Longitudinal Analysis of Maternal Plasma Apolipoproteins in Pregnancy: A Targeted Proteomics Approach

Abstract

Minimally invasive diagnostic tests are needed in obstetrics to identify women at risk for complications during delivery. The apolipoproteins fluctuate in complexity and abundance in maternal plasma during pregnancy and could be incorporated into a blood test to evaluate this risk. The objective of this study was to examine the relative plasma concentrations of apolipoproteins and their biochemically modified subtypes (i.e. proteolytically processed, sialylated, cysteinylated, dimerized) over gestational time using a targeted mass spectrometry approach. Relative abundance of modified and unmodified apolipoproteins A-I, A-II, C-I, C-II, and C-III was determined by surface-enhanced laser desorption/ionization-time of flight-mass spectrometry in plasma prospectively collected from 11 gravidas with uncomplicated pregnancies at 4–5 gestational time points per patient. Apolipoproteins were readily identifiable by spectral pattern. Apo C-III₂ and Apo C-III₁ (doubly and singly sialylated Apo C-III subtypes) increased with gestational age (r²>0.8). Unmodified Apo A-II, Apo C-I, and Apo C-III₀ showed no correlation (r² = 0.01–0.1). Pro-Apo C-II did not increase significantly until third trimester (140 ± 13% of first trimester), but proteolytically cleaved, mature Apo C-II increased in late pregnancy (702 ± 130% of first trimester). Mature Apo C-II represented 6.7 ± 0.9% of total Apo C-II in early gestation and increased to 33 ± 4.5% in third trimester. A label-free, semiquantitative targeted proteomics approach was developed using LTQ-Orbitrap mass spectrometry to confirm the relative quantitative differences observed by surface-enhanced laser desorption/ionization-time of flight-mass spectrometry in Apo C-III and Apo C-II isoforms between first and third trimesters. Targeted apolipoprotein screening was applied to a cohort of term and preterm patients. Modified Apo A-II isoforms were significantly elevated in plasma from mothers who delivered prematurely relative to term controls (p = 0.02). These results support a role for targeted proteomics profiling approaches in monitoring healthy pregnancies and assessing risk of adverse obstetric outcomes.

The maternal physiology during pregnancy is characterized by inflammation and hyperlipidemia. Plasma protein composition fluctuates dynamically throughout gestation to reflect these physiological changes. Apolipoproteins, a diverse subset of triglyceride transport proteins, contribute to the hyperlipidemia of pregnancy by modulating lipid homeostasis in maternal plasma (1–3). Exaggerated hyperlipidemia and peripheral apolipoprotein burden are associated with inflammatory insult and signal obstetric complications (4–5). Numerous post-translationally modified apolipoprotein isoforms are reported in plasma, but it is unclear how these modifications affect apolipoprotein function and plasma distribution. For example, changes in the glycosylation status of apolipoprotein variants predate the onset of clinical symptoms in patients with preeclampsia, a hypertensive disorder of pregnancy with clinical features in common with cardiovascular disease (6–8). The identification and functional characterization of plasma apolipoprotein isoforms and their post-translationally modified subtypes may reveal important diagnostic and/or therapeutic targets for hypertensive disorders of pregnancy (6).

Mass spectrometry and targeted proteomics analyses afford unprecedented sensitivity and specificity for detecting apolipoproteins and their numerous isoforms and subtypes (9–12). Mass spectrometry approaches overcome limitations inherent in biochemical approaches (e.g. ELISA [enzyme-linked immunosorbant assays] and Western blot analysis), especially the lack of specificity of antibodies for post-translationally modified variants of plasma proteins. The objective of this study was to longitudinally evaluate maternal plasma apolipoprotein profile over gestational time by SELDI-TOF-MS (surface-enhanced laser desorption/ionization-time of flight-mass spectrometry)¹ analysis of intact proteins and a complementary targeted LTQ-Orbitrap XL MS approach. We evaluate changes in 13 post-translationally modified subtypes of the plasma apolipoproteins A-II, C-I, C-II, and C-III over gestational time.

EXPERIMENTAL PROCEDURES

Patients

The study was approved by the Madigan Healthcare System Institutional Review Board (Joint Base Lewis-McChord, WA). Patients were invited to participate during their initial OB-GYN appointment if they presented under 12 weeks of gestation, were 18 years of age or older, and were nulliparous. Patients were excluded if they had a history of chronic disease (including hypertension) or infertility treatment. All study participants provided informed consent. Antepartum records were reviewed to determine whether patients experienced an uncomplicated course of pregnancy according to the guidelines of the American College of Obstetrics and Gynecologists (ACOG). Prematurity was defined as delivery at less than 37 weeks of age as per ACOG guidelines (13). Nonpregnant patients were healthy women in the same age range as the pregnant cohort. Plasma was collected during the menstrual, luteinizing hormone (LH) surge, and luteal phases of the menstrual cycle in each patient.

Plasma Collection

The target gestational age ranges for specimen collection were 4–12, 16–22, and 24–28, and 35–38 weeks. In some cases, patients provided two specimens in the 4–12 week range (see Supplemental Table SI). Blood was collected in EDTA Vacutainer tubes (BD Biosciences) and spun at 1500 × g for 15 min at 4 °C. Plasma supernatant was placed on ice and mixed with protease inhibitor mixture (Roche Diagnostics, Mannheim, Germany) and stored in single-use aliquots at −150 °C. Time from phlebotomy to storage was recorded. Specimens with freeze times exceeding 90 min were not analyzed. Typical processing time was ∼30–45 min.

Enzyme-Linked Immunosorbant Assays (ELISA)

Antibodies directed against the analytes Apo C-II, Apo C-III, Apo A-I, and Apo A-II were used in ELISA assays to probe n = 5 patient specimens (n = 2 technical replicates per patient) at 2–4 gestational ages according to manufacturer instructions (AssayPro, St. Charles, MO). Apolipoprotein concentration was determined by measuring optical densities at 450 nm absorption and interpolating to a standard calibration curve by microplate spectrophotometer (Biotek Instruments, Winooski, VT).

HDL, VLDL, and LDL Cholesterol Quantification Assays

High density lipoprotein (HDL) and combined low-density lipoprotein and very low-density lipoprotein (LDL/VLDL) fractions were measured in patient plasma using a colorimetric assay in which cholesterol oxidase reacts with free cholesterol to produce a color change in the presence of a cholesterol probe (BioVision, Mountain View, CA). Fractions were separated and quantified according to manufacturer's instructions.

SELDI-TOF-MS

Aliquots of plasma were denatured with 9 m urea, 2% CHAPS, 50 mm Tris-hydrochloride (pH 9) at room temperature with continuous vortexing for 20 min. SELDI-TOF-MS hydrophobic (H50) arrays (BioRad Laboratories, Hercules, CA) were washed in binding buffer (10.0:0.1:89.9% acetonitrile: trifluoroacetic acid [TFA]: water). One microliter of denatured plasma was diluted in 50 μl binding buffer and applied to the arrays. Arrays were shaken by Micromix (Siemens Medical Solutions Diagnostics, Los Angeles, CA), setting 20/5 for 30 min. To remove unbound proteins, each array was washed 3 times for 5 min in binding buffer on the Micromix, setting 20/7, and desalted by 2 quick washes in deionized distilled water. Sinapinic acid matrix dissolved in 50% acetonitrile/0.1% trifluoroacetic acid (TFA) was applied to spots containing specimens and intact protein calibrants. The instrument was externally calibrated to an array spotted with a standard protein mixture (Bio-Rad) containing hirudin (6963.5 Da), bovine cytochrome C (12230.9 Da), and equine cardiac myoglobin (16951.5 Da). Stringent quality control using manufacturer supplied reference serum and frequent calibration were employed to ensure instrument fidelity as per Bons and colleagues (14). Spectra were internally calibrated to Apo C-I (6630.6 Da), Apo C-III₀ (8764.7 Da), and Apo C-III₁ (9421.3 Da) as described (15). Accurate mass assignment for Apo C-III₀ and Apo C-III₁ was determined by incubating plasma samples overnight with O-glycosidase and/or neuraminidase (Roche Diagnostics). Resultant peaks were assigned mass values of 9421.3 Da (neuraminidase) or 8764.7 Da (neuraminidase and O-glycosidase) corresponding to Apo C-III₁ and Apo C-III₀, respectively, as described (8). Samples were analyzed by SELDI-TOF-MS using matrix attenuation of 1 kDa, mass range of 20 kDa, focus mass of 10 kDa, and laser energy determined empirically (range: 700–2000 nJ). Parameters for a second, mid-mass range analysis were applied for protein candidates greater than 15 kDa: matrix attenuation of 1 kDa, mass range of 10–100 kDa, focus mass of 25 kDa, and empirically determined laser energy (1250–2000 nJ).

RP HPLC MS/MS Analysis by NanoAcquity LC and Orbitrap XL MS

High abundance plasma proteins were extracted by acetonitrile precipitation as described (16). An AQUA peptide standard (10pmol, Sigma Chemical, St. Louis, MO) was added to the specimens. Samples were digested with trypsin as described (17) and desalted using UltraMicroSpin silica C18 column chromatography according to the manufacturer's instructions (The Nest Group, Inc., Southborough, MA, SUM SS18V). Nanoscale liquid chromatography-tandem mass spectrometry was performed using a NanoAquity system for chromatographic separation (Waters Corporation, Milford, MA) and an LTQ-Orbitrap XL mass spectrometer (ThermoFisher Scientific, West Palm Beach, FL). Peptides were concentrated and desalted on a reversed phase C18 trapping column (0.34 μm x 5 mm; Waters Corporation, Milford, MA) for 5 min at 25 μl/min (2.5 μl injection volume). Peptides were eluted from the reversed phase C18 analytical column (75 μm × 250 mm; Waters Corporation) at a flow rate of 300 nl/min using a gradient of 0–40% acetonitrile for 100 min and 40–90% acetonitrile for 12.5 min. A wash program was used to flush the column by 7 alternating saw-tooth gradients (85–2% water/0.1%formic acid and 2–85% acetonitrile, 5 min each gradient). The column was re-equilibrated for 51 min with 100% water/0.1% formic acid. Mass spectral survey scans were acquired by the Orbitrap operated in data dependent MS/MS mode at 60,000 resolution from 350–1800 m/z. Four precursor ions per full scan were selected for MS/MS by the linear trap with a normalized collision-induced dissociation (CID) energy of 35%.

Mass Assignment of Post-translationally Modified Isoforms of Tryptic Peptides

The m/z values for N-terminal tryptic peptides for Apo C-I' (Apo C-I lacking the N-terminal Thr-Pro sequence) and all Apo A-II subtypes were derived from experimentally determined mass assignments in the literature (15, 18) and sequence analysis by in silico tryptic digest calculations (2). The m/z for doubly and triply charged ions corresponding to the predicted glycoconjugate fragments of C-terminal Apo C-III₁ and Apo C-III₂ subtypes was calculated based on literature assignments (15, 19–21). Relative quantification was obtained by area under the extracted ion chromatogram (AUC) of the most abundant isotope normalized to an internal standard AQUA peptide FSEFWDLDPEVRPT(¹³C₆¹⁵N₂) SAAA (commercially synthesized, Sigma Chemical). To confirm identifications of singly and doubly sialylated Apo C-III₁ and Apo C-III₂ tryptic peptides, another (MS³) scan was triggered in the neutral loss mode when one of the 4 most intense peaks from the MS² spectrum corresponded to a neutral loss event of m/z 147.01, 227.97, or 329.39 (empirically derived from pilot data).

Data Processing and Database Searching

Data were processed using the Xcalibur software package, v. 2.0.7 (ThermoFisher Scientific) with post-processing by Proteome Discoverer, v. 1.2 (ThermoFisher Scientific). Searches were conducted against custom-built FASTA files from NCBI sequences for a total of 13 target apolipoprotein subtype sequences (v.2010.07.26). Database searches were applied with precursor m/z tolerances of 2.5 Da and fragment mass tolerances of 1.0 Da. Trypsin was the enzyme specified with an allowance of two missed cleavages. Cysteine residues were searched as carbamidomethylated (+ 57.021 Da). Oxidation of methionine residues was allowed (+ 15.995 Da) (22). False discovery rates (FDR) were determined by searching against a decoy database (target FDR of 0.01). Peptide assignments were filtered by peptide score based on false positive rate estimated by the target/decoy database approach described by Gygi and colleagues (23), (24).

Statistics

SELDI-TOF-MS data were evaluated by one-way analysis of variance by ranks (Kruskal-Wallis) with significance set at p < 0.05. Comparisons between groups were evaluated by analysis of variance by repeated measures. Pairwise comparisons were made with post hoc Bonferroni correction. Areas under the extracted ion chromatogram (AUC) for first and third trimester LTQ Orbitrap XL data were evaluated by nonparametric Wilcoxen signed rank test with significance set at p < 0.05.

RESULTS

SELDI-TOF-MS Identification of Target Apolipoproteins

Maternal plasma was longitudinally collected at 4 gestational ages from each of 11 nulliparous patients: 4–12 weeks, 16–22 weeks, 24–28 weeks, and 35–38 weeks gestation. Antepartum records review confirmed uncomplicated pregnancies in these patients. Patient characteristics are summarized in Supplemental Table SI. Plasma proteins were fractionated in triplicate on-spot using hydrophobic chemically derivatized surfaces (H50 arrays) and proteomic pattern was analyzed by SELDI-TOF-MS. Signature patterns of Apo C-I, Apo C-II, Apo A-II, Apo C-III, and Apo A-I in maternal spectra were identified based on molecular weight and spectral pattern (Fig. 1; Table I) (15, 18, 25–27). The apolipoproteins are stable features of the human plasma proteome readily detectable by published matrix-assisted laser desorption/ionization-time of flight-mass spectrometry (MALDI-TOF-MS) spectral patterns as previously described (Fig. 1; Table I) (10, 15, 19–21, 28). The patterns and relative abundances of Apo C-I, Apo C-II, Apo A-II, Apo C-III, transthyretin tetramers, and Apo A-II dimers were compared in plasma from a single representative patient in trimester 1 (Fig. 1). All presumptive apolipoprotein peaks had signal-to-noise ratios (S/N) greater than 2.5. Coefficients of variance averaged 12.5 ± 2% among replicates.

Fig. 1.

SELDI-TOF-MS Detection of apolipoproteins and related proteins in first trimester plasma. Target proteins were evaluated in trimester 1 plasma from a single representative individual in m/z ranges (A) m/z 6,000–10,000 Da: Apo A-II monomers, Apo C-I, Apo C-II, and Apo C-III isoforms; (B) m/z 10,000–15,000 Da: transthyretin tetramers; (C) m/z 17,000–18,000 Da: Apo A-II dimer isoforms; and (D) m/z 20,000–30,000 Da: mature Apo A-I.

Table I. Calculated and observed molecular masses for apolipoproteins in maternal plasma.

Peak identification	Accession #a	Peak code	Calculated (Da)b	Reference	SELDI measured (m/z)c	S.D.d	Delta M (m/z)f
Apo C-I without N-terminal Thr-Pro	GI178834	CI′	6432.4	15	6434.1	0.2	1.7
Apo C-I	GI178834	CI	6630.6	15, 18	6631.6	Calibe	Calib
Apo C-II, mature form	P02655	CII	8204.1	15, 27	8204.3	0.8	0.2
Apo A-II, cysteinylated, minus C-terminal Gln	P02652	AII′	8682.9	18	8689.9	0.9	7.0
Apo C-III0	P02656	CIII0	8764.7	15, 27	8765.7	Calib	Calib
Apo A-II, cysteinylated	P02652	AII	8810.9	18	8810.5	0.4	−0.4
pro Apo C-II	P02655	proCII	8914.9	15, 26	8915.6	0.4	0.7
Apo C-III1	P02656	CIII1	9421.3	15, 26	9422.3	Calib	Calib
Apo C-III2	P02656	CIII2	9712.6	15, 26	9713.4	0.4	0.8
Apo A-II dimer minus two C-terminal Gln	P02652	AIId′	17123.4	18	17130.2	3.7	6.8
Apo A-II dimer minus one C-terminal Gln	P02652	AIId″	17251.5	18	17249.6	1.2	−1.9
Apo A-II dimer	P02652	AIId″′	17379.5	18	17378.1	1.2	−1.4
Apo A-I, mature form	GI296635	AI	28078.6	18	28079.8	Calib	1.2

^a Accession numbers from NCBI database; 13 amino acid sequences were added to a custom-built FASTA file representing pertinent subtypes.

^b Calculated molecular mass.

^c Experimental molecular mass, this study.

^d S.D., standard deviation.

^e Calib, peak used for internal calibration.

^f Delta M, molecular mass, (observed - calculated).

Unmodified Apo C-I was the predominant form identified, with a lesser percentage of Apo C-I' lacking the N-terminal proline and threonine residues (Fig. 1A). Most of the apolipoproteins A-II, C-II, and C-III identified from maternal plasma corresponded to post-translationally modified isoforms (Figs. 1A and 1C). Transthyretin tetramers were also present in the 10,000–15,000 Da mass range (Fig. 1B). Both pro-Apo-CII and Apo-CII (mature, proteolytically cleaved) were present in detectable abundance (Table I, Fig. 1A) (10, 28). Both singly and doubly sialylated Apo C-III were the predominant forms of Apo C-III in maternal plasma, although an appreciable peak was detected at the predicted molecular weight for Apo C-III₀ (Fig. 1A; Table I) (2, 15). Mass assignment and spectral pattern for Apo C-III₀ and the sialylated subtypes were confirmed by combination neuraminidase and O-glycosidase enzyme treatment as described (data not shown and (8)). Linear regression analysis for Apo A-II relative to the 3 Apo A-II dimers showed a positive correlation, with slopes statistically different from zero (r² = 0.652, r² = 0.600, and r² = 0.305 for AII_d, AII_d', and AII_d″). Apo A-II' was not correlated with any of the dimerized subtypes (r² = 0.02–0.08).

Quantification of Apolipoproteins and Cholesterols in Plasma

HDL, VLDL/LDL cholesterol, Apo C-III, and Apo A-II were quantified by ELISA at 4 gestational age ranges in 5 pregnant individuals and three nonpregnant controls (Table II). Apo C-II was quantified in first and third trimesters only (Table II). In support of the current literature, apolipoproteins uniformly increased over the course of pregnancy (Table II and Supplemental Fig. S1).

Table II. Quantification of apolipoproteins and cholesterols in maternal plasma and nonpregnant controls.

	Apolipoprotein A-II (mg/dL)					Apolipoprotein A-I (mg/dL)
	Menstrual	4–12 weeks	16–22 weeks	24–28 weeks	35–38 weeks	Menstrual	4–12 weeks	35–38 weeks
Averages	25.3	23.7	30.0	32.9	34.1	93.1	104.1	241.2
S.E.	2.5	2.5	1.8	1.8	4.1	8.5	14.8	27.7
p valuea		p = 0.009					p = 0.009

	Apolipoprotein C-III (mg/dL)					Apolipoprotein C-II (mg/dL)
	Menstrual	4–12 weeks	16–22 weeks	24–28 weeks	35–38 weeks	Menstrual	4–12 weeks	35–38 weeks
Averages	10.5	9.6	16.6	24.3	41.5	8.1	8.7	12.7
S.E.	1.1	1.9	2.6	2.8	6.8	0.6	0.5	1.1
p valuea		p < 0.0001					p = 0.007

	VLDL/LDL cholesterol (mg/dL)					HDL (mg/dL)
	Menstrual	4–12 weeks	16–22 weeks	24–28 weeks	35–38 weeks	Menstrual	4–12 weeks	16–22 weeks	24–28 weeks	35–38 weeks
Averages	139.9	142.0	223.0	250.1	274.2	42.1	35.7	48.2	52.8	46.0
S.E.	18.4	15.8	15.1	27.8	23.6	5.8	8.2	6.3	5.0	5.5
p valuea		p = 0.0003					p = 0.0084

^a p values calculated by paired t-test (2 groups) or repeated measures ANOVA (4 groups) for pregnant cohorts.

SELDI-TOF-MS Relative Quantification

SELDI-TOF-MS ion intensity for each protein peak was used to determine relative abundance of the median peak amplitude for each presumptive apolipoprotein peak in patient plasma across gestational time. Ion intensities were plotted as a function of gestational age and correlation coefficients were calculated.

Pro-Apo-CII showed a modest, increasing correlation with gestational age (2B and Figs. 3A, 3B). The positive correlation between gestational age and ion intensities for mature Apo C-II (with N-terminal TQQPQQ residues proteolytically cleaved) was more striking, showing a statistically significant increase with greater interpatient variability as gestation progressed (r² = 0.342) (Fig. 2A). Mature Apo C-II composed an increasingly greater proportion of total Apo C-II (mature plus pro-Apo-CII) as gestation progressed (r² = 0.469) (Fig. 2C). Mature Apo C-II composed 6.7 ± 0.9% of total Apo C-II in early gestation and increased to 33 ± 4.5% in third trimester. By third trimester, pro-Apo-CII was 140 ± 13% of first trimester (4–12 weeks) abundance (Figs. 3A, 3B, p < 0.05). Ion intensity peak amplitudes for mature Apo C-II at 16–22, 26–28, and 36–38 weeks gestation were 252 ± 54, 312 ± 56, and 783 ± 152% higher than samplings at 4–12 weeks gestation (first trimester, p < 0.05) (Fig. 3A). Apo C-II peak amplitudes decreased relative to nonpregnant controls in the first trimester, then increased progressively over gestational time in the pregnant cohort (Fig. 3A). In contrast, pro-Apo-CII did not significantly differ from baseline first trimester until late pregnancy (140 ± 13% above first trimester at 35–38 weeks gestation) (p < 0.05) (Fig. 3B).

Fig. 3.

Repeated measures comparison of apolipoprotein subtypes by gestational age range. SELDI-TOF-MS peak amplitude (ion intensity) was plotted as a function of gestational age and normalized to 4–12 week ion intensities for each patient. Nonpregnant patient values were normalized to the average intensities of all 4–12 week intensities. A, Mature Apo C-II lacking N-terminal TQQPQQ sequence (m/z 8204.1), B, pro-Apo C-II (m/z 8915.6), C, Apo C-III₀ (m/z 8765.7), D, Apo C-III₁ (m/z 9422.3), E, Apo C-III₂ (m/z 9713.4), F, Apo A-II (cysteinylated) (m/z 8810.5), G, Apo A-II' (cysteinylated without C-terminal Gln) (m/z 8689.9). *, p < 0.05 relative to first trimester (100%), n = 11 patients at 4–5 time points per patient.

Fig. 2.

Correlation of maternal plasma Apo C-II and C-III isoforms with gestational age. SELDI-TOF-MS peak amplitude (ion intensity) was plotted as a function of gestational age. A, Mature Apo C-II lacking the N-terminal TQQPQQ sequence (m/z 8204.1) and B, pro-Apo C-II (m/z 8915.6) were plotted according to gestational age. C, SELDI-TOF-MS ion intensities for peaks corresponding to mature Apo C-II (m/z 8204.3) were plotted as percent total Apo C-II (mature + pro-Apo C-II) over gestational time. D, Apo C-III₀ (m/z 8765.7), E, Apo C-III₁ (m/z 9422.3), and (F) Apo C-III₂ (m/z 9713.4) were plotted as a function of gestational age and correlation coefficients were determined. n = 11 patients at 4–5 time points per patient.

The peak amplitudes for the a-sialylated, mono-, and di-sialylated isoforms of Apo C-III (Apo C-III₀, Apo C-III₁, and Apo C-III₂) were plotted as a function of gestational age (Figs. 2D–2F and Figs. 3C–3E). Relative abundance of Apo C-III₀ did not correlate with gestational age (r² = 0.066; Fig. 2D), whereas mono-sialylated Apo C-III₁ and di-sialylated Apo C-III₂ correlated strongly (r² = 0.805 and 0.815, Fig. 2E and Fig. 2F, respectively). Ion intensities for Apo C-III₀ and sialylated variants were normalized to Apo C-III₀ and plotted as a function of gestational age (r² = 0.725). Kruskal Wallis analysis of variance by ranks indicated a statistically significant increase in Apo C-III₁ and Apo C-III₂ over gestational time (Figs. 3D, 3E) with no significant difference in Apo C-III₀ (Fig. 3C). To quantify the elevation in Apo C-III concentrations in third trimester maternal plasma in terms of μg/ml, we generated standard curves by spiking escalating amounts of Apo C-III protein standard into first trimester maternal plasma as described (29). Calibration curves for 0, 12.5, 25.0, 50.0, and 100 μg/ml protein standard showed a linear increase in the 9421.3 Da peak at 2500 nJ laser energy (Supplemental Fig. S2). Based on interpolation to the standard curve, third trimester plasma was estimated to have 65.6 μg/ml more Apo C-III than first trimester plasma.

Relative abundance of Apo A-II increased slightly over gestational time whereas the subtype ApoA-II' showed no significant correlation with gestational age (Figs. 3F, 3G).

Comparison with Nonpregnant Controls

Apolipoprotein profiles were compared between a nonpregnant control group of n = 3 women at 3 successive stages in the menstrual cycle and pregnant cohorts (Fig. 3). Apo C-II (mature) was significantly elevated in all phases of the menstrual cycle relative to pregnant individuals at 4–12 weeks gestation (Fig. 3A). Pro-Apo C-II showed a slight but statistically insignificant increase (Fig. 3B). The sialylated variants of Apo C-III were differentially expressed in nonpregnant controls relative to first trimester (Fig. 3C–3E). Apo C-III₀ was lower in abundance in luteal and LH (luteinizing hormone) phases of the menstrual cycle relative to 4–12 weeks gestation (Fig. 3C), whereas Apo C-III₂ was elevated relative to 4–12 weeks gestation (Fig. 3E). No significant differences were observed for Apo A-II subtypes in nonpregnant and first trimester patients (Figs. 3F, 3G).

To determine whether apolipoprotein SELDI-TOF-MS pattern could be used to accurately predict gestational age range, a test set of plasma specimens derived from n = 5 patients at n = 4 sequential gestational ages per patient was prepared. The patients selected for the test set were different individuals than the training set of patients. An investigator blinded to gestational age predicted the gestational age ranges with 100% accuracy in all patients based solely on apolipoprotein variation according to the mass ranges and correlations predicted by the training set. The most useful predictors of gestational age were Apo C-III₁, Apo C-III₂, and mature Apo C-II.

Targeted MS/MS Proteomics Analysis of Apo C-III Glycovariants and Apo C-II

LTQ-Orbitrap XL tandem mass spectrometry was used to verify the results obtained from SELDI-TOF-MS profiling experiments. Identification of sialylated Apo C-III₁ and Apo C-III₂ required MS³ validation (Figs. 4A–4C and Supplement Table S2). To distinguish among the apolipoprotein subtypes in the tryptic peptides from the Orbitrap analysis, we quantified spectral counts (Supplemental Table S3) and area under the extracted ion chromatogram for only the peptides that effectively differentiate among the apolipoprotein subtypes (Supplemental Table S4). The area under the extracted ion chromatogram (AUC) for each differentiable N- or C-terminal peptide was normalized to an internal standard AQUA peptide (C-terminal tryptic peptide of Apo C-III₀ FSEFWDLDPEVRPTSA[V-C₆¹³N₂¹⁵]AA) as described, and relative abundances of pro-Apo C-II, mature Apo C-II, and the Apo C-III glycovariants were determined in first and third trimester plasma (17) (Fig. 4D, 4E, Supplement Table S4).

Fig. 4.

LTQ-Orbitrap XL MS detection of sialylated Apo C-III and Apo C-II in maternal plasma. Intact C-terminal tryptic sialylated peptides of Apo C-III₂ (m/z 1029.45, [M+H]³⁺) and Apo C-III₁ (m/z 932.51 [M+H]³⁺) were isolated in the first round of MS in the Orbitrap from maternal plasma. A, After fragmentation of m/z 1029.45 precursor in the ion trap (MS/MS), the resultant ions corresponded with m/z 1396.96, (GalNAc-Gal-conjugated, mono-sialylated Apo C-III₁), 1251.49 (GalNAc-Gal-conjugated Apo C-III₀), 1170.50 (GalNAc-conjugated Apo C-III₀), and 1069.12 (a-sialylated Apo C-III₀). B, Fragmentation of the m/z 932.51 precursor ion in the ion trap yielded ions m/z 1069.16 (a-sialylated, GalNAc-Gal-conjugated Apo C-III₀), 1170.74 (GalNAc-conjugated Apo C-III₀), and 1251.60 (GalNAc-Gal-conjugated Apo C-III₀). C, Further MS³ fragmentation of the m/z 1069.16 ion resulted in a fragmentation pattern with b/y ion ladder corresponding to the unmodified C-terminal peptide of Apo C-III₀. (open square, GalNAc [N-acetylgalactosamine]; filled circle, galactose; filled diamond, sialic acid; pep = C-terminal tryptic peptide of Apo C-III₀ (FSEFWDLDPEVRPTSAVAA). D, Relative abundances of N-terminal tryptic peptides of mature Apo C-II (m/z 746.9, [M+H]²⁺) and pro-Apo C-II (m/z 735.0, [M+H]³⁺) were estimated in maternal plasma over gestational time. E, Absolute quantification of a-sialylated Apo C-III₀ (m/z 1069.5, [M+H]²⁺⁾, C-terminal tryptic peptides mono-sialylated Apo C-III₁ (m/z 932.1, [M+H]³⁺⁾, di-sialylated C-terminal tryptic peptides of Apo C-III₂ (m/z 1029.5, [M+H]³⁺⁾ was achieved by normalization to an internal standard AQUA peptide. *, p < 0.05 by paired student's t test, first trimester relative to third trimester AQUA normalized data; n = 5 patients, 2 gestational ages each (trimester 1 [4–12 weeks] and trimester 3 [35–38 weeks]).

SELDI-TOF-MS Analysis of Apolipoproteins in Prematurity

A new cohort composed of n = 8 term and n = 3–4 preterm specimens in trimesters 1 and 2 was evaluated by SELDI-TOF-MS to test the ability of the apolipoprotein targeted mass spectrometry assay to differentiate among normal patients and patients with a significant obstetrics complication (i.e. prematurity) (Supplemental Table S5). Because all individuals in the preterm cohort delivered before 37 weeks, third trimester specimens were unavailable from this group of women. Apolipoprotein panels were evaluated in first and second trimester plasma specimens from both sets of patients. The unmodified Apo A-II cysteinylated dimer (AII_d) (m/z 17344) was significantly elevated in preterm cohorts relative to their gestationally age-matched term counterparts in both first and third trimester (59.4 ± 7.8 μA in term and 102.3 ± 6.6 μA in preterm patients in trimester 1 and 74.7 ± 6.3μA in term and 97.0 ± 3.8μA in trimester 2; Table III; Supplemental Table S5).

Table III. Comparison of ApoA-II cysteinylated dimer relative abundance by SELDI-TOF MS in maternal plasma from term relative to preterm patients (n = 10 term and n = 4 preterm patients).

	4–9 weeks gestation				16–22 weeks gestation
	Term		Preterm		Term		Preterm
	Mean	S.E.	Mean	S.E.	Mean	S.E.	Mean	S.E.
m/z	17342.3	0.8	17344.1	0.7	17341.8	2.8	17344.4	1.3
Intensity	65.6	8.5	102.3	6.6	70.3	4.8	97.0	3.8
p valuea				0.009			0.021

^a p values calculated for peak amplitude ion intensities in term relative to preterm individuals.

DISCUSSION

In this study, we use SELDI-TOF-MS and LTQ-Orbitrap targeted proteomics to isolate and semiquantitatively characterize post-translationally modified subtypes of plasma apolipoproteins throughout gestation. We demonstrate the novel findings that a mature, cleaved form of Apo C-II increases over gestational time, and provide evidence that a panel of apolipoprotein subtypes may be useful in predicting preterm birth.

To our knowledge, this study is the first to report a disproportionate increase in mature Apo C-II relative to pro-Apo C-II subtypes in pregnancy using the detection capabilities unique to mass spectrometry. Studies using SELDI-TOF-MS have reported the increase in abundance of an 8204 Da peak coincident with mature Apo C-II in clinical conditions associated with increased inflammation. Systemic IL-2 administration resulted in an increase in the 8204 Da Apo C-II peak, suggesting a correlation between mature Apo C-II and inflammatory insult (21). In normal physiology, only about 5% of Apo C-II exists in adult plasma in its cleaved, mature form (lacking the N-terminal TQQPQQ sequence) (25). Pro-Apo C-II represents the remaining 95%. Mass spectrometry-based studies confirm a proportionally smaller quantity of mature Apo C-II in plasma relative to pro-Apo-CII (28).

ELISA and immunoturbidimetric assays show that Apo A-I, C-I, and A-II increase proportionally between first and second trimesters (3, 30–33). As expected in pregnancy, we observe a significant decrease in ratios of Apo A-I:Apo A-II and Apo A-II_dimer subtypes from trimester 1 to trimester 3, and an appreciable increase in VLDL/LDL cholesterol but not HDL cholesterol. Immunological methods lack sufficient sensitivity to account for variation in apolipoprotein subtypes in pregnancy. The discrepancy between the SELDI-TOF-MS and ELISA data could also be explained by competitive binding of proteins present in high molar ratios to the array surfaces. The changes in subtypes reported in this study warrant further investigation in larger cohorts of complicated pregnancies.

Caution must be exercised when comparing proteomic profiles in physiological states such as pregnancy which are characterized by dynamic changes in plasma protein complement. Because SELDI-TOF-MS arrays preferentially bind proteins of greater molar abundance, significant changes in protein complement can artificially affect intensities of other spectral features (20). This limitation of the SELDI-TOF-MS technique may be overcome by fractionating specimens to limit the competitive inhibition of high molar abundance species in maternal plasma (34). However, one of the limitations of most of the mass spectrometry techniques described in this study is the ability to detect only relative changes in abundance. The concentrations of Apo C-III calculated by interpolation to an internal standard are lower than predicted by ELISA (an average difference of 290 μg/ml, Supplemental Table S4), suggesting that the sialyl groups caused significant attenuation of ionization in the SELDI-TOF technique. During laser desorption in mass spectrometry, variable loss of sialic acid occurs in proportion to laser energy and other factors. From a biochemical standpoint, the ELISA antibodies detect all subtypes of Apo C-III, whereas our mass spectrometry techniques examine the subtypes individually. Therefore, both the biochemical and mass spectrometry approaches have inherent limitations making it difficult to obtain accurate quantitation of individual subtypes in a given sample.

Bondarenko and colleagues demonstrated the utility of electrospray mass spectrometry in characterizing apolipoprotein subtypes by analyzing intact proteins (15, 35). More recently, novel fucosylated Apo C-III subtypes were identified by mass spectrometry analysis of Schistosoma mansoni from urine of infected individuals (36). Serum isoforms of Apo C-I, Apo C-II, Apo C-III, Apo A-II, dimerized Apo A-II, and Apo A-I were predictive markers for Alzheimers disease in MALDI-TOF-MS studies (35).

Although we acknowledge that our findings of changes in concentration for dimerized Apo A-II in preterm patients must be validated in a larger cohort, our results warrant further investigation into the use of apolipoprotein subtype analysis in obstetric populations. Shotgun proteomics strategies in obstetric cohorts have yielded some success in biomarker identification in blood for early diagnosis of pregnancy complications (6–7, 37–43), but most of these studies require prefractionation and depletion of the most abundant proteins in plasma. These preprocessing modifications risk concomitant depletion of lower abundance proteins. Acetonitrile precipitation of high abundance proteins may account for the discrepancies between the SELDI-TOF-MS and Orbitrap data in our study, particularly among the Apo A-II and Apo C-I subtypes. Further, we did not investigate differences of 2 atomic mass units among our discriminatory peptides to differentiate monomeric from dimeric Apo A-II. Our quantification methods using 10 pmol of internal standard AQUA peptides (Orbitrap) or protein standard (SELDI-TOF-MS) are also limited in that variable loss of sialic acid which could result from laser desorption (in the case of SELDI-TOF-MS), depending on factors such as the laser energy used (20). Thus, our quantification probably underestimates the true content of Apo C-III in plasma. Alternatively, the antibody used in the ELISA is not specific for multiple apolipoprotein subtypes. The elevated concentration reported in the ELISA relative to the SELDI-TOF-MS quantification could reflect greater specificity of the mass spectrometry approach in detecting unique subtypes.

In this study, we evaluate LTQ-Orbitrap XL mass spectrometry as a complementary experimental approach to estimate changes in apolipoprotein subtype abundance. Our approach relies on determining the area under the extracted ion chromatogram of differentiable peptides, thus allowing for discrimination between apolipoprotein subtypes. The accurate mass capability of the Orbitrap renders this level of specificity possible, allowing for improved differentiation among the subtypes in the Orbitrap platform.

In conclusion, this study demonstrates the utility of a SELDI-TOF-MS proteomic profiling strategy to (1) detect apolipoprotein subtypes in plasma with minimal sample manipulation and (2) observe changes in protein profile in an obstetrically relevant biological problem (prematurity). The LTQ-Orbitrap XL relative quantification and targeted identification of modified peptides provide a complementary experimental approach to SELDI-TOF-MS for the Apo C-III, Apo C-II, and Apo A-II subtypes investigated. This study reports novel evidence for the presence and fluctuation of modified isoforms of apolipoproteins in maternal plasma. In particular, we describe for the first time an increase in mature Apo C-II relative to pro-Apo C-II over gestational time. Because database searching algorithms frequently do not incorporate post translational modifications of apolipoprotein isoforms, this study provides important evidence that many isoforms of proteins might be overlooked in conventional shotgun proteomics approaches.