Spontaneous preterm birth significantly contributes to the overall neonatal morbidity associated with preterm deliveries. Nearly 50% of cases are associated with microbial invasion of the amniotic cavity followed by an inflammatory response. Robust diagnostic tools for neonates jeopardized by infection and inflammation may thus decrease the overall neonatal morbidity substantially. Amniotic fluid retrieved during labor retains fetal and pregnancy-related protein fingerprint and its sampling does not place any unwanted stress on women. Using exploratory and targeted methods we analyzed proteomes of amniotic fluid sampled at the end of spontaneous preterm labor prior to delivery from women with and without infection and inflammation. Exploratory data indicated several amniotic fluid proteins to be associated with infectious-inflammatory complications in spontaneous preterm birth. LC-SRM analysis subsequently verified statistically significant changes in lipocalin-1 (
Despite the progress in perinatal medicine and the increasing knowledge on the risk factors and underlying mechanisms associated with preterm birth (i.e., childbirth occurring in less than 37 completed weeks of gestation [
Several studies have shown that microbial invasion of the amniotic cavity (MIAC) is one of the risk factors for spontaneous preterm labor [
During the last decade, mass spectrometry-based proteomics has become a valued tool for conducting comprehensive analyses in biological and clinical samples [
The method of choice for conducting accurate and sensitive measurements of protein concentrations in clinical samples is to use antibody-based assays, such as ELISAs. Unfortunately, antibodies or validated immunoassays for new marker candidates are often not readily available. SRM is a versatile LC-MS/MS based technique with the potential to achieve accurate and fairly sensitive protein quantification [
Here, we perform exploratory and targeted proteomic analyses of amniotic fluid to reveal and verify proteins potentially useful for detection of newborns jeopardized by intra-amniotic infection and inflammation. The samples were collected at the end of the first stage of sPTB followed by delivery in women with and without both MIAC and HCA. Amniotic fluid retrieved during labor represents a reasonable diagnostic material as it fully retains a fetal and pregnancy-related protein fingerprint, but in contrast to transabdominal amniocentesis, the method and timing of the collection do not place any unwanted stress on the mother. To the best of our knowledge our study represents the first attempt of a comprehensive proteomic analysis of this unique clinical specimen.
Samples were collected between 2008 and 2011 as a part of a major study to examine the genetic and biomarker changes associated with sPTB and to determine the factors involved in the racial disparity between African Americans and Caucasians. This study was approved by the Institutional Review Board of Tristar Nashville and the Western Institutional Review Board of the Perinatal Research Center, Nashville, TN. All samples were collected after obtaining written informed consent from all participants.
All of the women included in the study were recruited at the Department of Obstetrics and Gynecology of Centennial Women’s Hospital, Nashville, TN, USA. Amniotic fluid samples were collected from the forebag via needle puncture of intact membranes at the end of first stage of labor. Amniotic fluid was centrifuged immediately for 10 min at 2,000 g and aliquots were stored at −70°C until analysis. Samples were supplemented with protease inhibitors (Roche Diagnostics, Basel, Switzerland) and filtered through a 0.22
sPTB was defined as the occurrence of regular uterine contractions at a minimum frequency of two contractions
MIAC was determined based on a positive aerobic/anaerobic cultivation of amniotic fluid or PCR detection of bacterial 16s rRNA. Histological examination of the placenta, the fetal membranes, and the umbilical cord was performed in all cases. The degree of polymorphonuclear leukocyte infiltration was assessed separately in the free membranes, in the placenta, and in the umbilical cord. The diagnosis of HCA was conducted based on the presence of diffuse neutrophil infiltration in the chorion decidua, chorionic plate, amnion, and umbilical cord.
All chemicals were from Sigma Aldrich in the highest grade available if not indicated otherwise (Sigma, St. Louis, MO). The solvents used for HPLC with UV detection were of HPLC grade, and those used for HPLC hyphenated to a mass spectrometer were of LC-MS grade (Sigma).
An overview on how the samples were processed for exploratory iTRAQ analysis is depicted in Figure
Sample processing for the exploratory iTRAQ analysis. Representative MIAC and HCA-positive and negative samples were created and processed as two duplicates. Samples were depleted from 14 high-abundance proteins and digested. Resulting peptides were iTRAQ labeled, combined, and processed according to the CysTRAQ protocol. Cysteinyl and noncysteinyl iTRAQ peptides were further fractionated using high pH reversed-phase chromatography. Eventually, each fraction was analyzed using LC-MALDI-MS/MS.
The 14 most abundant plasma proteins that could be anticipated to also be present in amniotic fluid at high concentrations were removed from the prepared representative samples using a MARS Hu-14 column (Agilent) in an Alliance 2695 HPLC system (Waters, Milford, MA), according to the manufacturer’s instructions. MARS buffer A was exchanged three times for water using 3 kDa cut-off Amicon Ultra filters. The retentates were collected, and the protein concentration was determined via BCA.
From each depleted sample, 200
Desalted cysteinyl and noncysteinyl peptide fractions were redissolved in 200
Each high pH reversed-phase fraction was dissolved in 40
The spectra were evaluated with ProteinPilot 2.0.1 software (AB Sciex) using the Paragon search algorithm, Pro Group algorithm, and the integrated FDR analysis function [
An overview on how the samples were processed for LC-SRM verification is depicted in Figure
Sample processing for the LC-SRM verification. Each sample was subjected to immunoaffinity depletion of 14 high-abundance proteins and digested by trypsin. Portion of each sample was used to create a global internal standard (GIS). GIS peptides were dimethylated by formaldehyde-d2 whereas peptides in each sample were labeled by formaldehyde. GIS was spiked in each individual sample prior to LC-SRM analysis.
For each sample, 65
Peptides in individual 50
Individual light-labeled samples were mixed with an equal amount of GIS, acidified by TFA and incubated for 30 min at 37°C to hydrolyze RapiGest, which was subsequently removed via centrifugation. Samples were desalted using 3 M Empore C18-SD 4 mm/1 mL cartridges and vacuum dried. Prior to the analysis, samples were redissolved in 38
The LC-SRM assay development, including method refinement, and final data analysis were all performed in Skyline software (University of Washington, Seattle, WA) [
First of all, development samples were used to build an MS/MS library of dimethylated peptides to facilitate and accelerate SRM transitions design. Each development sample (~1
Eluted peptides were electrosprayed into a Q-Exactive mass spectrometer using a Nanospray Flex ion source (Thermo Scientific, Bremen, Germany). The full MS/Top12 experimental setup was used. Positive ion full scan MS spectra (
Precursor ions of charge state ≥ 2 and threshold intensity of 1 × 105 counts were selected for HCD fragmentation, with an exclusion window of 60 s. The isolation window of 2 Da and normalized collision energy of 27 were used. Each product ion spectrum was acquired in the Orbitrap at 17 500 resolution, with a 1 × 106 AGC target and a maximum 60 ms injection time.
MS/MS spectra were searched in Proteome Discoverer software (Thermo Scientific) using MASCOT engine (Matrix Science, London, UK). The tryptic specificity was set to full and two missed cleavages were allowed. The mass tolerance was set to 10 ppm for precursors and 20 mmu for fragment ions. Oxidized Met and dimethylation of N-terminus and of Lys were set as dynamic modifications while Cys thiomethylation was set as a fixed modification. MS/MS spectra explained with a cut-off score 0.80 were extracted by the Skyline software from the MASCOT results and stored in an MS/MS library.
The FASTA sequences of proteins of interest were retrieved from UniProt and imported into the Skyline. The following constraining criteria were applied on peptide selection: digestion: trypsin (cleavage at Arg or Lys unless Pro is the next amino acid); no missed cleavage allowed; peptide length: between 6 and 25 amino acids; excluding the first five N-terminal amino acids; structural modifications: methylthio at Cys and dimethylation of N-terminus and Lys. Peptides sharing sequence with additional human protein(s) were excluded; Leu and Ile were treated as interchangeable. Only peptides matching the criteria and present in the MS/MS library were considered.
At the transition level, the criteria were as follows: precursor charge: 2 and 3; fragment charge: 1; ion types: y and b; product ions ranging from (
Peptides were then screened for detectability in development amniotic fluid samples using the Agilent 1260 LC system coupled to an Agilent 6490 Triple Quadrupole mass spectrometer, controlled by MassHunter acquisition software. For screening purposes, 8
Peptides detected as single chromatographic peak formed by overlapping transition signals, showing high similarity degree of transition signals with peak intensities in the MS/MS spectra, were considered for final SRM method. These peptides were subsequently used for retention time prediction, which was done for peptides with Arg and Lys separately because two methyl groups on Lys residues significantly influence retention on reversed-phase. Additional peptides detected in a predicted retention time window, with overlapping transition signals and with high similarity degree of transition and MS/MS signals, were considered as well. For each protein, the most intense precursor of three different peptides was included in the final LC-SRM method. Peptides without Met and Cys, peptides without Gln at N-terminus, peptides without Asn in a close proximity to a Gly, peptides with Lys at N-terminus, and doubly charged precursors of Arg peptides were strongly preferred. Up to five best-performing transitions were selected
Final amniotic fluid analyses were performed in two technical replicates, using 8
Recorded LC-SRM chromatograms were evaluated in Skyline software. The peak integration was done automatically using Savitzky-Golay smoothing. All peaks were manually inspected to confirm proper automatic selection and integration. Maximum of two most intense transitions
The minimum number of samples for the exploratory phase (
The four-dimensional nature of the analysis led to recording of 26.230 MS/MS spectra, identifying 10.864 distinct peptides at a maximum 1% FDR. Based on these peptides, 690 amniotic fluid proteins were identified at a maximum 1% FDR. Of these, 22 noncontaminant proteins significantly differed (
A candidate list for subsequent verification was compiled from proteins showing significantly altered abundance, potentially due to the presence of MIAC and HCA. The candidates were selected using the following criteria: only proteins with an identification FDR lower than 1% that showed a statistically significant change in abundance (
List of protein candidates for verification.
|
Unusedb | Access. # | Name | Peptides |
iTRAQ |
iTRAQ |
iTRAQ |
iTRAQ |
AVG |
---|---|---|---|---|---|---|---|---|---|
318 | 11.71 | P0CG48 | Polyubiquitin-C |
|
|
|
|
|
|
374 | 10.00 | P09228 | Cystatin-SA |
|
|
|
|
|
|
154 | 27.52 | P31025 | Lipocalin-1 |
|
|
|
|
|
|
110 | 35.41 | P03973 | Antileukoproteinase |
|
|
|
|
|
|
438 | 7.70 | P59666 | Neutrophil defensin 3 |
|
|
|
|
|
|
225 | 17.96 | P09466 | Glycodelin |
|
|
|
|
|
|
73 | 44.36 | P00915 | Carbonic anhydrase 1 |
|
|
|
|
|
|
167 | 25.81 | P07358 | Complement component C8 beta chain |
|
|
|
|
|
|
8 | 175.57 | P02452 | Collagen alpha-1(I) chain |
|
|
|
|
|
|
115 | 34.38 | P60174 | Triosephosphate isomerase |
|
|
|
|
|
|
37 | 73.59 | P60709 | Actin, cytoplasmic 1 |
|
|
|
|
|
|
228 | 17.59 | P43490 | Nicotinamide phosphoribosyltransferase |
|
|
|
|
|
|
16 | 113.54 | Q9HC84 | Mucin-5B |
|
|
|
|
|
|
385 | 9.51 | Q9BRK3 | Matrix-remodeling-associated protein 8 |
|
|
|
|
|
|
238 | 16.30 | P20930 | Filaggrin |
|
|
|
|
|
|
237 | 16.31 | P14384 | Carboxypeptidase M |
|
|
|
|
|
|
bProteinPilot protein score.
cNumber of iTRAQ peptides used for quantification of the protein.
dAverage protein iTRAQ ratios (MIAC and HCA positive versus MIAC and HCA negative).
eSignificance level for protein iTRAQ ratios.
fExtremity of the iTRAQ change calculated as the absolute log2 of the average iTRAQ ratio.
To facilitate development of a LC-SRM assay for verification of the exploratory iTRAQ findings, we first compiled an MS/MS spectra library of dimethylated peptides from which appropriate surrogate peptides were selected. For all but filaggrin at least one peptide matched to an MS/MS spectrum and totally 106 nonrepeated peptides were assigned to an MS/MS spectrum. All peptides found for polyubiquitin-C shared sequence with polyubiquitin-B, fusion proteins ubiquitin-60S ribosomal protein L40, and ubiquitin-40S ribosomal protein S27a. Because these proteins were reported by ProteinPilot in the same protein group with almost identical score and are of the same functionality they were kept in the Skyline document. Peptides of neutrophil defensin 3 (shared with neutrophil defensin 1) and actin, cytoplasmic 1 (shared with actin, cytoplasmic 2), were treated in the same way. Two of three peptides of lipocalin-1 had shared sequences with putative lipocalin-1-like protein 1. Existence of this protein is however very uncertain (information available in the neXtProt database) and peptides were thus kept for subsequent refinement. As a result, 95 peptides were screened for detectability in immunodepleted but additionally not fractionated development samples using 524 transitions of 112 precursors. Based on the obtained intensities and preferences towards most reliable peptides we set up the final method covering 15 proteins using 38 peptides, 76 precursors, and a total of 354 transitions.
Seven peptides did not provide any transition free of interference and two peptides showed low signal stability of the heavy form. Transitions used for quantification are listed in Table S3 in Supplementary Materials. Four proteins were assayed using single peptide, 8 proteins using two peptides and three proteins using three peptides (Table
LC-SRM assay results.
Acc. # | Protein name (peptide)a | Peptidesb | Concordancec |
|
|
Median ratio to GIS | Median ratio to GIS |
|
---|---|---|---|---|---|---|---|---|
Positive | Negative | Positive | Negative | |||||
P0CG48 | Polyubiquitin-C | 3 | 49 | 26 | 23 | 1.40 | 1.04 | 0.083 |
P09228 | Cystatin-SA | 1 | — | 27 | 23 | 0.53 | 1.09 | 0.321 |
P31025 | Lipocalin-1 | 2 | 46 | 26 | 20 | 0.99 | 1.61 | 0.047 |
P03973 | Antileukoproteinase | 1 | — | 27 | 23 | 0.49 | 0.28 | 0.087 |
P59666 | Neutrophil defensin 3 | 2 | 3 | — | — | — | — | — |
(IPACIAGER) | — | — | 27 | 23 | 0.15 | 0.07 | — | |
(YGTCIYQGR) | — | — | 27 | 23 | 0.45 | 0.15 | — | |
P09466 | Glycodelin | 2 | 42 | 23 | 19 | 1.64 | 0.49 | 0.013 |
P00915 | Carbonic anhydrase 1 | 2 | 50 | 27 | 23 | 0.58 | 0.78 | 0.756 |
P07358 | Complement component C8 beta chain | 2 | 36 | 19 | 17 | 1.07 | 0.95 | 0.568 |
P02452 | Collagen alpha-1(I) chain | 3 | 0 | — | — | — | — | — |
(ICVCDNGK) | — | — | 27 | 23 | 0.43 | 0.36 | — | |
(VLCDDVICDETK) | — | — | 27 | 23 | 0.79 | 0.64 | — | |
(SLSQQIENIR) | — | — | 27 | 23 | 0.97 | 1.25 | — | |
P60174 | Triosephosphate isomerase | 3 | 49 | 27 | 22 | 1.22 | 0.78 | 0.169 |
P60709 | Actin, cytoplasmic 1 | 2 | 37 | 19 | 18 | 0.70 | 0.42 | 0.076 |
P43490 | Nicotinamide phosphoribosyltransferase | 2 | 39 | 24 | 20 | 0.55 | 0.19 | 0.018 |
Q9HC84 | Mucin-5B | 2 | 35 | 22 | 15 | 0.94 | 0.91 | 0.805 |
Q9BRK3 | Matrix-remodeling-associated protein 8 | 1 | — | 27 | 23 | 1.07 | 1.06 | 0.832 |
P14384 | Carboxypeptidase M | 1 | — | 27 | 23 | 0.94 | 1.29 | 0.613 |
bNumber of peptides used for LC-SRM quantification of particular protein.
cNumber of samples in which concordance between corresponding peptides was below RSD of 20%.
dNumber of samples that fulfilled all requirements and were subjected to statistical analysis.
e
From 15 proteins analyzed in individual samples using LC-SRM coupled with dimethyl labeling, data for all but two proteins were admitted to statistical analysis (Table
Amniotic fluid concentration of lipocalin-1, glycodelin, and nicotinamide phosphoribosyltransferase relative to GIS and respective ROC curves. In concordance with the iTRAQ exploratory findings, women without MIAC and HCA had higher amniotic fluid lipocalin-1 levels than women in whom both conditions were confirmed whereas levels of glycodelin and nicotinamide phosphoribosyltransferase were higher in women with MIAC and HCA. Some measurements did not pass the rigorous evaluation criteria for reporting only the most confident data, leading to slightly different numbers of observations subjected to statistical analysis.
Development of diagnostic tests to reduce inflammatory complications linked to sPTB is highly topical for decreasing overall perinatal morbidity. Previous biomarker discovery studies, addressing proteome alterations in amniotic fluid from sPTB patients with and without intra-amniotic infection and inflammation, were performed using samples obtained transabdominally at admission, that is, shortly after diagnosing preterm labor [
To describe and confirm the changes in the amniotic fluid proteome related to the presence or absence of both MIAC and HCA at the end of the first stage of sPTB we employed global iTRAQ and targeted LC-SRM quantification techniques in a two-stage MS-based workflow (Figure
Global and targeted quantification techniques used for the exploratory and verification phase as demonstrated on lipocalin-1 peptide GLSTESILIPR. (a) MALDI-TOF/TOF MS/MS spectra of iTRAQ-labeled peptides were used for protein identification and (b) for global quantification based on the iTRAQ reporter ions 114–117. (c) For LC-SRM verification, a library of HCD MS/MS spectra of dimethylated peptides was compiled first. SRM transitions for each peptide were then selected based on the spectral information in the library (e.g., peptide GLSTESILIPR was quantified by fragments y9 and
First, we performed an exploratory analysis in representative amniotic fluid samples and employed CysTRAQ quantitation technology in combination with a multidimensional fractionation and separation, as described previously [
The obtained set of 690 amniotic fluid proteins identified at a 1% FDR (1019 proteins at a 5% FDR) represents one of the most detailed descriptions of the amniotic fluid proteome. Proteins assessed previously in amniotic fluid using ELISA in the tens of ng/mL (myeloperoxidase) or few ng/mL ranges (cathelicidin and M-CSF) were among the proteins identified at a 1% FDR [
In the second phase, we attempted to verify exploratory findings using LC-SRM in each individual amniotic fluid sample to convert the proteomic significance to a biostatistical significance. Rather than developing of the LC-SRM assay using synthetic stable isotope labeled peptides, we employed reductive dimethyl labeling of peptides using either formaldehyde or formaldehyde-d2. This straightforward and cost-effective technique for introduction of stable isotopes into peptides has been used for protein quantification in exploratory proteomic studies [
Although the nanoconfiguration is still preferred in mass spectrometry-based proteomics due to higher sensitivity, we performed LC-SRM using a narrow-bore column and a standard ESI probe. The advantages of a higher flow rate and standard ESI in terms of separation reproducibility and ionization stability represent a very compelling alternative, especially when focusing on a high-throughput protein quantitation platform [
Among the dysregulated proteins detected in the exploratory iTRAQ analysis, neutrophil defensin 3 (P59666), which can be cleaved to produce neutrophil defensin 2, was found to be altered following the same trend as reported by Buhimschi et al. [
Lipocalin-1 (P31025) was found by Bujold et al. to be overproduced in amniotic fluid from women exhibiting preterm labor without intra-amniotic infection or inflammation who subsequently delivered at term [
LC-SRM on dimethylated peptides of glycodelin (P09466) confirmed its significantly increased level in amniotic fluid of women with MIAC and HCA (
Nicotinamide phosphoribosyltransferase (P43490) is the third protein with increased abundance in amniotic fluid of women with MIAC and HCA that was confirmed independently in the verification phase of the study (
As outlined above, the design of the exploratory phase based on representative pooled samples can be considered a limitation of our study. However, even using the latest proteomic MS-based technologies, it would be very difficult to obtain such a detailed description of the amniotic fluid proteome across 57 samples through analysis of each sample individually. Because we are aware of constraints associated with the pooled strategy (e.g., a few outliers may significantly alter the average concentration in a pooled sample), in the second essential phase of the study, we attempted to assign biostatistical confidence to the top dysregulated proteins by means of LC-SRM. In this stage of the project, seven samples were unfortunately no longer available, which might have had an effect on the obtained statistical significances.
In the present study, we attributed certain potential for uncovering MIAC and HCA to several proteins from amniotic fluid that was retrieved at the end of the first stage of sPTB. Statistically different levels were verified for lipocalin-1, glycodelin, and nicotinamide phosphoribosyltransferase using LC-SRM on dimethylated peptides. These proteins are thus interesting targets for an independent validation in larger cohorts as potential peripartum markers for inflammatory complications in spontaneous preterm birth.
We revealed and verified significant change in levels of lipocalin-1, glycodelin, and nicotinamide phosphoribosyltransferase due to the presence of microbial invasion of the amniotic cavity and histological chorioamnionitis. Assessment of these proteins in amniotic fluid from women with spontaneous preterm labor sampled prior to delivery might thus help identify women and premature newborns who would benefit from specific perinatal and postnatal interventions.
The authors declare that there is no conflict of interests regarding the publication of this paper.
This work was supported by a grant from the Ministry of Education, Youth and Sports (ME 10025) and a grant from the Internal Grant Agency of the Ministry of Health, Czech Republic (NT/13599). The purchase of the Agilent 6490 Triple Quadrupole mass spectrometer was covered by the MH CZ - DRO (UHHK, 00179906). Salary of Vojtech Tambor was covered by Postdoctoral Fellowship Project no. 714 CZ.1.07/2.3.00/30.0012.