In perinatal medicine, our concept of amniotic fluid has evolved from a simple electrolyte solution made up of fetal urine and sloughed cells to a complex biological fluid that performs multiple functions. In prior work we have investigated amniotic fluid cell-free fetal RNA as a source of gene expression information about normal second-trimester development.1 Other studies from our group have demonstrated unique transcriptomic differences for trisomies 18 and 21, thus providing insights into their pathophysiology and generating hypotheses for further studies.2–4
To date, the vast majority of studies on amniotic fluid cell-free fetal RNA have been performed on second-trimester pregnancies with clinical indications for an invasive procedure.5 During the first 20 weeks of pregnancy, there is bidirectional diffusion between the fetus and the amniotic fluid across the unkeratinized fetal skin. Human studies have shown that the majority of cells and cellular fragments in second-trimester amniotic fluid are derived from fetal squamous epithelial surfaces (namely skin, gastrointestinal tract, genitourinary tract) and not from the amniotic epithelium.6 More recently, amniotic fluid has become recognized as a source of fetal stem cells.7 Many aspects of the biology of amniotic fluid cell-free fetal RNA are unknown, but it is probable that cell-free transcripts enter amniotic fluid directly from fetal organs as well as from free-floating cells and cellular fragments. In the second half of pregnancy, the composition and circulation of amniotic fluid are dramatically altered by keratinization of the fetal skin, increased fetal urine output, and increased fetal swallowing and respiratory movements. Although gene expression from amniotic fluid supernatant has been shown to vary by gestational age,8 this has not been explored in the third trimester. Given the profound changes in fetal organ development and amniotic fluid composition between the second trimester and term,9 we hypothesized that the cell-free transcripts in term amniotic fluid will reflect changes associated with fetal maturation.
In this study we investigated the third-trimester fetus by comparing amniotic fluid cell-free fetal RNA from normal-term pregnancies with second-trimester controls. Specifically, we aimed to determine whether the relative representation of specific organs in term amniotic fluid cell-free fetal RNA would differ from that in second-trimester amniotic fluid using a bioinformatics tool for tissue expression analysis and a gene expression atlas. We also hypothesized that functional genomic analysis of upregulated transcripts in term amniotic fluid would identify pathways and genes that reflect the real-time biology of the third trimester and provide novel information about the term fetus.
MATERIALS AND METHODS
This was a prospective study of women undergoing planned cesarean delivery between 37 and 40 weeks of gestation at Tufts Medical Center, Boston, Massachusetts, and a comparison group of women who underwent second-trimester amniocentesis. The Tufts Medical Center institutional review board approved the study. For the cesarean delivery group, exclusion criteria included gestational age less than 37 or greater than 41 weeks, diagnosis of congenital anomaly, fetal growth restriction, multiple pregnancy, onset of labor, prelabor rupture of membranes, antepartum hemorrhage, preeclampsia, and pre-existing maternal medical condition. One woman with gestational diabetes was included in the study.
Amniotic fluid was collected at cesarean delivery after entry into the uterus and before rupture of the amniotic membranes. Approximately 10 mL of amniotic fluid was aspirated through the intact membranes using a blunt plastic cannula attached to a 20-mL syringe. Specimens with gross maternal blood or meconium contamination were discarded. Amniotic fluid samples were centrifuged at 350 g for 10 minutes at 4°C. The supernatant samples were spun at 1,600 g for 10 minutes at 4°C to remove residual vernix and then stored at −80°C.
Second-trimester amniotic fluid supernatant samples were obtained as discarded residual clinical samples from the Cytogenetics Laboratory at Tufts Medical Center, also under an institutional review board-approved protocol. Only samples from structurally normal fetuses with euploid karyotypes were included. These samples were spun at 165 g for 10 minutes at room temperature to remove amniocytes for diagnostic testing. The residual fluid was stored at −20°C for up to 1 week and then archived at −80°C.
All included amniotic fluid samples were collected between November 2011 and June 2012. RNA was extracted from 5–10 mL of amniotic fluid supernatant within 4 months of collection according to a customized protocol.10 All samples were processed using the Qiagen Circulating Nucleic Acid kit with an on-column DNase digestion step to remove genomic DNA. The RNA was then purified and concentrated with the RNeasy MinElute Clean up kit and eluted in RNase-free water. RNA was converted to cDNA and amplified using the Ovation Pico WTA kit V2 and then purified with the QIAquick PCR Purification kit. cDNA yield and purity were measured with the Nanodrop 1000 Spectrophotometer. Five micrograms of cDNA from each sample were biotinylated and fragmented using the Encore Biotin Module and hybridized to a whole human genome expression array (Affymetrix GeneChip Human Genome U133 Plus 2.0).
Normalization was performed with the three-step command from the AffyPLM package in BioConductor using ideal mismatch background and signal adjustment, quantile normalization, and the Tukey biweight summary method.11 This summary method included a logarithmic transformation to improve the normality of the data. We identified genes that were differentially regulated in term amniotic fluid compared with second-trimester amniotic fluid samples using the independent t test and adjusted for multiple testing using the Benjamini-Hochberg correction.12 We defined genes as significantly differentially regulated if the P value using the Benjamini-Hochberg correction was <.01. Our microarray data sets are publicly available in the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/).
We used the BioGPS Gene Expression Atlas at http://biogps.org to identify upregulated genes that had tissue-specific expression patterns. This atlas of the human protein-encoding transcriptome contains the gene expression profiles of 78 normal human tissues.13 The BioGPS Gene Expression Atlas allowed us to assess which organs were specifically represented in the 10 most upregulated genes at term and the second trimester. We categorized genes as highly organ-specific if they mapped to a single organ with an expression value greater than 30 multiples of the median in accordance with a previously established stringency threshold.14
We also used the tissue expression function in the Database for Annotation, Visualization and Integrated Discovery (DAVID) to identify tissues that most highly expressed the genes upregulated in term and second trimester fetuses, respectively (http://david.abcc.ncifcrf.gov/).15 The DAVID integrates tissue expression databases from publicly available sources to allow the identification of the most enriched gene expression patterns across hundreds of normal and diseased tissues for any given gene list. For each resource, the expression values for a given gene in all tissues are ranked from greatest to least. All tissues with an expression value in the top quartile are associated with that gene and an enrichment P value called the EASE score (a modified Fisher’s exact P value) is calculated for that tissue.16 The DAVID also calculates a more conservative P value using the Benjamini-Hochberg correction to control the false discovery rate of family-wise enriched terms. We considered results significant if the tissue term was enriched by twofold or more with a Benjamini-Hochberg-corrected P<.05. We report the results derived from the UniProt database (www.uniprot.org).
Functional analysis was performed using IPA 9.0 software. This pathways analysis tool is a manually curated database that identifies overrepresented biological processes in a given data set and calculates a significance score for each result using the right-tailed Fisher’s exact test. The Affymetrix probe set identifiers for the significantly differentially regulated genes were uploaded to the pathways analysis software with their corresponding median fold change and Benjamini-Hochberg-corrected P values.
We used the pathways analysis software to identify any physiologic systems or molecular and cellular functions that were enriched in term fetuses by focusing on the genes that were significantly upregulated by at least fourfold in the term group. To reduce false-positive results, we applied the Benjamini-Hochberg correction for multiple comparisons within the pathways analysis as we were examining all potentially significant biological categories. We considered pathways significant if they had a corrected P<.05.
We also performed an upstream regulator analysis with the pathways analysis software using the list of all significantly upregulated and downregulated genes with a fold change value of 1.5 or more or −1.5 or less. This analysis uses prior published knowledge of expected effects between upstream regulators and their target genes to identify which upstream regulators are predicted to be activated or inhibited based on the direction of gene expression in the data set. The activation z-score is used to infer likely activation states of upstream regulators based on comparison with a model that assigns random regulation directions. We reported the bias-corrected activation z scores for upstream regulators that corrects for any skews in the direction of gene regulation in the data set or in the direction of molecular network relationships for a particular upstream regulator. We considered results with an activation z score greater than 2 or less than −2 as statistically significant according to thresholds recommended by the pathways analysis software. An overview of our analytical methods is provided in Figure 1.
We obtained amniotic fluid samples from eight term and eight second-trimester pregnancies. The median gestational ages were 38 weeks (range 37–39 weeks) and 17 weeks (range 16–19 weeks) in the term and second-trimester groups, respectively. The most common indication for second-trimester amniocentesis was increased risk of trisomy 21 (Appendix 1, available at http://links.lww.com/AOG/A387). The maternal age of the second-trimester group was significantly higher than the term group (mean age 37 years compared with 33 years, P=.04, independent t test). There were equal proportions of male to female fetuses in each group.
The average microarray hybridization rate for all samples was 41% (range 33.0–50.1%). There were a total of 2,871 genes that were significantly differentially regulated in term compared with second-trimester amniotic fluid; 1,307 genes were upregulated and 1,564 were downregulated (Appendix 2, available at http://links.lww.com/AOG/A388).
The 10 most upregulated genes by fold change value in the term group were dominated by lung and saliva-specific transcripts (Table 1). All five pulmonary surfactant protein genes (SFTPA1, SFTPA2, SFTPB, SFTPC, SFTPD) were significantly upregulated in term amniotic fluid. In contrast, the top 10 genes with known functions that were most upregulated in second-trimester amniotic fluid were associated with the lower gastrointestinal tract (Appendix 3, available at http://links.lww.com/AOG/A389).
The DAVID tissue expression analysis of the genes upregulated in term amniotic fluid showed statistically significant enrichment of transcripts highly expressed by saliva glands, trachea, and kidney. In contrast, amniotic fluid transcripts upregulated in the second trimester were highly enriched for fetal brain and Cajal-Retzius (embryonic neuronal) cells (Table 2).
There were 609 well-annotated genes that were upregulated by at least fourfold in term compared with second-trimester amniotic fluid. Core pathways analysis of these genes showed enrichment of multiple physiologic systems involved in newborn functions such as immune defense, eating, and breathing (Table 3 and Appendix 4, available at http://links.lww.com/AOG/A390).
The three most statistically significant molecular and cellular functions upregulated in term fetuses were carbohydrate metabolism (41 genes), cellular movement (139 genes), and lipid metabolism (93 genes) (Appendix 5, available at http://links.lww.com/AOG/A391). Within the lipid metabolism category were endocrine processes known to be involved in parturition such as phospholipid concentration and prostaglandin synthesis. Individual upregulated genes included the key enzyme in inducible prostaglandin synthesis, prostaglandin–endoperoxidase synthase 2 (also known as cyclooxygenase 2), phospholipase A2, and the genes for prostaglandin E receptors types 3 and 4 (PTGER3, PTGER4).
We identified 17 upstream regulators in the pathways analysis that were significantly upregulated in term neonates and predicted to be activated based on the differential regulation of target genes within our data set (Appendix 6, available at http://links.lww.com/AOG/A392). Of interest were several genes involved in proinflammatory pathways and immune activation such as nuclear factor-kappaB, interleukin 8, interleukin-1 β, and the interleukin 6 receptor. There were six upstream regulators that were statistically significantly predicted to be activated in the second trimester. These included genes involved in thyroid function, oxidative stress, and liver development (Appendix 7, available at http://links.lww.com/AOG/A393).
The results of our analysis of term compared with second-trimester amniotic fluid demonstrate upregulation of fetal maturation processes that normally occur at term, supporting our hypothesis that amniotic fluid cell-free fetal RNA reflects real-time developmental fetal physiology.
The major strength of this functional analysis of term amniotic fluid cell-free fetal RNA is the detailed biological interpretation in the context of normal fetal maturity. High-dimensional biology techniques such as functional genomics and proteomics have been applied to study abnormal pregnancy using a variety of maternal and fetal samples.17,18 In contrast, there are very few “-omic” studies of amniotic fluid that aim to characterize normal third-trimester fetal physiology. One proteomic study specifically examining the effect of gestational age on amniotic fluid failed to yield any significant information on the third trimester.19 A metabolomic study on second- and third-trimester amniotic fluid samples identified metabolites associated with increasing gestational age.20 Neither of these studies attempted to provide any detailed biological interpretation of the differentially expressed proteins in the third compared with the second trimester.
Gene expression atlas mapping and the DAVID tissue expression analysis of genes upregulated at term showed a dominance of lung, upper gastrointestinal tract, and renal transcripts. It is not surprising that the organs that actively secrete or excrete into the amniotic cavity appear to be the major sources of amniotic fluid cell-free RNA at term. The Database for Annotation, Visualization and Integrated Discovery tissue expression analysis also demonstrated that the relative abundance of cell-free transcripts associated with the fetal nervous system varies according to gestational age with more nervous system-related transcripts present at the second trimester compared with at term.
Many of the physiologic systems upregulated at term were highly specific for adaptations required after birth, including the digestive system and respiratory function. Preparation for postnatal energy requirements and thermal regulation was also suggested by the upregulation of carbohydrate metabolism and adipogenesis. Free radical scavenging was one of the enriched cellular and molecular functions at term, which is consistent with the positive linear association between gestational age and antioxidant capacity.21
One of the striking features of the term amniotic fluid results was the evidence of preparation for labor in the term pregnancies through significant upregulation of known inflammatory and endocrine pathways. Normal term labor is associated with acute inflammation of fetal membranes in the absence of overt infection.22 One of the fetal signals believed to initiate normal term labor is the increased production of surfactant protein A, which leads to macrophage activation and migration and subsequent nuclear factor-kappaB and interleukin-8 production in mice.23 We report here significant results for both immune cell trafficking function and SFTPA1 and SFTPA2 upregulation in term amniotic fluid, which fits this proposed fetal signaling model. The downstream effects analysis also identified increased activity of three proinflammatory genes with key roles in labor, nuclear factor-kappaB, interleukin-1β, and interleukin-8.
Of the endocrine processes involved in the onset of labor, lipid metabolism and prostaglandin synthesis were significantly upregulated functions in term amniotic fluid. The increase in prostaglandin production before labor is attributable largely to increased activity of PTGS2, which was upregulated 15-fold in term amniotic fluid. Taken in total, our findings are very consistent with current models of the immune and endocrine influences on parturition.24
One of the challenges of studying amniotic fluid cell-free RNA is ascertaining the tissue(s) of origin of transcripts. There is currently no model that allows us to observe transcripts trafficking from specific fetal organs into the amniotic cavity in vivo. We therefore adopted an “in silico” approach using bioinformatics tools to obtain information on the tissue expression profiles in amniotic fluid cell-free fetal RNA. We took care to avoid maternal cellular contamination during sample collection so that we could be confident that the RNA was only fetal in origin. Although maternal microchimerism (the presence of maternal cells in the fetus) is a recognized phenomenon, we feel that any potential contribution of these rare maternal cells to our functional analysis of amniotic fluid would be negligible. We also cannot exclude the theoretical possibility of cell-free transcripts in maternal plasma entering amniotic fluid through unknown routes. However, because we limited our core pathways analysis to those transcripts upregulated by at least fourfold at term, it is unlikely that any such maternal transcripts would have a major effect on results.
Our cross-sectional study design was a result of logistic constraints related to patient recruitment and sample collection. Ideally, we would have used a longitudinal design with paired second-trimester and term amniotic fluid samples from the same pregnancies, but this was impractical to implement in our single-center study. Another limitation of our study is the sample size. Our target of eight samples per group was based on findings that near-maximal levels of statistical stability are obtained with eight to 15 biological replicates.25 We acknowledge that, in general, the more replicates, the stronger the inferences that can be made from the data. However, even at our minimum target sample size, we were able to demonstrate significant differential gene expression in 7.44% of the total probe sets at a very stringent false discovery rate (Benjamini-Hochberg P<.01).
Determining the reproducibility of our findings in a larger independent data set would be an important step to validate our approach to studying fetal gene expression in amniotic fluid.
Gene expression studies of the human fetus are frequently limited by practical and ethical considerations. Most genomic studies of the human fetus have relied on tissue specimens obtained after spontaneous abortions, terminations of pregnancy, or cultured extraembryonic cells. In contrast, the results presented here demonstrate that amniotic fluid cell-free RNA is a source of biologically meaningful gene expression data obtainable from the live human fetus. This has clinical relevance for future human studies in which amniotic fluid may provide a feasible alternative to direct tissue biopsy, eg, in investigations of fetal renal function at term. The findings of this study thus advance the concept of amniotic fluid as a real-time gene expression summary fluid and support its potential for future studies of abnormal fetal development.
1. Hui L, Slonim DK, Wick HC, Johnson KL, Bianchi DW. The amniotic fluid transcriptome: a source of novel information about human fetal development. Obstet Gynecol 2012;119:111–8.
2. Slonim DK, Koide K, Johnson KL, Tantravahi U, Cowan JM, Jarrah Z, et al.. Functional genomic analysis of amniotic fluid cell-free mRNA suggests that oxidative stress is significant in Down syndrome fetuses. Proc Natl Acad Sci U S A 2009;106:9425–9.
3. Koide K, Slonim DK, Johnson KL, Tantravahi U, Cowan JM, Bianchi DW. Transcriptomic analysis of cell-free fetal RNA suggests a specific molecular phenotype in trisomy 18. Hum Genet 2011;129:295–305.
4. Hui L, Slonim DK, Wick HC, Johnson KL, Koide K, Bianchi DW. Novel neurodevelopmental information revealed in amniotic fluid supernatant transcripts from fetuses with trisomies 18 and 21. Hum Genet 2012;131:1751–9.
5. Hui L, Bianchi DW. Cell-free fetal nucleic acids in amniotic fluid. Hum Reprod Update 2011;17:362–71.
6. Tyden O, Bergstrom S, Nilsson BA. Origin of amniotic fluid cells in mid-trimester pregnancies. Br J Obstet Gynaecol 1981;88:278–86.
7. Jezierski A, Gruslin A, Tremblay R, Ly D, Smith C, Turksen K, et al.. Probing stemness and neural commitment in human amniotic fluid cells. Stem Cell Rev 2010;6:199–214.
8. Larrabee PB, Johnson KL, Lai C, Ordovas J, Cowan JM, Tantravahi U, et al.. Global gene expression analysis of the living human fetus using cell-free messenger RNA in amniotic fluid. JAMA 2005;293:836–42.
9. Underwood MA, Gilbert WM, Sherman MP. Amniotic fluid: not just fetal urine anymore. J Perinatol 2005;25:341–8.
10. Dietz JA, Johnson KL, Massingham LJ, Schaper J, Horlitz M, Cowan J, et al.. Comparison of extraction techniques for amniotic fluid supernatant demonstrates improved yield of cell-free fetal RNA. Prenat Diagn 2011;31:598–9.
11. Gentleman R. Reproducible research: a bioinformatics case study. Stat Appl Genet Mol Biol 2005;4:Article 2.
12. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B Stat Methodol 1995;57:289–300.
13. Su AI, Wiltshire T, Batalov S, Lapp H, Ching KA, Block D, et al.. A gene atlas of the mouse and human protein-encoding transcriptomes. Proc Natl Acad Sci U S A 2004;101:6062–7.
14. Pennings JL, Kuc S, Rodenburg W, Koster MP, Schielen PC, de Vries A. Integrative data mining to identify novel candidate serum biomarkers for pre-eclampsia screening. Prenat Diagn 2011;31:1153–9.
15. Dennis G Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, et al.. DAVID: database for annotation, visualization, and integrated discovery. Genome Biol 2003;4:P3.
16. Hosack DA, Dennis G Jr, Sherman BT, Lane HC, Lempicki RA. Identifying biological themes within lists of genes with EASE. Genome Biol 2003;4:R70.
17. Romero R, Espinoza J, Gotsch F, Kusanovic JP, Friel LA, Erez O, et al.. The use of high-dimensional biology (genomics, transcriptomics, proteomics, and metabolomics) to understand the preterm parturition syndrome. BJOG 2006;113(suppl 3):118–35.
18. Buhimschi IA, Buhimschi CS. Proteomics of the amniotic fluid in assessment of the placenta. Relevance for preterm birth. Placenta 2008;29(suppl A):S95–101.
19. Michaels JE, Dasari S, Pereira L, Reddy AP, Lapidus JA, Lu X, et al.. Comprehensive proteomic analysis of the human amniotic fluid proteome: gestational age-dependent changes. J Proteome Res 2007;6:1277–85.
20. Cohn BR, Fukuchi EY, Joe BN, Swanson MG, Kurhanewicz J, Yu J, et al.. Calculation of gestational age in late second and third trimesters by ex vivo magnetic resonance spectroscopy of amniotic fluid. Am J Obstet Gynecol 2010;203:76.e1–76.e10.
21. Rogers S, Witz G, Anwar M, Hiatt M, Hegyi T. Antioxidant capacity and oxygen radical diseases in the preterm newborn. Arch Pediatr Adolesc Med 2000;154:544–8.
22. Romero R, Espinoza J, Goncalves LF, Kusanovic JP, Friel LA, Nien JK. Inflammation in preterm and term labour and delivery. Semin Fetal Neonatal Med 2006;11:317–26.
23. Condon JC, Jeyasuria P, Faust JM, Mendelson CR. Surfactant protein secreted by the maturing mouse fetal lung acts as a hormone that signals the initiation of parturition. Proc Natl Acad Sci U S A 2004;101:4978–83.
24. Golightly E, Jabbour HN, Norman JE. Endocrine immune interactions in human parturition. Mol Cell Endocrinol 2011;335:52–9.
25. Pavlidis P, Li Q, Noble WS. The effect of replication on gene expression microarray experiments. Bioinformatics 2003;19:1620–7.