The placenta is readily available yet underutilized in epidemiology as a biological marker of the effects of in utero exposures on the fetal/placental unit. As an endocrine organ, which plays a critical role in all aspects of pregnancy maintenance and fetal development, the placenta can provide insight into mechanisms by which specific exposures and risk factors are associated with outcomes measured at delivery. Outcomes which are placentally-mediated include the timing of labor 1, preeclampsia 23 intrauterine growth restriction 45, and possibly endocrine 67 and neurologic diseases 8 which develop later in life. In the present study, we explored the utility of placental gene expression measured by real-time quantitative polymerase chain reaction (qPCR) as a novel biomarker for use in environmental epidemiology, including in our research on prenatal exposures to common endocrine disrupting chemicals called phthalates.

Phthalates have been shown to be reproductive and developmental toxicants in animal models with some preliminary evidence of effects in humans 910. Urinary phthalate metabolites can be characterized as low dose chronic exposures in >90% of pregnant women 111213. We chose to study the placenta as a potential target of phthalate toxicity during pregnancy. The placenta is a transient endocrine organ which assumes a wide range of functions to facilitate maternal-fetal interactions 14. It supplants the ovary in the production, metabolism and regulation of steroid and other hormones necessary for pregnancy maintenance and fetal development 15. The placenta, through autocrine and paracrine signaling, helps to maintain uterine quiescence until late pregnancy when a tightly regulated signaling cascade between the placenta, the fetus, and the uterus is initiated to stimulate uterine contractions 16. Placental transporters can both block and facilitate xenobiotic entry into the fetal compartment 17.

The placenta is an extremely heterogeneous tissue and consists of several distinct cell populations. The main placenta cell type is the trophoblast which is fetal in origin and carries out most of the functions listed above. The maternal side of the placenta or basal plate emanates from the uterine wall and consists of trophoblastic and endometrium-derived cells. The fetal component or the placental villous is composed of a mixture of cell types including trophoblasts, the trophoblastic basement membrane, stroma (mesenchymal cells, macrophages, fibroblasts, and matrix components) and fetal vessels 18. In the application of biomarkers of transcription, it is of critical importance to identify tissue type and cell type of interest and design methods to maximize the presence of these cells in any given biopsy. We were interested in targeting villous tissue and specifically gene targets expressed in the syncytiotrophoblast.

Existing, previously-validated methods for identifying toxic or pathologic responses in the human placenta include measuring the residues of the compound of interest in placental tissues 192021; documenting changes in morphology and histology as is done in routine pathological exams 222324; and looking at associations between the exposure and clinical cases of preeclampsia, placenta abruption and other clinical outcomes relevant to placental function 252627. DNA adducts have been measured in placental tissue and were associated with environmental exposures but not with enzyme activity, suggesting that they may not be good biomarkers of biochemical effects 17.

We were interested in identifying a biomarker that could be measured reliably and accurately in the placenta in an epidemiologic setting that was related to phthalate exposure and potentially relevant to effects at the molecular and clinical levels. Messenger RNA was chosen given the feasibility in measuring it in a large number of placentas and its clear physiologic relevance. qPCR was chosen over global gene expression microarrays, another commonly used method in studies using human placental tissue 2829303132, for two reasons. We had identified our gene targets based on apriori evidence from the toxicologic literature. Secondly, qPCR was more cost-effective given that our ultimate goal was to analyze a large enough number of samples (N>150) to be able to detect associations with common, low dose environmental exposures.

This validation study was undertaken to quantify underlying variability for two gene targets that will be used in our epidemiologic research: CYP19 (aromatase) and PPARϒ (peroxisome proliferator receptor protein gamma). CYP19 codes for the enzyme aromatase that is responsible for the conversion of androstenedione to estradiol in the placenta. Estradiol is essential for fetal development and parturition signaling 15. CYP19 may also be important in the metabolism of xenobiotic compounds 3334. PPARγ is a transcription factor that has been shown to play an essential role in placental development and function through the regulation of genes involved in trophoblast differentiation, angiogenesis, fatty acid transport, and inflammation 35. PPARγ null mice die during embryogenesis due to gross placental malformation and cardiac defects in the mouse 36. The expression of both CYP19 and PPARγ were altered in response to phthalate exposure in rodent models 373839404142. Within the chorionic villi, PPARγ and CYP19 expression are expressed primarily by trophoblasts. In term placentas, PPARγ is localized to a large degree to the syncytiotrophoblast 4344

The purpose to the study was four-fold: (1) to evaluate variability in gene expression by placenta, by quadrant and by inner and outer regions of the chorionic plate; (2) to determine which of three potential internal controls was most appropriate; (3) to evaluate the effects of sampling characteristics on mRNA levels such as delivery method and time elapsed from delivery to sample collection; and (4), given the extremely heterogeneous nature of placental tissue, to assess the cell type composition in our tissue biopsies using histologic methods. Due to the fact that this study was conducted on discarded tissue and exempt from human subject protection, we were not able to collect subject-specific information other than time of delivery and delivery method.

Placental Samples: Ten full-term, discarded placentas were collected in the labor and delivery room at Morgan Stanley Children's Hospital of New York Presbyterian. Placentas were collected anonymously and thus IRB (Institutional Review Board) exemption was granted by Harvard School of Public Health (HSPH) IRB and the Columbia Medical Center IRB. The placentas included Caesarian deliveries (C-section) (n = 3), vaginal normal deliveries (n = 3), vaginal induced deliveries (n = 3), and one vaginal augmented delivery (n = 1). Twelve tissue samples were collected per placenta according to the scheme in Figure 1a. A placental sampling device was used to orientate the fetal side of each placenta in relation to the umbilical cord (Figure 1b). The S1–S4 samples were categorized as 'inner' or from the area closest to the umbilical cord insertion. The Q1–Q8 samples were categorized as 'outer'. Care was taken by visual examination and dissection to minimize contamination from fetal membranes or maternal decidua and to maximize the amount of villous tissue in the sample. Each biopsy was taken approximately 1 – 1.5 cm below the fetal membrane to avoid membrane contamination as well as decidua contamination. The general dimension of each biopsy was 1–2 cubic centimeters and less than 1 gram in weight. Each sample was preserved in RNALater (Ambion, Austin, TX USA) to stabilize the RNA and stored at 4°C. Within 30 days, samples were transferred to -80°C. Samples were collected between 3 and 25 minutes after delivery (<= 10 minutes, n = 3; 11 – 20 minutes, n = 3; 21 – 25 minutes, n = 4). Samples were collected over one month in June – July 2005.

RNA Analysis: Total RNA was isolated from approximately 300 mg of tissue using the RNeasy-Midi Kit (Qiagen, Valencia, CA). Genomic DNA contamination in the sample was minimized with a DNase digestion step. Total RNA was measured by determining absorbance at 260 nm using an Ultrospec 2100 Pro UV/Visible Spectrophotometer (GE Healthcare, Piscataway, NJ USA). RNA purity was assessed by the A₂₆₀/A₂₈₀ ratio (mean = 1.8, SD = 0.13, n = 119). We also ran the isolated RNA samples on analytical agarose gels. If the gel showed clear bands for 28S and 18S, the samples were used for reverse transcription (RT). If not, we re-isolated the total RNA isolation from the same sample. Approximately 3 μg total RNA were used in a RT reaction to synthesize cDNA using the First Strand SuperScript from Invitrogen (Carlsbad, CA USA). Finally, qPCR was used to quantitate differences in mRNA levels in each sample for individual genes. Cyclophilin (CYC) and 18S ribosomal RNA (18S) were included as housekeeping genes to serve as internal controls for quantity and quality of cDNA going into the RT reaction. CYC was chosen based on a previous study that evaluated 3 housekeeping genes for use in human placental tissue analysis and found no difference between them 45. 18S was later chosen as the consensus housekeeping gene based on the more recent placenta literature 30324647. Total RNA (μg total RNA/sample) was also evaluated as a potential internal control 4849. All samples were analyzed for PPARγ, CYP19, CYC and 18S mRNA using the ABI Prism 7500 Sequence Detection System (Applied Biosystems, Foster City, CA USA). Cycling conditions were the same for all four transcripts: 95.0 C for 5:00 minutes for activation of the enzyme, 95.0 C for 30 seconds for denaturation and 60.0 C for 1:00 minute for annealing/extension for 40 cycles, followed by a dissociation step. Forward and reverse primers (Sigma, St. Louis, MO USA) were either designed by Primer 3 50 or selected for each gene to maximize specificity and efficiency in the reaction: CYP19 (249 base pairs (BP)) (forward) 5'-ATACCAGGTCCTGGCTACTG-3' and CYP19 (reverse) 5'-TCTCATGCATACCGATGCACTG-3'51; PPARγ (225 BP) (forward) 5'-GCTGTGCAGGAGATCACAGA-3' and PPARγ (reverse) 5'-GGGCTCCATAAAGTCACCAA-3'; CYC (116 BP) (forward) 5'-CCCACCGTGTTCTTCGACAT-3' and CYC (reverse) 5'-CCAGTGCTCAGAGCACGAAA-3'52; and 18S (forward) 5'-CGGCTACCACATCCAAGGAA-3' and 18S (187 BP) (reverse) 5'-GCTGGAATTACCGCGGCT-3' 53. Each reaction used 2 μl cDNA, forward and reverse primers at optimized concentrations, and SYBR Green PCR Core Reagents kit for a total reaction volume of 25 μl.

Specificity and Quantitation: Each sample was run in duplicate. The duplicate values not falling within 50% of their mean were rerun. Specificity of the PCR product was evaluated using the melting curve generated at the end of amplification and by running a 2% agarose gel to visualize the PCR product. Absolute quantitation of mRNA concentration in the original sample was achieved using a standard curve generated for each batch 54. Each standard curve included 2 non-template controls and 8 serial dilutions covering the range of 1000 molecules/μl – 1 × 10⁷ molecules/μl. The standards for each gene were prepared as described previously 55. The R² for the standard curve was between 0.98–1.00. The plate was rerun if it fell below 0.95. Ct (threshold cycle) values were use to evaluate batch effects. These are the raw data generated by the qPCR and which are used in combination with the standard curve and the internal control to calculate the number of mRNA molecules in each sample. The number of target gene mRNA molecules for each sample was divided by the number of CYC or 18S molecules or μg total RNA for that same sample and expressed as a unitless ratio.

Histologic confirmation: Histologic confirmation of cell type composition was done by using a subset of 5 representative placentas selected from the full set of 10 placentas used for qPCR. From each placenta, a piece of tissue was cut from an inner sample (closer to umbilical insertion point) and an outer sample (closer to outer margin of the chorionic plate), treated with RNA-Later and stored long-term at -80 degrees Celsius. Later they were rinsed and fixed in formalin. They were then paraffin embedded, cut and stained with hematoxylin and eosin and examined by a placental pathologist.

Statistical analysis: Gene expression values were transformed to more closely approximate a normal distribution; for CYP19/CYC and CYP19/total RNA, we were able to approximate normality with a square root transformation and for all others we used a natural log transformation. One CYP19 value for placenta ID 2 and sample Q7 was extremely low (<2 molecules mRNA/sample). We concluded that the original mRNA concentration was below detection and excluded it from the analysis. Two tissue samples were missing completely and three samples were missing 18S values, reducing sample size from N (placentas) = 10 and n (tissues samples) = 120 to N = 10 and n = 115. The distributions of the untransformed values were summarized using means per placenta. Spearman correlation coefficients were used to evaluate correlations between genes within samples. Mixed effects models were used to estimate differences in gene expression by delivery method and time elapsed since delivery, and regional (inner versus outer placenta) differences. The correlation structure assumed equal variance between any 2 of the 12 samples. Variance component analysis was used to estimate between-placenta versus within-placenta variance. Statistical significance was set at p-value equal to or less than 0.05. SAS 9.1 (SAS Institute, Cary, NC USA) software was used to conduct all analyses.

In a small study to validate methods for measuring CYP19 and PPARγ placental gene expression for application in an epidemiologic study, we identified methodologic challenges which we were able to partly resolve for these two transcripts. The application of placental biomarkers of transcription in epidemiologic studies will allow for the testing of a wide range of hypotheses relating to environmental hazards in pregnancy and their potential role in fetal origins of disease. The primary challenge was developing a reproducible sampling scheme that can give a representative snapshot of gene expression for a given placenta while not being overpowered by the high within-placenta variability in these two transcripts. In an epidemiologic analysis, the statistical ability to identify predictors of gene expression can be greatly diminished as the ratio of within placenta variance/total variance increases much over 50%.

Another challenge was the selection of an appropriate internal control that represents mRNA levels in the cell at the time of sampling independent of factors regulating the target mRNA. Thirdly, given that the placenta is a highly dynamic tissue that reaches its physiologic climax immediately prior to being sampled, characteristics such as delivery method and time elapsed from delivery can potentially introduce unwanted variability. As our goal was to measure gene expression in the fully differentiated endocrine cell type, the syncytiotrophoblast, we included an additional step to confirm that this cell type was present.

Based on the results of this study, we recommend sampling from different regions within the area of interest (i.e. fetal or maternal side) to capture physiologically relevant spatial variability. In order to target trophoblast expression, we devised an efficient design by sampling the inner and outer regions of the fetal side or chorionic plate. If more than two samples are collected, mean values for the inner and outer regions may be modelled as the dependent variable. Methods can be further tailored to reduce the impact of variability within a placenta when the objective is to compare gene expression profiles between placentas. Although, we cannot recommend a universally appropriate internal control for placental studies, 18S seems to give the least within-placenta variation for CYP19 and PPARγ.

CYP19: aromatase; PPARγ: peroxisome proliferator activated receptor gamma; qPCR: quantitative real-time polymerase chain reaction; CYC: cyclophilin; 18S: 18S ribosomal RNA; IRB: Institutional Review Board; SD: standard deviation.

The authors declare that they have no competing interests.

JA conducted the placenta sampling, all lab analyses, data analysis and drafted the manuscript. HK provided mentorship and expertise in qPCR. RH, PW, RW, HN, RH and HK contributed to study design, data analysis and manuscript preparation. HT conducted the histologic analysis.