinement. One possibility is to use an increased amount of antibodies against the apoB-100 peptide FPEVDVLTK to increase the analyte signal and reduce the number of outliers as a result of low analyte signal. In addition, the average ratio between endogenous and SIL-FPEVDVLK of 1.7 indicates that the concentration of SIL-FPEVDVLTK can be increased to improve both signal intensity and precision in response ratio. Lastly, the digestion of apoB-100 can be optimized to increase peptide recovery and analyte signal. Although the here presented results on within-sample precision and between-sample linearity for quantification of apoB-100 with the SISCAPA–MALDI-TOF-MS workflow indicate only a small effect from digestion incompleteness on assay performance, it is anticipated that an improved peptide recovery will further improve the reproducibility and robustness of the workflow. Another possibility to improve peptide recovery is the selection of a different signature peptide with more efficient cleavage, as described in a previous report [36]. Another factor that can be optimized for future SISCAPA–MALDI-TOF-MS assays is the specific selection of a diagnostic signature peptide with a MALDI assay in mind. The current peptide-antibody systems have been particularly selected and developed for LC–ESI–SRM applications. However, peptide abundances can vary between MALDI and ESI, and a future assay that particularly considers the suitability of the signature peptide for MALDI can further increase the sensitivity and dynamic range of SISCAPA–MALDI-TOF-MS assays. In conclusion, the presented SISCAPA–MALDI-TOF-MS workflow for quantification of apoA-I and apoB-100 in clinical serum samples demonstrated accurate and precise measurement in agreement with clinical immunoturbidimetric analyses. Although the quantification of apoB-100 by SISCAPA–MALDI-TOF-MS needs further refinement to increase sensitivity and reproducibility, the SISCAPA–MALDI-TOF-MS workflow is a promising approach for accurate, robust, and high-throughput quantification of apoA-I and apoB-100 in large sample cohorts.

Acknowledgements The authors like to thank Simone Nicolardi and Hans Dalebout from the Center for Proteomics and Metabolomics (Leiden University Medical Center, Leiden, The Netherlands) for their support with MALDI-FTICR-MS and MALDI-TOF-MS analyses, respectively.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ymeth.2015.03. 001.

I. van den Broek et al. / Methods 81 (2015) 74–85

References [1] J.R. Barr, V.L. Maggio, D.G. Patterson Jr., G.R. Cooper, L.O. Henderson, W.E. Turner, S.J. Smith, W.H. Hannon, L.L. Needham, E.J. Sampson, Clin. Chem. 42 (1996) 1676–1682. [2] J.O. Becker, A.N. Hoofnagle, Bioanalysis 4 (2012) 281–290. [3] A.N. Hoofnagle, M.Y. Roth, J. Clin. Endocrinol. Metab. 98 (2013) 1343–1352. [4] A.N. Hoofnagle, M.H. Wener, J. Immunol. Methods 347 (2009) 3–11. [5] S.A. Agger, L.C. Marney, A.N. Hoofnagle, Clin. Chem. 56 (2010) 1804–1813. [6] U. Ceglarek, J. Dittrich, S. Becker, F. Baumann, L. Kortz, J. Thiery, Proteomics Clin. Appl. 7 (2013) 794–801. [7] M.E. Lassman, T.M. McLaughlin, E.P. Somers, A.C. Stefanni, Z. Chen, B.A. Murphy, K.K. Bierilo, A.M. Flattery, K.K. Wong, J.M. Castro-Perez, B.K. Hubbard, T.P. Roddy, Rapid Commun. Mass Spectrom. 26 (2012) 101–108. [8] N.P. Smit, F.P. Romijn, I. van den Broek, J.W. Drijfhout, M. Haex, A. van der Laarse, Y.E. van der Burgt, C.M. Cobbaert, J. Proteomics 109C (2014) 143–161. [9] R.G. Kay, B. Gregory, P.B. Grace, S. Pleasance, Rapid Commun. Mass Spectrom. 21 (2007) 2585–2593. [10] M.A. Kuzyk, D. Smith, J. Yang, T.J. Cross, A.M. Jackson, D.B. Hardie, N.L. Anderson, C.H. Borchers, Mol. Cell. Proteomics 8 (2009) 1860–1877. [11] M. Rezeli, A. Vegvari, F. Donnarumma, O. Gidlof, J.G. Smith, D. Erlinge, G. Marko-Varga, J. Proteomics 87 (2013) 16–25. [12] L. Anderson, C.L. Hunter, Mol. Cell. Proteomics 5 (2006) 573–588. [13] G. Walldius, I. Jungner, J. Intern. Med. 259 (2006) 493–519. [14] I. van den Broek, F.P. Romijn, N.P. Smit, A. van der Laarse, J.W. Drijfhout, Y.E. van der Burgt, C.M. Cobbaert, J. Proteome Res. 14 (2015) 928–942. [15] N.L. Anderson, N.G. Anderson, L.R. Haines, D.B. Hardie, R.W. Olafson, T.W. Pearson, J. Proteome Res. 3 (2004) 235–244. [16] M. Razavi, L.E. Frick, W.A. LaMarr, M.E. Pope, C.A. Miller, N.L. Anderson, T.W. Pearson, J. Proteome Res. 11 (2012) 5642–5649. [17] M. Razavi, L.D. Johnson, J.J. Lum, G. Kruppa, N.L. Anderson, T.W. Pearson, Clin. Chem. 59 (2013) 1514–1522. [18] N.L. Anderson, M. Razavi, T.W. Pearson, G. Kruppa, R. Paape, D. Suckau, J. Proteome Res. 11 (2012) 1868–1878. [19] M.S. Ritorto, R. Ewan, A.B. Perez-Oliva, A. Knebel, S.J. Buhrlage, M. Wightman, S.M. Kelly, N.T. Wood, S. Virdee, N.S. Gray, N.A. Morrice, D.R. Alessi, M. Trost, Nat. Commun. 5 (2014) 4763. [20] R. Patel, Clin. Chem. 61 (2015) 100–111.

85

[21] A.G. Camenzind, J.G. van der Gugten, R. Popp, D.T. Holmes, C.H. Borchers, Clin. Proteomics 10 (2013) 20. [22] J. Jiang, C.E. Parker, K.A. Hoadley, C.M. Perou, G. Boysen, C.H. Borchers, Proteomics Clin. Appl. 1 (2007) 1651–1659. [23] P.E. Oran, O. Trenchevska, D. Nedelkov, C.R. Borges, M.R. Schaab, D.S. Rehder, J.W. Jarvis, N.D. Sherma, L.H. Shen, B. Krastins, M.F. Lopez, D.C. Schwenke, P.D. Reaven, R.W. Nelson, PLoS One 9 (2014). [24] E.E. Niederkofler, K.A. Tubbs, K. Gruber, D. Nedelkov, U.A. Kiernan, P. Williams, R.W. Nelson, Anal. Chem. 73 (2001) 3294–3299. [25] CLSI, Preparation and validation of commutable frozen human serum pools as secondary reference materials for cholesterol measurement procedures; approved guideline, CLSI Document C37-A, Clinical and Laboratory Standards Institute, Wayne, PA, 1999. [26] S.M. Marcovina, J.J. Albers, L.O. Henderson, W.H. Hannon, Clin. Chem. 39 (1993) 773–781. [27] S.M. Marcovina, J.J. Albers, H. Kennedy, J.V. Mei, L.O. Henderson, W.H. Hannon, Clin. Chem. 40 (1994) 586–592. [28] M. Razavi, M.E. Pope, M.V. Soste, B.A. Eyford, A.M. Jackson, N.L. Anderson, T.W. Pearson, J. Immunol. Methods 364 (2011) 50–64. [29] M.R. Bladergroen, R.J. Derks, S. Nicolardi, B. de Visser, S. van Berloo, Y.E. van der Burgt, A.M. Deelder, J. Proteomics 77 (2012) 144–153. [30] CLSI, Measurement procedure comparison and bias estimation using patient samples; approved guideline – third ed., CLSI document EP09-A3, Clinical and Laboratory Standards Institute, Wayne, PA, 2013. [31] CLSI, User verification of performance for precision and trueness; approved guideline – second ed., CLSI document EP15-A2, Clinical and Laboratory and Standards Institute, Wayne, PA, 2005. [32] CLSI, Clinical and Laboratory Standards Institute, Interference testing in clinical chemistry; approved guideline – second ed., CLSI document EP7-A2, Clinical and Laboratory Standards Institute, Wayne, 2005. [33] EMA, Guideline on bioanalytical method validation, European Medicines Agency, committee for medicinal products for human use, 2011. [34] R.P. Grant, A.N. Hoofnagle, Clin. Chem. 60 (2014) 941–944. [35] J.L. Proc, M.A. Kuzyk, D.B. Hardie, J. Yang, D.S. Smith, A.M. Jackson, C.E. Parker, C.H. Borchers, J. Proteome Res. 9 (2010) 5422–5437. [36] I. van den Broek, N.P. Smit, F.P. Romijn, A. van der Laarse, A.M. Deelder, Y.E. van der Burgt, C.M. Cobbaert, J. Proteome Res. 12 (2013) 5760–5774. [37] C.G. Fraser, American Association of Clinical Chemistry, ISBN 1-890883-49-2, 2001.