Relative Quantitation of Glycoisoforms of Intact Apolipoprotein C3 in

Article pubs.acs.org/ac

Relative Quantitation of Glycoisoforms of Intact Apolipoprotein C3 in Human Plasma by Liquid Chromatography−High-Resolution Mass Spectrometry Wenying Jian,† Richard W. Edom,† Dai Wang,† Naidong Weng,† and Stanley (Weihua) Zhang*,‡ †

Janssen Research and Development, Johnson & Johnson, 1000 Route 202 South, Raritan, New Jersey 08869, United States Ortho-Clinical Diagnostics, Johnson & Johnson, 1001 Route 202 North, Raritan, New Jersey 08869, United States

‡

S Supporting Information *

ABSTRACT: Glycosylation is one of the most important posttranslational modifications to mammalian proteins. Distribution of different glycoisoforms of certain proteins may reflect disease conditions and, therefore, can potentially be utilized as biomarkers. Apolipoprotein C3 (ApoC3) is one of the many plasma glycoproteins extensively studied for association with disease states. ApoC3 exists in three main glycoisoforms, including ApoC3-1 and ApoC3-2, which contain an O-linked carbohydrate moiety consisting of three and four monosaccharide residues, respectively, and ApoC3-0 that lacks the entire glycosylation chain. Changes in the ratio of different glycoisoforms of ApoC3 have been observed in pathological conditions such as kidney disease, liver disease, and diabetes. They may provide important information for diagnosis, prognosis, and evaluation of therapeutic response for metabolic conditions. In this current work, a liquid chromatography(LC)−high-resolution (HR) time-of-flight (TOF) mass spectrometry (MS) method was developed for relative quantitation of different glycoisoforms of intact ApoC3 in human plasma. The samples were processed using a solid-phase extraction (SPE) method and then subjected to LC−full scan HRMS analysis. Isotope peaks for each targeted glycoisoform at two charge states were extracted using a window of 50 mDa and integrated into a chromatographic peak. The peak area ratios of ApoC3-1/ApoC3-0 and ApoC3-2/ApoC3-0 were calculated and evaluated for assay performance. The results indicated that the ratio can be determined with excellent reproducibility in multiple subjects. It has also been observed that the ratios remained constant in plasma exposed to room temperature, freeze−thaw cycles, and long-term frozen storage. The method was applied in preliminary biomarker research of diabetes by analyzing plasma samples collected from normal, prediabetic, and diabetic subjects. Significant differences were revealed in the ApoC3-1/ ApoC3-0 ratio and in the ApoC3-2/ApoC3-0 ratio among the three groups. The workflow of intact protein analysis using full scan HRMS established in this current work can be potentially extended to relative quantitation of other glycosylated proteins. To our best knowledge, this is the first time that a systematic approach of relative quantitation of targeted intact protein glycoisoforms using LC−MS has been established and utilized in biomarker research.

M

utility as disease biomarkers. ApoC3 is a 79 amino acid protein (Figure 1) synthesized by liver and intestine. It is an essential component of circulating particles rich in triglycerides, such as very low density lipoprotein (VLDL) and chylomicrometer remnants. ApoC3 modulates the binding of these lipid-rich particles to receptors and to low-density lipoprotein (LDL), thus decreasing their uptake by liver cells and subsequent degradation.6,7 ApoC3 gene and protein expression has been indicated in various diseases and conditions such as hypertriglyceridemia, insulin resistance, cardiovascular diseases, and so on.8−13 ApoC3 in human plasma exists mainly in three glycoisoforms.14,15 The O-linked sugar moiety is bound to threonine in position 74 and consists of one residue of galactose, one residue of N-acetyl-galatosamine, and one and two residues of N-acetylneuraminic acid (NeuNAc, known as sialic acid)

ammalian proteins exist in a high level of complexity due to variations in genetic codes, alternative splicing and processing, as well as co- or post-translational modifications. Among them, glycosylation is the most frequent modification, and it presents the most complicated co- or post-translational modification that a protein can undergo.1 Glycosylation is an enzyme-directed, site-specific process which results in glycans attached to protein molecules. The majority of proteins synthesized in the rough endoplasmic reticulum undergo glycosylation, which presents a high level of heterogeneity and potential association with disease processes. Certain diseases may selectively alter biosynthetic steps of glycosylated proteins in the endoplasmic reticulum and Golgi, and their export into plasma, therefore causing changes in the distribution of different glycosylation isoforms. Recent research has shown that glycoisoforms of certain proteins may be utilized as disease biomarkers, as demonstrated in studies of cancers, cardiovascular diseases, and liver diseases.2−5 Apolipoprotein C3 (ApoC3) is one of many plasma glycoproteins which have been extensively studied for potential © 2013 American Chemical Society

Received: November 30, 2012 Accepted: January 31, 2013 Published: January 31, 2013 2867

dx.doi.org/10.1021/ac3034757 | Anal. Chem. 2013, 85, 2867−2874

Analytical Chemistry

Article

information regarding the structure can be obtained. Matrixassisted laser desorption ionization (MALDI) time-of-flight (TOF) and surface-enhanced laser desorption ionization (SELDI)-TOF approaches have been adopted to study the polymorphism of intact apolipoproteins and their association with diseases.27,28 In work presented by Harvey et al., MALDITOF profiling of glycoisoforms of ApoC3 has been conducted to probe glycomic changes associated with obesity, liver diseases, and sepsis.17 However, MALDI-TOF and SELDITOF are low-resolution techniques, requiring off-line purification. In addition, they are not well-accepted tools for quantitation due to their lack of essential quantitative properties such as reproducibility and linearity. Therefore, the applications of MALDI-TOF and SELDI-TOF have mainly focused on proteomics-based profiling approaches. Recently, Mazur et al. profiled intact apolipoproteins isolated from human highdensity lipoprotein (HDL) using LC−LTQ−FT (linear trap quadrupole Fourier transform) mass spectrometry and identified ApoC3-1 as a protein associated with coronary artery disease.29 Overall, most glycosylation research based on MS techniques has been focused on structure analysis and glycomics profiling.30,31 No LC−MS bioanalytical method has been reported for targeted quantitation of ratios of different intact ApoC3 glycoisoforms. To our best knowledge, there is no such method for any other glycoprotein either. The purpose of the current work was to establish such a method for ApoC3, as well as to develop a generic workflow for the targeted quantitation of ratios of protein glycoisoforms. What is unique about this work is that our LC−MS-based method is utilized for targeted quantitation rather than structure analysis or profiling of the glycoproteins. Quantitation parameters essential for bioanalysis, including reproducibility and stability, were evaluated. It was decided that a relative quantitation approach would be most appropriate for the current task. There are two main reasons: (1) No reference standards are available for discrete glycoisoforms of ApoC3. All of the commercially available ApoC3 standards are isolated from human plasma, and they consist of a mixture of the three different forms in variable proportions. It is also technically unfeasible to isolate or to synthesize the specific glycoisoforms. (2) More importantly, since determining the ratio is the goal of the assay, a method that can determine the relative abundance of different glycoisoforms in a reproducible manner would fulfill the need. In the current work, a solid-phase extraction method was developed to cleanup the plasma samples, which were further subjected to LC−high-resolution MS analysis. This “top-down” approach provided intact analysis of ApoC3 glycoisoforms and potential for data mining, while HRMS afforded excellent specificity. The capability of the assay to reproducibly determine the ratio was further evaluated, as was the stability of the ratio in plasma samples that were exposed to various storage conditions. The assay was also applied to analysis of plasma samples collected from normal, prediabetic, and diabetic subjects for preliminary evaluation of the biomarker potential of ApoC3 glycoisoforms for early diagnosis of diabetes. To our best knowledge, this is the first time that a systematic approach of relative quantitation of targeted intact protein glycoisoforms using LC−MS has been established and utilized in biomarker research.

Figure 1. Amino acid sequence of ApoC3 and the structures of different glycoisoforms: Gal, galactose; GalNAc, N-acetyl-galactose; NeuNAc, N-acetyl-neuraminic acid. T (Thr) indicates the position for glycosylation.

for ApoC3-1 and ApoC3-2, respectively, while ApoC3-0 lacks the entire sugar chain (Figure 1). Changes in ApoC3 glycoisoform ratios have been observed in obesity, kidney diseases, liver diseases, sepsis, and so on, and may provide important information for diagnosis, prognosis, and evaluation of therapeutic response for metabolic conditions.16−18 Our preliminary internal research has also demonstrated the potential for utilization of the ratio of ApoC3 glycoisoforms as biomarkers for early diagnosis of prediabetes and diabetes. It is highly challenging to analyze and quantify glycoisoforms of proteins. Traditional immuno-based assays may lack sufficient specificity due to the similarity in structure among the different glycoisoforms. For example, it has been reported that an anti-ApoC3-1 antibody developed in rabbits crossreacted completely with other ApoC3 isoforms.19 Electrophoresis-based methods, including isoelectric focusing (IEF)19,20 and two-dimensional electrophoresis (IEF and sodium dodecyl sulfate gel),21,22 have been utilized for glycoisoform analysis. It has been shown that analytical IEF can resolve the desialylated (ApoC3-0) from the sialylated (ApoC3-1 and ApoC3-2) subspecies of ApoC3 and the resolved bands can be measured by densitometric scanning.19 However, electrophoresis techniques are only semiquantitative and lack the capability to differentiate electrically neutral sugar moieties, electrically neutral amino acid substitutions, oxidation, and some other post-translation modifications.23 Developments in liquid chromatography−mass spectrometry (LC−MS)technology have provided a platform for profiling the serum glycome with higher specificity, either as sugar chains released from the protein or as peptides conjugated with sugar chains generated from proteolytic digestion.24−26 Alternatively, intact glycoprotein analysis by mass spectrometry provides a complementary and attractive approach, as more comprehensive

■

MATERIALS AND METHODS Materials. The K2EDTA control human plasma and plasma from normal, diabetic, and prediabetic patients were purchased from Bioreclamation (Hicksville, NY). The samples were obtained

2868

dx.doi.org/10.1021/ac3034757 | Anal. Chem. 2013, 85, 2867−2874

Analytical Chemistry

Article

Figure 2. LC−TOF MS analysis of ApoC3 in human plasma. (A) TIC of full TOF MS scan. (B) Mass spectrum at retention time of 3.5 min (as selected in panel A). (C) Enlarged spectrum of ApoC3-1 at charge state 6 (as indicated by ∗ in panel B). The extraction window for peak integration was shown by the shade. (D) Chromatographic peak of ApoC3-1 by integrating three most abundant isotopic peaks at each of charge states 5 and 6.

column (1.0 mm × 50 mm, 5 μm, 300 Å, Phenomenex, Torrance, CA). Mobile phase A was 0.1% trifluoroacetic acid (TFA) and 0.5% acetic acid in water, and mobile phase B was 0.1% TFA and 0.5% acetic acid in acetonitrile. The needle rinse solvent was 0.1% TFA in 50% acetonitrile in water (v/v/v). The gradient elution started at 5% mobile phase B for the first 0.2 min, ramped linearly to 90% B in 3.8 min, held at 90% for 0.5 min, and then returned to 5% B in 0.1 min. The HPLC flow rate was 0.3 mL/min, and the total run time was 6 min per sample. The LC flow was diverted to waste before 2.0 min and after 4.5 min. TOF MS analysis was carried out with an API 5600 triple TOF (Q-TOF) mass spectrometer (AB Sciex, Foster City, CA). The ion source parameters in positive turbo ionspray mode were as follows: curtain gas 40 psi, GAS1 40 psi, GAS2 50 psi, ionspray voltage 5500 V, and source temperature 500 °C. The declustering potential (DP) and collision energy were 165 and 10 V, respectively. The TOF mass range was set to m/z 600−2000, and the accumulation time was 0.25 s. Mass calibration was conducted using APCI positive calibration standard (Applied Biosystems, Foster City, CA) delivered at speed of 50 μL/min for 2 min by the calibration delivery system (CDS) every few (5−10) injections. Peak integration was conducted using MultiQuant on the Analyst software system. The multiple reaction monitoring (MRM) analysis was carried out with an API 4000 triple-quadrupole mass spectrometer (Applied Biosystems, Foster City, CA). The operating

by the vendor upon consent of the patients. The information accompanying the samples included age, gender, medications, blood hemoglobin A1c, and blood glucose levels. Stable isotope labeled ApoC3 protein (D12-ApoC3-0 in which Ala76, Ala78, and Ala79 were replaced with D4-Ala) was synthesized by NeoBioSci (Cambridge, MA). Strata-XL 100 μm polyermic reversed-phase solid-phase extraction (SPE) 96-well plates (30 mg/1 mL) were obtained from Phenomenex (Torrance, CA). Sample Preparation. The plasma samples were processed using SPE. An aliquot of 0.1 mL of each plasma sample was fortified with 0.3 mL of freshly prepared 10% formic acid in water for denaturation, and 20 μL of internal standard (IS) working solution (60 μg/mL of D12-ApoC3 dissolved in horse plasma) was added, mixed thoroughly, and loaded onto the 96-well SPE plate. Each well on the plate was preconditioned with 0.8 mL of methanol and 0.8 mL of water. After loading, the SPE was washed with 0.8 mL of water, followed by 0.8 mL of 30% methanol, and then eluted with 0.4 mL of 0.2% trifluoroacetic acid in methanol. The eluent was evaporated to dryness under a gentle nitrogen stream at a temperature of 30 °C and then reconstituted with 0.1 mL of 10% acetonitrile in water. An aliquot of 15 μL was injected for LC−MS analysis. LC−MS Conditions. The HPLC system consisted of Shimadzu LC20AD pumps and a SIL-HTC autosampler (Columbia, MD). The HPLC system employed a Jupiter C18 2869

dx.doi.org/10.1021/ac3034757 | Anal. Chem. 2013, 85, 2867−2874

Analytical Chemistry

Article

other glycoisoforms and the internal standard. The extraction windows are listed in Table 1. Then, the ratio of ApoC3-1 chromatographic peak area to that of ApoC3-0 and the ratio of ApoC3-2 peak area to that of ApoC3-0 were calculated and further used in assay evaluation and sample analysis. Our full scan HRMS-based “top-down” approach affords several major advantages in comparison to “bottom-up”, as well as MRM-based “top-down” approaches: (1) In comparison to the “bottom-up” approach, where selected peptides generated from enzymatic digestion are monitored as surrogates of the protein,34,35 the current work flow allows acquisition of information from the whole protein, as well as multiple posttranslational modification species in a single run. (2) Compared to the MRM-based approach,36,37 it is more efficient to use full scan HRMS. There is no prerequisite to know the expected parent ion and product ion, and a generic MS method can be used. (3) The current approach can provide better specificity than an MRM-based approach. On an MRM workflow, the specificity of analysis relies on fragmentation of the selected parent ion to produce unique product ions. However, this process is often of low efficiency for proteins. For example, it was found that there were no suitable product ions that could be generated in collision-induced dissociation (CID) when we used MRM for ApoC3 proteins. As a result, “pseudo-MRM”, i.e., monitoring of parent ion to parent ion with very low collision energy, had to be conducted, which significantly sacrificed the specificity of the assay. In comparison, the specificity of HRMS was independent of fragmentation. The TOF instrument can provide high resolution to separate the isotope peaks of the targeted glycoisoform from the noise, as well as give high mass accuracy. As a result, the multiple isotope peaks at each charge state can be extracted using a very narrow window to avoid interference. As demonstrated in Figure 3, the interference peaks were at a much lower relative level in LC− TOF than in LC−MRM analysis, especially for ApoC3-0, which was present at the lowest abundance among all three species. (4) In addition, the current nontargeted approach allows the possibility to conduct postacquisition data mining to reveal other components, such as biotransformation products, different posttranslational modification species, or even other proteins. For example, it was found that the ions m/z 1105.9308 (z = 6) and m/z 1326.9148 (z = 5) were in good agreement with the theoretical m/z of ApoC1 (difference from theoretical value 1461.9 (C3-0, 6+), 1885.3 > 1885.3 (C3-1, 5+), 1571.2 > 1571.2 (C3-1, 6+), 1943.5 > 1943.5 (C3-2, 5+), 1619.8 > 1619.8 (C3-2, 6+), 1756.4 > 1756.4 (IS, 5+), and 1463.9 > 1463.9 (IS, 6+). Method Evaluation. Intraday and Interday Assay Precision. Six individual lots of human plasma were analyzed in triplicate for three runs on separate days. The mean peak area ratios of ApoC3-1/ApoC3-0 and ApoC3-2/ApoC3-0 and their percent coefficients of variation (% CVs) (intraday for an individual run and interday for all three runs) were calculated for each lot. Stability. Six individual lots of human plasma samples were kept at room temperature for 24 h, frozen (−20 °C) and thawed (room temperature) for three cycles, or kept frozen at −20 °C for 131 days, and then analyzed in triplicate. The mean peak area ratio of ApoC3-1/ApoC3-0 and ApoC3-2/ApoC3-0 for each lot under each condition were compared to the values determined in the same lots of plasma before they were subjected to these conditions (control).

■

RESULTS AND DISCUSSION LC−MS Workflow. In recent years, “top-down” MS analysis of intact proteins as whole has emerged and gained significance, especially for proteins of relatively smaller sizes, where molecular weight is not a limitation on current MS instrumentation.32,33 In the current work, we used a full scan high-resolution MS approach for “top-down” analysis of ApoC3 proteins and relative quantitation of its different glycoisoforms. Human plasma was denatured by addition of acidic buffer and processed using SPE to remove large proteins and salts. The resulting samples were subjected to a short LC run on an analytical column followed by high-resolution full TOF MS scanning. Figure 2A demonstrates the total ion current (TIC) of a typical TOF MS scan. By processing the TIC using extraction of one of the theoretical m/z of ApoC3 (Table 2), it was found the ApoC3 species eluted at about 3.5 min. Figure 2B is the highresolution MS spectra of a time span of 0.1 min at retention time 3.5 min (as selected in Figure 2A). The spectra showed that glycosylated ApoC3, including ApoC3-1 and ApoC3-2, were the more abundant species, and they were present at charge states of 5, 6, 7, 8, and 9, while ApoC3-0 gave a much weaker signal, and it could only be observed at charge states 5 and 6 for some of the samples containing low levels of ApoC3-0. The theoretical and observed m/z of the most abundant monoisotopic peak of each species from charge states 5−9 are listed in Table 1. The calculated difference between the observed and theoretical values was less than 10 ppm for all ions, indicating excellent MS accuracy. Figure 2C illustrates the zoomed-in HRMS spectra around m/z 1571 (indicated by ∗ in Figure 2B), the 6 charge state of ApoC3-1, as a representative. Near-baseline separation of the isotopic peaks was achieved. Peak integration was conducted on the three most abundant ions, 1570.9104, 1571.0773, and 1571.2444 (theoretical values), using an extraction window of 50 mDa, as illustrated in the figure. It can be seen that the noise between the isotope peaks was excluded, thus eliminating potential interference. The three most abundant isotopic peaks at charge state 5, 1884.8924, 1885.0927, and 1885.2933 (theoretical values), were also extracted. The chromatogram resulting from the combination of all six ions can be seen in Figure 2D, which is a distinct chromatographic peak at a retention time of 3.5 min. The same procedure was done for the 2870

dx.doi.org/10.1021/ac3034757 | Anal. Chem. 2013, 85, 2867−2874

obsd

theor

1753.8472

1461.7060

1253.0337

1096.5295

974.8040

charge state

5

6

7

8

9

2871

1461.5505

1461.7167

1461.8833

1461.5391

1461.7060

1461.8731

1754.0551

1754.0477

6

1753.8574

1753.8472

5

obsd

1753.6551

theor

1753.6469

charge state

1047.7182

1178.5580

1346.7805

1571.0773

1885.0927

theor

1047.7283

1178.5674

1346.7879

1571.0863

1885.0990

obsd

9420.4637

ApoC3-1

9.65

8.01

5.47

5.74

3.32

difference (ppm)

extraction window

1461.8533−1461.9033

1461.6875−1461.7375

1461.5280−1461.5780

1754.0281−1754.0781

1753.8340−1753.8840

theor

1571.2444

1571.0773

1570.9104

1885.2933

1885.0927

1884.8924

1571.2536

1571.0863

1570.9181

1885.3028

1885.0990

1884.8981

obsd

extraction window

1571.2255−1571.2755

1571.0586−1571.1086

1570.8915−1570.9415

1885.2677−1885.3177

1885.0690−1885.1190

1884.8685−1884.9185

ApoC3-1 theor

1619.7603

1619.5932

1619.4263

1943.5123

1943.3118

1943.1115

obsd

1619.7675

1619.6006

1619.4332

1943.5152

1943.3162

extraction window

8.40

7.91

6.85

4.58

2.25

difference (ppm)

1619.7414−1619.7914

1619.5722−1619.6222

1619.4055−1619.4555

1943.4865−1943.5365

1943.2880−1943.3380

1943.0876−1943.1376

ApoC3-2

1080.0712

1214.9545

1388.3751

1619.6006

1943.3162

obsd

9711.5591

ApoC3-2

1943.1161

1080.0621

1214.9449

1388.3656

1619.5932

1943.3118

theor

(B) Extraction Window for Each Extracted Isotopic Peak (Width = 50 mDa)

7.99

6.65

5.56

7.31

5.80

difference (ppm)

1753.6306−1753.6806

ApoC3-0

974.8118

1096.5368

1253.0407

1461.7167

1753.8574

8764.2361

MW

ApoC3-0

(A) Molecular Weight, Theoretical and Observed m/z at Different Charging States

1463.8852

1463.7181

1463.5512

1756.4623

1756.2618

1756.0614

theor

obsd

obsd

1463.8917

1463.7232

1463.5589

1756.4672

1756.2665

8.89

6.65

5.18

3.46

2.70

difference (ppm)

1463.8654−1463.9154

1463.7003−1463.7503

1463.5312−1463.5812

1756.4422−1756.4922

1756.2404−1756.2904

1756.0396−1756.0896

extraction window

internal standard

976.1541

1098.0459

1254.7649

1463.7232

1756.2665

1756.0654

976.1454

1098.0386

1254.7584

1463.7181

1756.2618

theor

8776.3088

internal standard

Table 1. (A) Calculated Molecular Weight, Theoretical m/z, and Observed m/z at Different Charge States for Most Abundant Monoisotopic Peak of ApoC3-0, ApoC3-1, ApoC3-2 (from a Representative Sample), and the Stable Isotope Labeled Internal Standard (from an Injection of Pure Protein): (B) Extraction Windows (50 mDa) for Each Extracted Isotope Peak Used in MultiQuant Software for Peak Integration

Analytical Chemistry Article

dx.doi.org/10.1021/ac3034757 | Anal. Chem. 2013, 85, 2867−2874

Analytical Chemistry

Article

Figure 3. Comparison of chromatography obtained from LC−TOF analysis (left panel) and LC−MRM analysis (right panel) of ApoC3 in human plasma.

contain no interference components to human ApoC3. The theoretical and observed m/z and extraction window of the IS are listed Table 1. The purpose of having an internal standard was to ensure general quality of the assay, such as monitoring if there was variability in the SPE procedure or LC injection. However, it was not used in relative quantitation because the peak areas of ApoC3-0, ApoC3-1, and ApoC3-2 were directly used in peak area ratio calculations. Since the IS signal was the same for all three species in each sample, it would be canceled out if peak area ratio of glycoisoform/IS was used in calculation. Another purpose for having the IS in the sample was for potential evaluation of absolute abundance. The ratio of each species to IS observed in different patients can be calculated to compare the absolute abundance of that glycoisoform in different patients. In fact, upon sample analysis for normal, prediabetic, and diabetic subjects, we calculated ApoC3-0/IS, ApoC3-1/IS, and ApoC3-2/IS. However, the data did not indicate any trend (data not shown), and this approach was therefore not further pursued. However, this technique may be suitable for other protein biomarkers for which the absolute rather than relative abundance is more relevant to the pathological or pharmacological effects to be investigated. Method Evaluation. The analytical target in the current work is the ratio of ApoC3-1/ApoC3-0 and ApoC3-2/ApoC3-0, rather than the absolute abundance of each individual isoform. The method evaluation was conducted to elucidate the capability of the assay to obtain the ratio values reproducibly, upon repeated analysis (intraday and interday), as well as after exposure of the samples to room temperature, freeze−thaw cycles, and long-term storage. The results for ApoC3-1/ApoC3-0 in plasma from six individual normal humans are shown in Table 2A. The ratios ranged between 5.66 and 11.00. The intraday and interday % CVs were within 6.87% and 4.21%, respectively, indicating that the ratios can be determined in a highly reproducible manner. In the stability test, the % difference from initial value upon exposure to room temperature for 24 h, upon freeze−thaw for three cycles, and frozen for 131 days were within ±11.21%, ±4.23%, and ±8.58%, respectively, well within the acceptance criteria for LC− MS-based bioanalytical assays. Similar observations were made for ApoC3-2/ApoC3-0 (Table 2B).

Table 2. Assay Evaluation Results (n = 3) (A) ApoC3-1/ApoC3-0 lot 1

lot 2

lot 3

lot 4

lot 5

lot 6

10.35 2.32 1.61

5.66 3.83 3.55

11.00 2.20 4.21

10.35 10.73 3.66

5.66 5.95 5.08

11.00 11.84 7.66

10.35 10.43 0.77

5.66 5.90 4.23

11.00 10.63 −3.32

10.58 10.83 2.42

6.05 5.82 −3.89

11.17 10.21 −8.58

3.56 6.50 3.05

4.53 5.26 0.64

4.90 4.25 5.37

11.61 4.22 2.24

3.56 4.00 12.37

4.53 4.79 5.80

4.90 5.18 5.75

11.61 13.28 14.37

3.56 3.55 −0.23

4.53 4.58 1.12

4.90 4.89 −0.12

11.61 11.06 −4.76

3.60 3.79 5.32

4.64 4.81 3.80

5.17 5.14 −0.48

12.40 10.59 −14.63

Precision av 6.01 7.04 8.52 % CV (intraday) 3.42 1.24 6.87 % CV (interday) 2.25 1.71 2.75 Room Temp 24 h reference 6.01 7.04 8.52 observed 6.08 6.25 8.56 % difference 1.19 −11.21 0.54 Freeze−Thaw Three Cycles reference 6.01 7.04 8.52 observed 5.88 6.99 8.25 % difference −2.30 −0.70 −3.12 Frozen 131 Days reference 6.15 7.15 8.43 observed 6.16 7.17 8.87 % difference 0.09 0.25 5.24 (B) ApoC3-2/ApoC3-0 Precision av 5.90 4.03 % CV (intraday) 8.70 1.71 % CV (interday) 4.93 4.14 Room Temp 24 h reference 5.90 4.03 observed 6.23 3.66 % difference 5.57 −9.25 Freeze−Thaw Three Cycles reference 5.90 4.03 observed 5.64 4.00 % difference −4.33 −0.81 Frozen 131 Days reference 6.19 4.32 observed 6.33 4.49 % difference 2.21 4.07

Initial assay evaluation was conducted for six lots of plasma from normal subjects. The capability of the assay to measure the ratio in more diversified samples with a broader distribution 2872

dx.doi.org/10.1021/ac3034757 | Anal. Chem. 2013, 85, 2867−2874

Analytical Chemistry

Article

Figure 4. Distribution of log-transformed ApoC3-1/ApoC3-0 ratio.

Receiver operating characteristic (ROC) curves were also constructed for ApoC3-1/ApoC3-0, ApoC3-2/ApoC3-0, hemoglobin A1c, and glucose to compare the clinical utility. The results indicated that neither ApoC3-1/ApoC3-0 nor ApoC3-2/ ApoC3-0 had better sensitivity or specificity than hemoglobin A1c or glucose for distinguishing the normal, prediabetic, or diabetic subjects (data not shown). The hemoglobin A1c value, which is the percentage of glycosylated hemoglobin to total hemoglobin,38 was the best among all of the evaluated parameters to separate the three groups. In addition, the combination of ApoC3-1/ApoC3-0 or ApoC3-2/ApoC3-0 with hemoglobin A1c did not show any further improvement in diagnostic power compared to hemoglobin A1c alone. Overall, the statistical analysis indicated that there was a significant difference in ApoC3-1/ApoC3-0 among normal, prediabetic, and diabetic subjects. A similar observation was made for ApoC3-2/ApoC3-0, but to a lesser extent. However, the clinical utility of these ratios in the current study was shown to be less valuable than current existing diagnostic tools, i.e., hemoglobin A1c and glucose.38,39 Part of the reason for this result was that the samples used in the analysis were categorized based on hemoglobin A1c and glucose. As a result, statistical analysis was intrinsically biased and favored hemoglobin A1c and glucose. Further exploration using a larger population and an optimized study design may be needed to clearly elucidate the clinical utility of ApoC3 glycoisoform ratios. Previous studies have shown that these ratios for each person were very stable over time.40 Individuals with a unique glycoisoform ratio in the normal range could experience a several-fold change before diagnosis with disease. Thus, it was suggested that a longitudinal assay of each individual and diagnosis by personal change will provide the most sensitive analysis.17

of the values was further confirmed in the following sample analysis for normal, prediabetic, and diabetic subjects. Totally 75 samples, 25 from each category, were analyzed three times on three separate days, in singlet on each day (results are shown in Supporting Information Tables S1 and S2). The observed average ApoC3-1/ApoC3-0 ranged from 2.27 to 47.14. The % CVs of 69 samples (92% of total) were within 10%, and the maximum % CV of all the samples was 15.92%. The observed ApoC3-2/ApoC3-0 ranged from 1.08 to 30.75. The % CVs of 62 samples (82.7% of the total) were within 10%, of nine samples (12% of the total) were between 10% and 15%, while the remaining four were between 15% and 19.5%. Overall, the results demonstrated that the current assay can measure the ApoC3-1/ApoC3-0 ratio and ApoC3-2/ApoC3-0 ratio reproducibly in plasma from normal as well as prediabetic and diabetic subjects. Sample Analysis. As mentioned above, 25 plasma samples were analyzed in each category of normal, prediabetic, and diabetic subjects. The ratios of the different glycoisoforms were determined and compared to evaluate if there was any correlation with the status of the disease. It was determined that the range of ApoC3-1/ApoC3-0 was 2.27−24.90 (with an outlier of 47.14), 3.04−21.47, and 5.64− 39.47, in normal, prediabetic, and diabetic subjects, respectively (see data in Supporting Information Table S1). The average values were calculated to be 9.58 (the outlier was included in calculation), 10.47, and 15.09, respectively. Figure 4 demonstrates the log-transformed ApoC3-1/ApoC3-0 ratio. The logtransformed data fit normal distribution better than untransformed data and was further used in statistical analysis. The analysis of variance (ANOVA) F test indicated a highly significant difference (p < 0.01) in the ApoC3-1/ApoC3-0 ratio among diabetic, prediabetic, and normal subjects with a p value of 0.0008. The range of ApoC3-2/ApoC3-0 ratios were 1.08− 16.53, 1.99−27.09, and 1.97−30.75, for normal, prediabetic, and diabetic subjects, respectively (see data in Supporting Information Table S2). The average values were calculated to be 5.48, 6.59, and 9.43, respectively. The ANOVA F test also indicated a significant difference (p < 0.05) among the three groups with a p value of 0.0109.

■

CONCLUSION AND PERSPECTIVE A relative quantitation method for glycoisoforms of intact ApoC3 in human plasma using LC−HRMS was developed. A solid-phase extraction procedure was utilized to prepare the plasma samples, which were then subjected to LC−HR full scan MS analysis. The three most abundant isotopic peaks at charge 2873

dx.doi.org/10.1021/ac3034757 | Anal. Chem. 2013, 85, 2867−2874

Analytical Chemistry

Article

(10) Chan, D. C.; Chen, M. M.; Ooi, E. M.; Watts, G. F. Int. J. Clin. Pract. 2008, 62, 799−809. (11) Ooi, E. M.; Barrett, P. H.; Chan, D. C.; Watts, G. F. Clin. Sci. (London) 2008, 114, 611−624. (12) Pollin, T. I.; Damcott, C. M.; Shen, H.; Ott, S. H.; Shelton, J.; Horenstein, R. B.; Post, W.; McLenithan, J. C.; Bielak, L. F.; Peyser, P. A.; Mitchell, B. D.; Miller, M.; O’Connell, J. R.; Shuldiner, A. R. Science 2008, 322, 1702−1705. (13) Gervaise, N.; Garrigue, M. A.; Lasfargues, G.; Lecomte, P. Diabetologia 2000, 43, 703−708. (14) Vaith, P.; Assmann, G.; Uhlenbruck, G. Biochim. Biophys. Acta 1978, 541, 234−240. (15) Ito, Y.; Breslow, J. L.; Chait, B. T. J. Lipid Res. 1989, 30, 1781− 1787. (16) Bosello, O.; Cominacini, L.; Zocca, I.; Garbin, U.; Ferrari, F.; Davoli, A. Ann. Nutr. Metab. 1985, 29, 33−39. (17) Harvey, S. B.; Zhang, Y.; Wilson-Grady, J.; Monkkonen, T.; Nelsestuen, G. L.; Kasthuri, R. S.; Verneris, M. R.; Lund, T. C.; Ely, E. W.; Bernard, G. R.; Zeisler, H.; Homoncik, M.; Jilma, B.; Swan, T.; Kellogg, T. A. J. Proteome Res. 2009, 8, 603−612. (18) Atger, V.; Beyne, P.; Frommherz, K.; Roullet, J. B.; Drueke, T. Ann. Biol. Clin. (Paris) 1989, 47, 497−501. (19) Kashyap, M. L.; Srivastava, L. S.; Hynd, B. A.; Gartside, P. S.; Perisutti, G. J. Lipid Res. 1981, 22, 800−810. (20) Harake, B.; Caines, P. S.; Thibert, R. J.; Cheung, R. M. Clin. Biochem. 1991, 24, 255−260. (21) Schmitz, G.; Ilsemann, K.; Melnik, B.; Assmann, G. J. Lipid Res. 1983, 24, 1021−1029. (22) Sprecher, D. L.; Taam, L.; Brewer, H. B., Jr. Clin. Chem. 1984, 30, 2084−2092. (23) Haase, R.; Menke-Mollers, I.; Oette, K. Electrophoresis 1988, 9, 569−575. (24) Kim, Y. G.; Jang, K. S.; Joo, H. S.; Kim, H. K.; Lee, C. S.; Kim, B. G. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2007, 850, 109− 119. (25) Kita, Y.; Miura, Y.; Furukawa, J.; Nakano, M.; Shinohara, Y.; Ohno, M.; Takimoto, A.; Nishimura, S. Mol. Cell. Proteomics 2007, 6, 1437−1445. (26) Atwood, J. A., III; Cheng, L.; Alvarez-Manilla, G.; Warren, N. L.; York, W. S.; Orlando, R. J. Proteome Res. 2008, 7, 367−374. (27) Bondarenko, P. V.; Cockrill, S. L.; Watkins, L. K.; Cruzado, I. D.; Macfarlane, R. D. J. Lipid Res. 1999, 40, 543−555. (28) Fan, Y.; Shi, L.; Liu, Q.; Dong, R.; Zhang, Q.; Yang, S.; Yang, H.; Wu, P.; Yu, J.; Zheng, S.; Yang, F.; Wang, J. Mol. Cancer 2009, 8, 79. (29) Mazur, M. T.; Cardasis, H. L.; Spellman, D. S.; Liaw, A.; Yates, N. A.; Hendrickson, R. C. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 7728−7733. (30) Wuhrer, M.; de Boer, A. R.; Deelder, A. M. Mass Spectrom. Rev. 2009, 28, 192−206. (31) Leymarie, N.; Zaia, J. Anal. Chem. 2012, 84, 3040−3048. (32) Ruan, Q.; Ji, Q. C.; Arnold, M. E.; Humphreys, W. G.; Zhu, M. Anal. Chem. 2011, 83, 8937−8944. (33) Gucinski, A. C.; Boyne, M. T., II. Anal. Chem. 2012, 84, 8045− 8051. (34) Ciccimaro, E.; Blair, I. A. Bioanalysis 2010, 2, 311−341. (35) Li, F.; Fast, D.; Michael, S. Bioanalysis 2011, 3, 2459−2480. (36) Ji, Q. C.; Rodila, R.; Gage, E. M.; El-Shourbagy, T. A. Anal. Chem. 2003, 75, 7008−7014. (37) Ji, Q. C.; Rodila, R.; Morgan, S. J.; Humerickhouse, R. A.; ElShourbagy, T. A. Anal. Chem. 2005, 77, 5529−5533. (38) The International Expert Committee. Diabetes Care 2009, 32, 1327−1334. (39) American Diabetes Association. Diabetes Care 2010, 33 (Suppl. 1), S4−S10. (40) Nelsestuen, G. L.; Zhang, Y.; Martinez, M. B.; Key, N. S.; Jilma, B.; Verneris, M.; Sinaiko, A.; Kasthuri, R. S. Proteomics 2005, 5, 4012− 4024.

state 5 and 6 were extracted using a narrow window (50 mDa) and integrated to generate the chromatographic peak. The peak area ratio of different glycoisoforms was then calculated and used in assay evaluation and a biomarker study. It was demonstrated that reproducibility of the peak area ratio was excellent in plasma obtained from multiple subjects and over extended times and various storage conditions. The method was applied in a preliminary biomarker research study for sample analysis of plasma obtained from normal, prediabetic, and diabetic subjects. The results showed that there was significant difference in ApoC3-1/ApoC3-0 and ApoC3-2/ ApoC3-0 among the different groups. The workflow established in the current study is generic, and the data can mined for other post-translational modifications and proteins, owing to acquisition of a complete full scan data set for all the components in the extracted plasma and the specificity provided by HRMS. Utilization of an internal standard also afforded the possibility of evaluating the association of absolute abundance of the glycoisoforms to diseases. More importantly, the workflow can be easily extended to other glycoprotein biomarkers of similar size. For glycoproteins of larger sizes, immunoaffinity enrichment of the targeted protein and proteolytic digestion may need to be employed. Nevertheless, the full scan HR MS approach would still provide more comprehensive information and more flexible data-mining than targeted MRM detection. The relative quantitation approach in these cases is expected to be valid for biomarker research of certain diseases for which a ratio rather than absolute abundance of glycoisoforms is more relevant.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

REFERENCES

(1) Spiro, R. G. Adv. Protein Chem. 1973, 27, 349−467. (2) Desmyter, L.; Fan, Y. D.; Praet, M.; Jaworski, T.; Vervecken, W.; De Hemptinne, B.; Contreras, R.; Chen, C. J Gastroenterol. Hepatol. 2007, 22, 1148−1154. (3) Zhao, J.; Qiu, W.; Simeone, D. M.; Lubman, D. M. J. Proteome Res. 2007, 6, 1126−1138. (4) Qiu, Y.; Patwa, T. H.; Xu, L.; Shedden, K.; Misek, D. E.; Tuck, M.; Jin, G.; Ruffin, M. T.; Turgeon, D. K.; Synal, S.; Bresalier, R.; Marcon, N.; Brenner, D. E.; Lubman, D. M. J. Proteome Res. 2008, 7, 1693−1703. (5) Mauger, J. F.; Couture, P.; Bergeron, N.; Lamarche, B. J. Lipid Res. 2006, 47, 1212−1218. (6) Ginsberg, H. N.; Le, N. A.; Goldberg, I. J.; Gibson, J. C.; Rubinstein, A.; Wang-Iverson, P.; Norum, R.; Brown, W. V. J Clin. Invest. 1986, 78, 1287−1295. (7) McConathy, W. J.; Gesquiere, J. C.; Bass, H.; Tartar, A.; Fruchart, J. C.; Wang, C. S. J. Lipid Res. 1992, 33, 995−1003. (8) Olivieri, O.; Bassi, A.; Stranieri, C.; Trabetti, E.; Martinelli, N.; Pizzolo, F.; Girelli, D.; Friso, S.; Pignatti, P. F.; Corrocher, R. J. Lipid Res. 2003, 44, 2374−2381. (9) Joven, J.; Vilella, E.; Costa, B.; Turner, P. R.; Richart, C.; Masana, L. Clin. Chem. 1989, 35, 813−816. 2874

dx.doi.org/10.1021/ac3034757 | Anal. Chem. 2013, 85, 2867−2874