Proteomics Analysis of Human Amniotic Fluid

Amniotic fluid is a dynamic and complex mixture that reflects the physiological status of the developing fetus. In this study, the human amniotic fluid (AF) proteome of a 16–18-week normal pregnancy was profiled and analyzed to investigate the composition and functions of this fluid. Due to the complexity of AF, we utilized three different fractionation strategies to provide greater coverage. Two types of two-dimensional LC/MS/MS as well as an LC-SDS-PAGE-LC-MS/MS platform were used. A total of 16 AF samples between gestational ages of 16 and 18 weeks from women carrying chromosomally normal fetuses were analyzed by one of the three fractionation methods followed by a common reverse phase LC-MS/MS step. Mascot and The Global Proteome Machine engines were used to search the International Protein Index human database for peptide sequence identification. The list of proteins was generated by combining the results of both engines through the PeptideProphet of Scaffold software. All identified proteins were combined to generate the AF proteome comprising 1,026 unique gene matches or 842 non-redundant proteins. This list includes most of the currently used biomarkers for pregnancy-associated pathologic conditions such as preterm delivery, intra-amniotic infection, and chromosomal anomalies of the fetus. The subcellular localization, tissue expression, functions, and networks of the AF proteome were analyzed by various bioinformatic tools. These data will contribute to the better understanding of amniotic fluid function and to the discovery of novel biomarkers for prenatal diagnosis of fetal abnormalities. Amniotic fluid is a dynamic and complex mixture that reflects the physiological status of the developing fetus. In this study, the human amniotic fluid (AF) proteome of a 16–18-week normal pregnancy was profiled and analyzed to investigate the composition and functions of this fluid. Due to the complexity of AF, we utilized three different fractionation strategies to provide greater coverage. Two types of two-dimensional LC/MS/MS as well as an LC-SDS-PAGE-LC-MS/MS platform were used. A total of 16 AF samples between gestational ages of 16 and 18 weeks from women carrying chromosomally normal fetuses were analyzed by one of the three fractionation methods followed by a common reverse phase LC-MS/MS step. Mascot and The Global Proteome Machine engines were used to search the International Protein Index human database for peptide sequence identification. The list of proteins was generated by combining the results of both engines through the PeptideProphet of Scaffold software. All identified proteins were combined to generate the AF proteome comprising 1,026 unique gene matches or 842 non-redundant proteins. This list includes most of the currently used biomarkers for pregnancy-associated pathologic conditions such as preterm delivery, intra-amniotic infection, and chromosomal anomalies of the fetus. The subcellular localization, tissue expression, functions, and networks of the AF proteome were analyzed by various bioinformatic tools. These data will contribute to the better understanding of amniotic fluid function and to the discovery of novel biomarkers for prenatal diagnosis of fetal abnormalities. Amniotic fluid (AF) 1The abbreviations used are: AF, amniotic fluid; AFP, α-fetoprotein; FPLC, fast protein liquid chromatography; Ig, immunoglobulin; IPI, International Protein Index; SAX, strong anion exchange; SCX, strong cation exchange; 2DE, two-dimensional electrophoresis; bis-Tris, 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol; hCG, human chorionic gonadotropin. 1The abbreviations used are: AF, amniotic fluid; AFP, α-fetoprotein; FPLC, fast protein liquid chromatography; Ig, immunoglobulin; IPI, International Protein Index; SAX, strong anion exchange; SCX, strong cation exchange; 2DE, two-dimensional electrophoresis; bis-Tris, 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol; hCG, human chorionic gonadotropin. is a complex and dynamic biological fluid that provides mechanical protection and nutrients as well as other molecules required for fetal growth and well being. Therefore, both the quantitative and qualitative integrity of AF are essential for normal development of the human fetus during pregnancy. During embryogenesis, AF is initially formed from maternal plasma that passed through fetal membranes. Because free diffusion occurs bidirectionally between the AF and the fetus across fetal skin, placenta, and umbilical cord from 10 to 20 weeks of gestation, AF composition becomes similar to that of fetal plasma during this period. Therefore, analysis of AF composition before skin keratinization that occurs between 19 and 20 weeks of gestation would reveal valuable information that may indicate physiological or pathological conditions of the fetus. AF contains water, proteins, peptides, carbohydrates, lipids, hormones, and electrolytes. Among these components, studies have been done on the amino acids in AF, such as taurine, glutamine, and arginine, and some trophic mediators such as epidermal growth factor, transforming growth factor α and β-1, and insulin-like growth factor I (1Underwood M.A. Gilbert W.M. Sherman M.P. Amniotic fluid: not just fetal urine anymore.J. Perinatol. 2005; 25: 341-348Crossref PubMed Scopus (439) Google Scholar). Additionally expression levels of several proteins and cytokines were studied using immunoassays. For example, the levels of biomarkers for the most common chromosomal anomaly, Down syndrome, such as α-fetoprotein (AFP), were thoroughly investigated both in maternal serum and in AF. Nevertheless little is known about the functions of these proteins or the majority of constituents of AF. The collective profile of AF proteins has not as yet been assembled. Thus, little is known regarding the possible molecular interactions of proteins and their contributions in fetal development. Therefore, it is desirable to identify more proteins to explore their functions within the AF. Recent technological advances in proteomics have been actively utilized to investigate AF proteins to better understand this complex biological fluid and to discover disease-specific biomarkers. Consequently many putative markers for such anomalies as premature rupture of amnion, intra-amniotic infection, and Down syndrome have been reanalyzed or newly discovered (2Vuadens F. Benay C. Crettaz D. Gallot D. Sapin V. Schneider P. Bienvenut W.V. Lemery D. Quadroni M. Dastugue B. Tissot J.D. Identification of biologic markers of the premature rupture of fetal membranes: proteomic approach.Proteomics. 2003; 3: 1521-1525Crossref PubMed Scopus (112) Google Scholar, 3Thadikkaran L. Crettaz D. Siegenthaler M.A. Gallot D. Sapin V. Iozzo R.V. Queloz P.A. Schneider P. Tissot J.D. The role of proteomics in the assessment of premature rupture of fetal membranes.Clin. Chim. Acta. 2005; 360: 27-36Crossref PubMed Scopus (58) Google Scholar, 4Tsangaris G.T. Karamessinis P. Kolialexi A. Garbis S.D. Antsaklis A. Mavrou A. Fountoulakis M. Proteomic analysis of amniotic fluid in pregnancies with Down syndrome.Proteomics. 2006; 6: 4410-4419Crossref PubMed Scopus (95) Google Scholar). However, none of these biomarkers have a high enough individual detection rate to be used on their own. Thus, the discovery of more efficient biomarkers that have a higher detection rate and specificity is highly desirable. Proteomics analysis of AF, therefore, could be an ideal first step in efforts to elucidate changes related to pathological conditions in the fetus. Proteomic profiles of AF have been generated by several groups using different methods since 1997 (5Liberatori S. Bini L. De Felice C. Magi B. Marzocchi B. Raggiaschi R. Frutiger S. Sanchez J.C. Wilkins M.R. Hughes G. Hochstrasser D.F. Bracci R. Pallini V. A two-dimensional protein map of human amniotic fluid at 17 weeks' gestation.Electrophoresis. 1997; 18: 2816-2822Crossref PubMed Scopus (50) Google Scholar, 6Nilsson S. Ramstrom M. Palmblad M. Axelsson O. Bergquist J. Explorative study of the protein composition of amniotic fluid by liquid chromatography electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry.J. Proteome Res. 2004; 3: 884-889Crossref PubMed Scopus (39) Google Scholar, 7Park S.J. Yoon W.G. Song J.S. Jung H.S. Kim C.J. Oh S.Y. Yoon B.H. Jung G. Kim H.J. Nirasawa T. Proteome analysis of human amnion and amniotic fluid by two-dimensional electrophoresis and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.Proteomics. 2006; 6: 349-363Crossref PubMed Scopus (72) Google Scholar, 8Michel P.E. Crettaz D. Morier P. Heller M. Gallot D. Tissot J.D. Reymond F. Rossier J.S. Proteome analysis of human plasma and amniotic fluid by off-gel isoelectric focusing followed by nano-LC-MS/MS.Electrophoresis. 2006; 27: 1169-1181Crossref PubMed Scopus (100) Google Scholar, 9Tsangaris G.T. Kolialexi A. Karamessinis P. Anagnostopoulos A.K. Antsaklis A. Fountoulakis M. Mavrou A. The normal human amniotic fluid supernatant proteome.In vivo. 2006; 20: 479-490PubMed Google Scholar). More specifically, Liberatori et al. (5Liberatori S. Bini L. De Felice C. Magi B. Marzocchi B. Raggiaschi R. Frutiger S. Sanchez J.C. Wilkins M.R. Hughes G. Hochstrasser D.F. Bracci R. Pallini V. A two-dimensional protein map of human amniotic fluid at 17 weeks' gestation.Electrophoresis. 1997; 18: 2816-2822Crossref PubMed Scopus (50) Google Scholar) identified 31 proteins by two-dimensional electrophoresis (2DE) followed by postseparation analysis techniques such as N-terminal sequencing. Nilsson et al. (6Nilsson S. Ramstrom M. Palmblad M. Axelsson O. Bergquist J. Explorative study of the protein composition of amniotic fluid by liquid chromatography electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry.J. Proteome Res. 2004; 3: 884-889Crossref PubMed Scopus (39) Google Scholar) were the first to use MS for profiling AF. They used LC-Fourier transform-ion cyclotron resonance MS to identify 58 proteins from AF. They were also the first to deplete albumin from AF to identify more proteins. In 2006, three groups analyzed normal AF supernatant. Park et al. (7Park S.J. Yoon W.G. Song J.S. Jung H.S. Kim C.J. Oh S.Y. Yoon B.H. Jung G. Kim H.J. Nirasawa T. Proteome analysis of human amnion and amniotic fluid by two-dimensional electrophoresis and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.Proteomics. 2006; 6: 349-363Crossref PubMed Scopus (72) Google Scholar) reported 37 proteins by using 2DE followed by matrix-assisted laser desorption/ionization time-of-flight MS; Michel et al. (8Michel P.E. Crettaz D. Morier P. Heller M. Gallot D. Tissot J.D. Reymond F. Rossier J.S. Proteome analysis of human plasma and amniotic fluid by off-gel isoelectric focusing followed by nano-LC-MS/MS.Electrophoresis. 2006; 27: 1169-1181Crossref PubMed Scopus (100) Google Scholar) identified 69 proteins from albumin-depleted AF by Off-Gel™ electrophoresis/LC-MS/MS; and Tsangaris et al. (9Tsangaris G.T. Kolialexi A. Karamessinis P. Anagnostopoulos A.K. Antsaklis A. Fountoulakis M. Mavrou A. The normal human amniotic fluid supernatant proteome.In vivo. 2006; 20: 479-490PubMed Google Scholar) reported 136 proteins by 2DE followed by matrix-assisted laser desorption/ionization-MS/MS. These different groups not only utilized different approaches as well as different protein databases but also used different levels of stringency for protein identification, making it difficult to assess the accuracy of each data set. For example, the group who identified the highest number of proteins searched their queries against the collective protein database and consequently identified many non-human proteins from human AF (9Tsangaris G.T. Kolialexi A. Karamessinis P. Anagnostopoulos A.K. Antsaklis A. Fountoulakis M. Mavrou A. The normal human amniotic fluid supernatant proteome.In vivo. 2006; 20: 479-490PubMed Google Scholar). In this study, our objective was to generate an extensive profile of the normal human AF proteome. Similar to other biological fluids, AF contains a few major proteins that complicate proteome profiling. Depletion of these proteins certainly improves protein identification. However, the "sponge effect" of these major proteins (and especially of albumin) could lead to the loss of many low abundance proteins that may be valuable biomarkers. Therefore, we designed three complementary platforms to profile the AF proteome to maximize the number of proteins identified. As a result, we report here the most extensive list of normal human amniotic fluid proteins comprising 1,026 unique gene matches from at least 754 (and probably up to 842) different genes. The large number of the AF proteins allowed qualitative bioinformatics analysis and may provide the basis for further compositional and functional studies of AF. Human AF samples (8–10 ml) were obtained by amniocentesis from women at 16–18 weeks of gestation undergoing prenatal diagnosis mostly due to advanced maternal age ranging from 30 to 45 years after written informed consent. AF samples were centrifuged to collect amniocytes for cytogenetic analysis, and the cell-free supernatants were stored at −80 °C until use. 16 samples from chromosomally normal pregnancies were chosen randomly. The samples were fractionated by one of three methods as follows. Three samples were fractionated by strong anion exchange (SAX) liquid chromatography, another three were fractionated by strong cation exchange (SCX) liquid chromatography, and 10 samples were pooled together and fractionated by LC-SDS-PAGE (see Fig. 1). Samples were thawed and filtered through a 0.22-μm × 25-mm syringe-driven filter unit (Millipore). The pooled sample only was also filtered through an Amicon 50-kDa-cutoff centrifugal ultrafiltration device (Millipore), and the filtrate was collected. All samples were depleted for IgG with a Protein A/G column (Bio-Rad). Samples were then dialyzed at 4 °C using a membrane with 3.5-kDa molecular mass cutoff (Spectra/Por) as follows: for 12 h in 5 liters of 1 mm ammonium bicarbonate buffer with one buffer exchange in the same buffer for the SCX approach and in 5 liters of 20 mm Tris buffer, pH 9.6, for the SAX approach. For the SAX and the LC-SDS-PAGE approaches, the dialyzed AF samples were directly loaded onto an HR10/10 column packed with SOURCE15Q media for strong anionic exchange (GE Healthcare). Fractionation was performed using a fast performance liquid chromatography (FPLC) system using 50 mm Tris-HCl, pH 9.6, as the running buffer and 1 m NaCl as the elution buffer for 1 h using a linear gradient at a flow rate of 3 ml/min. A total of 12 fractions were collected, dialyzed in 5 liters of 1 mm ammonium bicarbonate buffer for 12 h, and lyophilized to dryness for trypsin digestion. Each lyophilized sample was denatured using 8 m urea, reduced with 200 mm dithiothreitol at 50 °C, and alkylated in 500 mm iodoacetamide at room temperature in a dark room. The sample was desalted using a NAP5 column (GE Healthcare). The sample was then lyophilized and resuspended in trypsin buffer (1:50, trypsin:protein concentration; 120 μl of 50 mm ammonium bicarbonate, 100 μl of methanol, 150 μl of H2O) overnight at 37 °C (Promega, sequencing grade modified porcine trypsin). The sample was lyophilized to dryness. For the SCX approach only, the dried peptides were resuspended in 120 μl of mobile phase A (0.26 m formic acid in 10% ACN) to be injected into the HPLC system. Briefly each of the 12 fractions was dissolved in SDS-PAGE loading buffer, DTT was added, and samples were heated at 80 °C for 5 min and loaded in a single lane on a 1-mm-thick 4–12% bis-Tris gel (Invitrogen). After separation, the gel was stained with SimplyBlue™ SafeStain (Invitrogen). The intensely stained bands below 50 kDa were excised separately, and all of the in-between areas were cut into slices, resulting in a total of 10–12 slices per fraction. In-gel digestion was performed as follows. Bands were washed in 200 μl of 30% methanol for 5 min, and then 200 μl of 100% ACN was applied for 10 min. The bands were incubated in 10 mm DTT in 100 mm NH4HCO3 for 30 min at 60 °C. The DTT solution was removed, and the bands were incubated with 200 μl of 50 mm iodoacetamide in 100 mm NH4HCO3 for 30 min in the dark. The iodoacetamide was then removed, and washes were performed with 500 μl of distilled water followed by addition of 100 μl of ACN. Then ACN was removed, and 50 μl of the 0.01 μg/μl trypsin solution was added. Proteins were digested overnight at 37 °C. For this approach, the samples were first digested with trypsin to generate peptides. The digested peptides, resuspended in 120 μl of mobile phase A, were directly loaded onto a PolySULFOETHYL A™ column (The Nest Group, Inc.) containing hydrophilic anionic polymer (poly(2-sulfoethyl aspartamide)) with a pore size of 200 Å and a diameter of 5 μm. Fractionation was performed using an HPLC system (Agilent 1100) for 1 h at a flow rate of 200 μl/min. 1 m ammonium formate and 0.26 m formic acid in 10% ACN (mobile phase B) were added in a linear gradient. A protein cation exchange standard (Bio-Rad) was applied before each run to evaluate column performance. The eluent was monitored by UV absorbance at 280 nm. A total of 12 or 24 fractions were collected and lyophilized to dryness. Each fraction from all fractionation schemes was resuspended in 80 μl of 95% water, 0.1% formic acid, 5% ACN, 0.02% trifluoroacetic acid (Buffer A) and desalted using a ZipTip C18 pipette tip (Millipore). The peptides were eluted in 4 μl of 90% ACN, 0.1% formic acid, 10% water, 0.02% trifluoroacetic acid (Buffer B), and 80 μl of Buffer A was added on top. Half of this volume (40 μl) was loaded on an Agilent 1100 HPLC system by the autosampler onto a 2-cm C18 trap column (inner diameter, 200 μm), and the peptides were eluted onto a resolving 5-cm analytical C18 column (inner diameter, 75 um). The 120-min gradient began at 0% Buffer B at 15 μl/min for 5 min and then changed to 40% Buffer B for 103 min with a linear gradient, then to 65% Buffer B for 4 min, and finally to 100% Buffer B for 13 min. The peptides were subjected to nanoelectrospray ionization followed by MS/MS in an LTQ two-dimensional linear ion trap (Thermo Scientific) coupled on line to the HPLC system. The resulting spectra from each fraction were searched separately against the IPI human database Version 3.16 by two database search engines: Mascot, Version 2.1.03 (Matrix Science) and The Global Proteome Machine Version 2.0.0.4 (Beavis Informatics Ltd.). The following parameters were used: (i) enzyme, trypsin; (ii) one missed cleavage allowed; (iii) fixed modification, carbamidomethylation of cysteines; (iv) variable modification, oxidation of methionine; (v) peptide tolerance, 3.0 Da; and (vi) MS/MS tolerance, 0.4 Da. The files were run on Scaffold Version 01_05_19 (Proteome Software). All DAT files, from Mascot, that were searched for all fractions from a single chromatography run (HPLC or FPLC) were loaded together as one "biological sample" within Scaffold, and all XML files, from The Global Proteome Machine, were loaded together as another biological sample. 95% peptide identification probability and 80% protein identification probability were used as the cutoffs for Scaffold, excluding proteins identified with lesser probability. The sample reports were exported to Microsoft Excel from Scaffold, and relevant information and annotations for each protein were searched from databases including Swiss-Prot, Human Protein Reference Database, Entrez Gene, and the Plasma Proteome Database. The false-positive error rate was calculated by analyzing all files with the same method except against a "sequence-reversed" IPI human database. The false-positive rate (FPR) was calculated as: FPR = number of false peptides/(number of true peptides + number of false peptides). Complex biological fluids such as serum or AF contain thousands of proteins ranging from high to low abundance. The major challenge for analyzing such proteomes is posed by high abundance proteins, especially albumin. In the case of serum, nearly 50% of the protein content is comprised by albumin. We verified by total protein and albumin analysis that albumin comprises nearly 70% of the protein content of AF with immunoglobulins (Igs) being the second most abundant fraction (data not shown). When Igs were removed by Protein A/G beads, the number of identified proteins increased by 38% (data not shown). Because removal of the bulk of Igs is straightforward, this step was included in all three fractionation protocols (Fig. 1). Albumin, unlike Igs, cannot be readily removed, and it also binds other proteins and peptides. Therefore, three different fractionation protocols were used, and different variations were tried within the context of each method (Fig. 1) to extend the depth of proteomics analysis. The data are summarized below. Three multidimensional separation methods coupled with MS/MS were used (Fig. 1). Raw files were searched by both Mascot and X!Tandem for more confident identification. To generate a statistically valid list of proteins, Scaffold was used to accommodate differences in algorithm and score calculation of the two search engines (10Keller A. Nesvizhskii A.I. Kolker E. Aebersold R. Empirical statistical model to estimate the accuracy of peptide identifications made by MS/MS and database search.Anal. Chem. 2002; 74: 5383-5392Crossref PubMed Scopus (3797) Google Scholar). The resulting lists of proteins from each run on Scaffold were combined into one collective list of proteins. Also all output files were searched with the reverse IPI human database, yielding the false-positive rate of ∼3%. To select only distinct proteins, we applied various elimination steps to generate the final non-redundant list. 1) We sorted the entries by IPI number and removed redundancies. 2) We removed all keratin entries (contaminants). 3) All Ig chains were removed because they are of lesser significance for proteome analysis, and they significantly increase redundancy and hinder statistical analysis of the proteome. Despite our effort to remove Igs by Protein A/G affinity chromatography, ∼11% of all initial entries were attributed to Igs. This is not surprising because a similar trend has been observed with the plasma proteome (11Anderson N.L. Polanski M. Pieper R. Gatlin T. Tirumalai R.S. Conrads T.P. Veenstra T.D. Adkins J.N. Pounds J.G. Fagan R. Lobley A. The human plasma proteome: a nonredundant list developed by combination of four separate sources.Mol. Cell. Proteomics. 2004; 3: 311-326Abstract Full Text Full Text PDF PubMed Scopus (746) Google Scholar). 4) Because one protein may have multiple IPI numbers, we sorted the proteins by their names and molecular weights and removed identical ones. After cleaning, the list included 1,026 proteins with distinct IPI number, name, and molecular weight (supplemental data: unique gene matches). Different matches of one gene may reflect biologically significant different gene products such as splice variants, sequence variants, and cleavage products. In our method such a distinction is often difficult because protein prediction is based on peptide sequence searches. Therefore, a more stringent method was used to include only non-redundant proteins. We searched for gene names of each entry and removed all but one entry with the highest number of unique sequences. We acknowledge that this approach may remove some legitimate protein variants. Our final list included 842 proteins from 754 distinct genes and 88 proteins from uncharacterized genes (supplemental data: unique genes). Among the 842 unique proteins, 445 were identified with two or more unique peptides. Among 397 proteins that were identified by one unique peptide, at least 67 proteins were identified multiple times via more than one of the three methods. In total, at least 512 proteins were identified with high confidence, and the rest (330 proteins) were identified with at least 80% probability. Of the 512 high confidence proteins, 424 were identified for the first time in AF (supplemental data: unique genes). Of 842 non-redundant proteins, 167 (20%) were identified via all three fractionation platforms (Fig. 2). Most, but not all, of the 167 proteins were of high abundance. Other proteins of high abundance were detected by two platforms. For example, α1B-glycoprotein (molecular mass, ∼80 kDa) and α1-antichymotrypsin (molecular mass, ∼65 kDa) were identified with over 10 unique peptides from each of the SAX-FPLC and SCX-HPLC platforms, but they were not detected by the SAX-FPLC-SDS-PAGE platform. This is likely due to the fact that the latter method was designed to remove albumin and other proteins of molecular mass of 50 kDa or higher. 195 (23%) of the proteins were identified by two platforms. The rest of the proteins (480 proteins, 57%) were identified by one platform (Fig. 2). This finding by no means reflects the reproducibility of our approach because the three different fractionation protocols were designed to complement rather than reproduce each other. Each identified protein was assigned a subcellular localization based on information from Swiss-Prot, Entrez Gene, and Gene Ontology databases. When one protein is known to be localized in more than one cellular compartment, all of the categories were counted non-exclusively. Fig. 3 shows the cellular distribution of 558 identified proteins with known localization. The majority are extracellular (42%) and membrane (26%) proteins. By searching the Plasma Proteome Database we found that 304 (36%) of the 842 proteins have also been found in plasma. This does not mean that the remaining 538 proteins are exclusive to AF because the plasma proteome list is still growing. Tissue expression of each protein was searched from Swiss-Prot, Entrez Gene, and Gene Ontology databases. 10 functional categories were selected based on the number of hits per organ. When one protein is expressed in more than one tissue, only the major tissue of expression was counted (otherwise we would have identified too many tissues for the majority of proteins). Fig. 4 shows 301 proteins with known tissue expression information (supplemental data: unique genes). Some of the organs to which many proteins were attributed include kidney, placenta, lung, liver, and heart. 24 proteins were specifically annotated as being expressed from embryonic organs/tissues (supplemental data: unique genes). We utilized Ingenuity (Ingenuity Systems) to retrieve known functions of each protein. 221 (of 842) were matched with functions. Because one protein may have multiple functions we selected the functions with p value <0.015. Fig. 5 shows the top 15 of 72 different functions based on significance. Major categories included cellular movement, development and function of organs, cellular growth and proliferation, cancer, and cell-to-cell signaling. More specific functions and the names of genes for each function are shown in the supplemental data. For example, the function "cancer" includes specific functions such as apoptosis, cell cycle regulation, and migration. Assignment of biological processes and subsequent construction of networks was done using Ingenuity software. We found that at least 227 different proteins are components of existing molecular networks. A total of 27 networks were constructed. One with the highest score is shown in Fig. 6. This particular network shows 35 genes that work together for cardiovascular system development, and proteins of all of these 35 genes were found in our AF proteome. A table with functions, involved genes, and significance scores of all 27 networks is presented in the supplemental data. We combined human AF proteome entries from previous publications to generate one list (2Vuadens F. Benay C. Crettaz D. Gallot D. Sapin V. Schneider P. Bienvenut W.V. Lemery D. Quadroni M. Dastugue B. Tissot J.D. Identification of biologic markers of the premature rupture of fetal membranes: proteomic approach.Proteomics. 2003; 3: 1521-1525Crossref PubMed Scopus (112) Google Scholar, 4Tsangaris G.T. Karamessinis P. Kolialexi A. Garbis S.D. Antsaklis A. Mavrou A. Fountoulakis M. Proteomic analysis of amniotic fluid in pregnancies with Down syndrome.Proteomics. 2006; 6: 4410-4419Crossref PubMed Scopus (95) Google Scholar, 5Liberatori S. Bini L. De Felice C. Magi B. Marzocchi B. Raggiaschi R. Frutiger S. Sanchez J.C. Wilkins M.R. Hughes G. Hochstrasser D.F. Bracci R. Pallini V. A two-dimensional protein map of human amniotic fluid at 17 weeks' gestation.Electrophoresis. 1997; 18: 2816-2822Crossref PubMed Scopus (50) Google Scholar, 6Nilsson S. Ramstrom M. Palmblad M. Axelsson O. Bergquist J. Explorative study of the protein composition of amniotic fluid by liquid chromatography electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry.J. Proteome Res. 2004; 3: 884-889Crossref PubMed Scopus (39) Google Scholar, 7Park S.J. Yoon W.G. Song J.S. Jung H.S. Kim C.J. Oh S.Y. Yoon B.H. Jung G. Kim H.J. Nirasawa T. Proteome analysis of human amnion and amniotic fluid by two-dimensional electrophoresis and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.Proteomics. 2006; 6: 349-363Crossref PubMed Scopus (72) Google Scholar, 8Michel P.E. Crettaz D. Morier P. Heller M. Gallot D. Tissot J.D. Reymond F. Rossier J.S. Proteome analysis of human plasma and amniotic fluid by off-gel isoelectric focusing followed by nano-LC-MS/MS.Electrophoresis. 2006; 27: 1169-1181Crossref PubMed Scopus (100) Google Scholar, 9Tsangaris G.T. Kolialexi A. Karamessinis P. Anagnostopoulos A.K. Antsaklis A. Fountoulakis M. Mavrou A. The normal human amniotic fluid supernatant proteome.In vivo. 2006; 20: 479-490PubMed Googl