Objective
We analyzed the vaginal fluid proteome to identify biomarkers of intraamniotic infection among women in preterm labor.
Study Design
Proteome analysis was performed on vaginal fluid specimens from women with preterm labor, using multidimensional liquid chromatography, tandem mass spectrometry, and label-free quantification. Enzyme immunoassays were used to quantify candidate proteins. Classification accuracy for intraamniotic infection (positive amniotic fluid bacterial culture and/or interleukin-6 >2 ng/mL) was evaluated using receiver-operator characteristic curves obtained by logistic regression.
Results
Of 170 subjects, 30 (18%) had intraamniotic infection. Vaginal fluid proteome analysis revealed 338 unique proteins. Label-free quantification identified 15 proteins differentially expressed in intraamniotic infection, including acute-phase reactants, immune modulators, high-abundance amniotic fluid proteins and extracellular matrix–signaling factors; these findings were confirmed by enzyme immunoassay. A multi-analyte algorithm showed accurate classification of intraamniotic infection.
Conclusion
Vaginal fluid proteome analyses identified proteins capable of discriminating between patients with and without intraamniotic infection.
Preterm birth is a major unsolved problem in the United States, where in 2004, 1 in 8 births occurred at <37 weeks gestation. Preterm birth, which is a leading cause of neonatal morbidity and death, has actually increased in recent years, despite concerted prevention efforts. A substantial proportion of spontaneous preterm births are associated with intraamniotic infection, which is defined by the recovery of bacterial pathogens or the presence of high concentrations of proinflammatory cytokines such as interleukin-6 in amniotic fluid. The diagnosis is difficult because most women with intraamniotic infection do not have fever, uterine tenderness, or other clinical signs of infection other than preterm labor and because they tend to deliver within 48 hours of examination. Evidence is increasing that neonates who are exposed to intraamniotic infection have an increased risk of adverse outcomes (such as neonatal sepsis, respiratory distress syndrome, and intraventricular hemorrhage) when compared with nonexposed infants of similar birthweight.
For Editors’ Commentary, see Table of Contents
Currently, the definitive diagnosis of intraamniotic infection requires a transabdominal amniocentesis to perform a direct examination of amniotic fluid. However, many health care providers are reluctant to perform amniocentesis because of perceived risks such as bleeding, rupture of the fetal membranes, or increased uterine contractions.
Recent developments in proteomic analysis provide the opportunity to examine the vaginal fluid proteome as a less invasive predictor of intraamniotic infection and preterm birth. To date, this technique has been examined in a nonhuman primate model and in a small series of women in spontaneous preterm labor. We hypothesized that systematic analysis of the vaginal fluid proteome would identify sensitive, novel markers of intraamniotic infection among women in spontaneous preterm labor with intact membranes. If successful, such markers could limit the role of amniocentesis as a confirmatory test among cases with high suspicion of intraamniotic infection. A proteomic approach from vaginal fluid samples ultimately might lead to the development of less invasive tests for intraamniotic infection among women in preterm labor and potentially influence clinical treatment.
Materials and Methods
We conducted a secondary analysis of 170 archived vaginal fluid samples from a prospective observational cohort of women in spontaneous preterm labor. Participants were at gestational ages of 20-34 weeks by obstetric estimate, which was determined from menstrual dating or from the earliest available ultrasound scan. Preterm labor was defined as regular uterine contractions at a frequency of <10 minutes with either documented cervical change or a cervical dilation of >1 cm or effacement of >50%. All participants had intact membranes at study enrollment that was confirmed by sterile speculum examination. Women with cervical dilation >4 cm or ruptured membranes at admission were not eligible for study inclusion. The University of Washington Institutional Review Board approved the original study protocol, and the subjects provided written informed consent at the time of original study enrollment. The present analyses were approved administratively by the University of Washington Human Subjects Division and considered exempt from further review because they involved secondary analysis of existing, deidentified data and specimens.
At study entry (after speculum and cervical digital examination), amniotic fluid from all participants was obtained by transabdominal amniocentesis; vaginal fluid was obtained by the saturation of a Dacron swab with fluid from the posterior vaginal fornix. Amniotic and vaginal fluid specimens were stored at −70°C in pyrogen-free containers until assayed. Amniotic and vaginal fluid bacterial cultures, demographic and reproductive history data, and pregnancy outcomes were available from the original study cohort. Women received tocolytics, corticosteroids, and antibiotics according to the judgment of their clinical providers. Amniotic fluid interleukin-6 concentrations were determined by commercial enzyme immunoassay (Genzyme Diagnostics, Cambridge, MA). Intraamniotic infection was defined by an “expanded gold standard” as a positive amniotic fluid bacterial culture and/or interleukin-6 >2 ng/mL, as previously reported. Early preterm birth was defined as delivery at ≤34 weeks gestation, because most neonatal morbidity occurs at this gestational age.
Of 220 archived vaginal fluid samples that were available from the original study cohort, 43 samples had insufficient protein (<0.17 μg/μL), and 7 samples did not have detectable human albumin and/or detectable concentrations for at least 8 of the 15 biomarkers of interest, which left 170 samples to be included in these analyses. There were no clinically important differences in subject characteristics or pregnancy outcomes between included and excluded samples. The 170 samples that were included were from 30 subjects with intraamniotic infection (12 subjects with a positive amniotic fluid bacterial culture and 18 subjects with a negative bacterial culture and interleukin-6 concentration >2 ng/mL) all of whom had an early preterm birth (intraamniotic infection group), 55 subjects who had an early preterm birth without intraamniotic infection (early preterm birth group), and 85 subjects without intraamniotic infection who had symptoms of preterm labor but delivered at >34 weeks gestation (preterm labor group).
Previously published approaches that were used for mass spectrometric analysis are briefly described later. Pools of vaginal fluid samples were created from 100 μL each of 21 randomly selected samples from each group or 63 total samples (intraamniotic infection, early preterm birth, and preterm labor). Pooled samples were subjected to 2-dimensional liquid chromatography with tandem mass spectrometry and label-free quantification to identify differentially expressed proteins between the groups.
For mass spectrometry, 600 μg protein from each pooled sample was digested with trypsin, and the resulting peptides were first separated with the use of strong cation exchange column into 32 fractions. These fractions were analyzed with an Agilent 1100 liquid chromatographer that was connected to a tandem mass spectrometer (Thermo Finnegan, San Jose, CA). A total of 101,377 mass spectra were collected that represented the 3 subject groups.
Peptides that were present in each sample were identified by a search of the corresponding mass spectra against a protein database that contained forward and reverse entries of the Swiss-Prot human database (version 46.6) with the use of 2 independent search engines: TurboSequest (Thermo Finnegan) and X! Tandem (The Global Proteome Organization; www.thegpm.org/tandem/index.html ). Peptide identifications from a sample were assembled into protein identifications with Scaffold software (version1.3.2; Proteome Software, Portland, OR). Protein identifications that had at least 2 independent peptide identifications were considered to be present in the sample.
The total number of mass spectra that were matched to a particular protein was used to assess the relative abundance of a protein in a sample with the use of a label-free quantification method that had been described previously. This method compares the spectral counts of a protein between 2 samples with either an independent 2 × 2 χ 2 test or a Fisher’s Exact test. Proteins that passed the quantification method with a probability value of ≤ .05 were considered to be expressed differentially between the samples. Fold changes of differentially expressed proteins were determined with a reference formula for calculating spectral count ratios.
Confirmation of the presence of differentially expressed biomarkers of interest was validated by enzyme-linked immunosorbent assays for 15 candidate biomarkers on all 170 individual samples, by previously described techniques. Available commercial antibodies and antigens were purchased from various vendors to prepare immunoassays. Standard curves were developed with the use of known quantities of recombinant proteins or standards provided by manufacturer to reference sample concentrations. All assays were performed on 100 μL samples of vaginal fluid in triplicate. Interassay and intraassay coefficient of variations ranged from 3-7%. The mean protein concentration across triplicates was used in subsequent analyses.
One-way analyses of variance were conducted to compare natural log-transformed enzyme immunoassay values of samples from women in 2 groups: those with intraamniotic infection vs those without infection (early preterm birth and preterm labor groups combined). For presentation, we transformed the mean log value back to original units (geometric mean and geometric standard deviation). We applied the Bonferroni correction to account for multiple comparisons (15 proteins were compared). We evaluated the classification performance of several different combinations of 2, 3, or 4 proteins using logistic regression models. Predicted values (ie, risk scores) were computed for each woman from each multianalyte model, and receiver operating characteristic (ROC) curves were plotted. Areas under the ROC curves with 95% confidence intervals (CIs) were calculated with bootstrap methods. Proteins were selected for multianalyte models based on their ability to discriminate individually between patients with intraamniotic infection and no intraamniotic infection and/or based on the improvement in classification that was observed by adding them to the models. Descriptive and inferential statistics were computed with SAS software (version 9.1: SAS Institute Inc, Cary, NC); ROC curves were produced and compared with customized STATA modules (Stata Corp, College Station, TX).
Results
Maternal and pregnancy characteristics that were stratified by infection status and timing of delivery are presented in Table 1 . Maternal age and race and occurrence of other pregnancy complications (such as twin gestation) were comparable across groups. The 30 subjects with intraamniotic infection were seen and delivered at an earlier gestational age, were more likely to have bacterial vaginosis, and had a substantially shorter time between enrollment and delivery than the 140 subjects without infection.
| Characteristic | Intraamniotic infection (n = 30) | No intraamniotic infection | P value a | ||
|---|---|---|---|---|---|
| All (n = 140) | Early preterm birth (n = 55) | Preterm labor (n = 85) | |||
| Median maternal age, y | 25 | 24 | 26 | 24 | .47 |
| Maternal race, n (%) | .38 | ||||
| White | 20 (67) | 88 (63) | 37 (67) | 51 (60) | |
| African American | 7 (23) | 24 (17) | 8 (15) | 16 (19) | |
| Other | 3 (10) | 28 (20) | 10 (18) | 18 (21) | |
| Parity, n (%) | 12 (40) | 83 (59) | 33 (60) | 50 (59) | .05 |
| Previous delivery ≤34 wk, n (%) b | 2 (17) | 29 (35) | 13 (39) | 16 (33) | .32 |
| Twin gestation, n (%) | 4 (13) | 19 (14) | 10 (18) | 9 (11) | .97 |
| Bacterial vaginosis, n (%) | 7 (24) | 17 (12) | 6 (11) | 11 (13) | .009 |
| Current cigarette smoker, n (%) | 3 (10) | 24 (17) | 17 (32) | 7 (8) | .42 |
| Mean gestational age at enrollment, wk | 28 | 32 | 30 | 32 | .0004 |
| Mean gestational age at delivery, wk | 28 | 35.5 | 32 | 37 | < .0001 |
| Days between enrollment and delivery, n | 3 | 21 | 6 | 32 | < .0001 |
a Compares subjects with intraamniotic infection (n = 30) with those with no intraamniotic infection (n = 140); values were based on χ 2 tests for categoric variables, on independent samples t tests for normally distributed variables (mean presented), and on Wilcoxon (nonparametric) tests for nonnormally distributed variables (median presented);
We detected 338 unique proteins in vaginal fluid. Functional annotation of the vaginal fluid proteome using GeneOntology terms (DAVID; version 2.1; david.abcc.ncifcrf.gov/ ) revealed that most of these proteins were associated with metabolism (34%) and immune response (18%). Significant pair-wise differences were seen in relative abundance between pooled samples from women with intraamniotic infection, compared with the early preterm birth or preterm labor groups, for the 26 vaginal fluid proteins by spectral count analysis ( Table 2 ). Twenty proteins were expressed differentially between those patients with intraamniotic infection and those with preterm labor, and 18 proteins were expressed differentially between the intraamniotic infection and early preterm birth groups. These differentially abundant proteins included acute-phase reactants (such as alpha-1-acid glycoprotein), immune modulators (such as calgranulin C and cystatin A), high-abundance amniotic fluid proteins (such as insulin-like growth factor binding protein-1 and vitamin D binding protein), and extracellular matrix–signaling factors (such as fatty acid binding protein). In general, acute-phase reactants and amniotic fluid proteins were more abundant and extracellular matrix–signaling proteins less abundant among those with intraamniotic infection ( Table 2 ). Immune modulators such as calgranulin C were more abundant with intraamniotic infection, and some such as cystatin A were relatively depleted in the context of intraamniotic infection. Although changes in protein abundance were in a similar direction for both the preterm labor and the early preterm birth groups, when compared with those with intraamniotic infection, total fibronectin was less abundant among those with intraamniotic infection when compared with those with early preterm birth and more abundant among those with intraamniotic infection when compared with those with preterm labor.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
