Current approaches and developments in transcript profiling of the human placenta

In most pregnancies, the villous placenta is genetically identical to the foetus and is the first foetal-placental tissue in direct contact with the maternal exposome. The placental transcriptome may, therefore, represent to an extent both inherent foetal characteristics and the foetal response to the intrauterine environment. Data generated from genome-wide profiling of the human placenta have several uses. Firstly, analyses of normal placentas throughout gestation enhances understanding of healthy development, which serves as a reference point for studies of how the placenta responds and adapts to various exposures and challenges in complicated pregnancies. Secondly, by analysing placentas from compromised pregnancies, pathological changes linked with different clinical phenotypes can be identified and utilized for developing biomarkers or targets for prediction, diagnosis and therapeutic interventions. For instance, placental-specific gene products that are secreted into the maternal circulation can serve as a non-invasive direct readout of placental function and an indirect measure of foetal wellbeing ( Cox et al. , 2015 ). One such success story of transcript profiling is sFLT1, which was first identified to be up-regulated in pre-eclamptic placentas by microarray ( Maynard et al. , 2003 ) and is now being trialled in clinical screening to predict if a pregnant woman is at risk of developing pre-eclampsia ( Zeisler et al. , 2016 ). Additionally, knowledge of the dysfunctional molecular processes at the maternal–foetal interface provides novel insights into the potential causal mechanisms underlying placental pathologies, which may open up new avenues for developing preventative measures for pregnancy complications. Moreover, the placenta functions as the intermediary between mother and child, participating in the normal programming of the developing foetus to face the prevailing environmental conditions of ex utero life ( Burton et al. , 2016 ). Understanding these programming mechanisms and deviations in pathological conditions opens up the possibility of modifying offspring growth and health trajectories arising from compromised intrauterine environments, through interventions that target the placenta, or identifying at-risk children who will benefit from close follow-up and early childhood interventions.

The human placenta undergoes rapid growth and development usually over a span of 9 months. Serving as the maternal–foetal interface, the placenta facilitates communication between mother and child throughout gestation. Therefore, investigating the placenta provides a window into how the pregnancy has progressed and an insight into the potential health trajectory of the child. To better understand maternal–placental–foetal health, especially in the context of pregnancy complications, numerous microarrays and RNA-sequencing studies have been performed to profile the placental transcriptome. Recent rapid technological advancements in ‘omics’ have enabled parallel gathering of large datasets from maternal, placental and foetal tissues across gestation. This review summarizes genome-wide human placental transcriptome studies performed over the last two decades. We provide an overview of transcriptome study methods, outline the important considerations for study design and interpretation, discuss past study results and highlight knowledge gaps that should be addressed in future studies. As the review is focussed on human placental studies, animal studies will not be referred to. Studies in other mammals ( Barreto et al. , 2011 ; Buckberry et al. , 2017 ; Carter, 2018 ) have undoubtedly provided additional valuable perspectives on placental health and disease and have added to knowledge on comparative placentation across species.

This work was supported by funding from the Singapore Institute for Clinical Sciences, Agency for Science, Technology and Research. S.Y.C. is supported by a Clinician Scientist Award from the Singapore National Medical Research Council (NMRC/CSA-INV/0010/2016).

H.E.J.Y. has no conflict of interest to declare. S.Y.C. is part of the EPIGEN academic consortium that has received research funding from Nestec.

Good study design is critical to harness the potential of genome-wide transcript profiling of the placenta. Key aspects to be considered in study design are subject recruitment, sample processing at delivery, transcript profiling and data analysis methods and data validation ( Fig. 1 ), all of which may represent potential pitfalls and limit the validity of conclusions that can be drawn. Key aspects to consider for placental transcriptome studies. Two main points to consider in subject recruitment are selection criteria and sample size. Firstly, the selection process in case–control studies should ensure suitable controls are chosen to compare with pathological cases identified by well-defined and established clinical definitions. As will be discussed in subsequent sections, varied clinical criteria can impact study findings and reproducibility. Hence, a consensus about research definitions of common pregnancy complications should be reached to make full use of available resources. The selection process must also determine if variables, such as gestational age, sex, labour status, mode of delivery and treatment modalities, which are known to affect placental gene transcription, are part of the inclusion/exclusion criteria. Researchers should also recognize a caveat of sampling placenta from the first half of pregnancy, is that the pregnancy outcome of an electively terminated pregnancy cannot truly be guaranteed as healthy as the final outcome cannot be determined, thus interpretation of findings involving such samples must take this into account. Secondly, inadequate sample size may affect statistical power and study reproducibility. Although the average number of placentas profiled in each study is ∼28, this is largely skewed by eight large studies of more than 100 placentas each. The median number of placentas profiled per study is merely 13 (interquartile range 8–29). It is noted that while around one in five studies used fewer than 10 placentas, a considerable number of these smaller studies were performed when the use of profiling technologies were in their infancy and they were valuable in providing an early proof of concept for application in the field. Nevertheless, meta-analysis may be a means to overcome the effects of sample size in some of these earlier studies and small studies of rare conditions. While the choice of study inclusion ultimately rests on the researchers performing the meta-analysis, we strongly recommend caution with including very small studies with fewer than five samples ( Supplementary Table SI ) as study batch effects are unlikely to be sufficiently corrected for in such cases. Hence, future studies should aim for much larger sample sizes that are adequately powered to answer the study question, to improve reproducibility and verify the findings of past studies going forward. Another important factor is the placental sampling procedure at delivery. The region of the placenta to biopsy is dependent on the question asked ( Fig. 2 ). For instance, studies addressing invasion of the extravillous trophoblast into the maternal decidua sample the basal plate ( Winn et al. , 2009 ), while those investigating the maternal–foetal transfer across the syncytiotrophoblast would sample the villous placenta ( Bari et al. , 2016 ). Unwanted variation may arise from inappropriate sampling of the placenta, such as non-removal of the decidua for studies of the villous placenta, inclusion of infarcted areas and insufficient cleaning to remove excess blood. Another consideration is the importance of multisite sampling of the same region as several studies established intra-placental variation of gene expression ( Pidoux et al. , 2004 ; Hughes et al. , 2015 ). The ideal way is to run replicate samples of each placenta, although this may not always be practical given that the cost of genome-wide transcript profiling is still relatively high per sample. An alternative is to pool RNA from multiple sites of the region of interest for each placenta and to have a large enough number of different placentas, so as to ensure differential expression patterns identified are related to the condition studied, rather than normal intra- and inter-individual biological variability. Placental regions commonly sampled for transcriptome analyses. RNA integrity is also vital for proper interpretation of transcriptome data. RNA integrity is determined by measuring the 28S to 18S rRNA ratio or assessing the RNA integrity number (RIN) by a commercially available algorithm. Some placental RNA transcripts are more susceptible to degradation than others due to differences in post-transcriptional regulation ( Reiman et al. , 2017 ). Rapidly degrading transcripts are enriched among those encoding membrane components or proteins with transporter function, while stable transcripts are primarily those involved in intracellular function ( Reiman et al. , 2017 ). A major determinant of RNA integrity is the time taken to fully immerse in RNAlater or snap-freeze placental biopsies following delivery, which shows an inverse correlation with RIN values ( Fajardy et al. , 2009 ; Jobarteh et al. , 2014 ). Biopsies are often rinsed in ice-cold buffer to remove maternal blood contamination prior to RNA preservation, but this step may also alter the transcript profile, particularly at the villous sprouts, which are transcriptionally sensitive to mechanical disturbances ( Burton et al. , 2014 ). Storage length prior to RNA extraction may affect the RIN value as well, depending on preservation methods used ( Martin et al. , 2017 ). Flash-frozen samples appear more sensitive to RNA degradation over a long period of storage, compared with RNAlater-preserved samples ( Fajardy et al. , 2009 ; Martin et al. , 2017 ). Therefore, RNAlater-preservation is preferential for better placental RNA quality as compared to snap-freezing in liquid nitrogen ( Wolfe et al. , 2014 ; Pisarska et al. , 2016 ). However, if the sample is limited and there are plans to utilize the tissue for other types of analysis, a potential drawback is that preservation in RNAlater, which has a high salt content and denatures proteins, may interfere with future analysis of native proteins and other techniques. Therefore, researchers should ideally work as efficiently as possible with RNase-free equipment and consumables while handling the placenta, although it is ultimately up to the researcher how they wish to process and store their placental samples. Following sample collection, various RNA isolation and possibly enrichment methods may be required, depending on which RNA species (e.g. long non-coding RNA or mRNA) are being studied. After RNA isolation, the choice of profiling platform and data analysis is another consideration. While either microarray or RNA sequencing can be used to determine genome-wide transcriptomes, as mentioned earlier, they each have their advantages and disadvantages (thoroughly reviewed by Cox et al. , 2015 ). Data analysis considerations include whether to identify differentially expressed genes or broad categories of dysregulated pathways in case–control studies and to ensure adequate statistical adjustment to account for multiple testing, i.e. false discovery correction. Technical notes and methods to visualize these data are further discussed in a recent review by Konwar et al. (2019) . Another key aspect is to validate the findings from transcriptome analyses. Expression changes should ideally be confirmed at the RNA and protein levels by techniques, such as real-time qPCR, immunoblotting, enzyme-linked immunosorbent assays and immunohistochemistry. Furthermore, in vitro functional assays with primary explant cultures or isolated cells and in vivo animal models would provide deeper mechanistic insights into the role of identified genes.

Cellular heterogeneity of the human placenta is likely a major contributing factor to inconsistent transcriptome findings between studies. With the advent of single-cell transcriptomics, it may be possible to deconvolute the placental tissue transcriptome more readily and account for differences in tissue sampling in the near future. Deconvolution may also highlight differences in placental cell composition due to the underlying pathology. Single-cell human placental transcriptome profiles across all trimesters are now available and serve as a basis to develop algorithms for deconvolutions ( Table I ). However, sample sizes in these studies are relatively small between two and ten placentas and with the current technical limitations requiring cell dissociation, available datasets may include artefactual changes and not fully capture the transcriptome of all placental cell types, particularly the large multi-nucleated syncytiotrophoblast. Meta-analysis of past studies enhances statistical power, which may highlight novel molecular pathways or strengthen the evidence for previously identified genes. Indeed, multiple meta-analyses have been performed for pre-eclampsia ( Kleinrouweler et al. , 2013 ; Moslehi et al. , 2013 ; Vaiman et al. , 2013 ; van Uitert et al. , 2015 ; Vaiman and Miralles, 2016 ; Brew et al. , 2016 ) and preterm birth ( Eidem et al. , 2015 ; Paquette et al. , 2018 ), as well as for investigating sexual dimorphism of the placenta ( Buckberry et al. , 2014 ) and understanding trophoblast differentiation in the context of hydatidiform moles ( Desterke et al. , 2018 ). Nevertheless, lack of data access can hamper the ability to perform powerful meta-analyses. For example, of the 10 transcript profiling studies performed for GDM, only data from three studies are publically available, representing less than a third of the profiled placentas. Efforts are ongoing to make placental data more accessible. For instance, the newly developed Placenta Atlas Tool centralized database simplifies the search for relevant placental transcriptome datasets and allows some basic analysis to be performed within the site ( Ilekis et al. , 2019 ), providing a useful starting point for researchers. Consistent and clear reporting of experimental details, such as specific microarray platforms utilized and poly A+ selection for mRNA enrichment in RNA sequencing studies, and of key clinical information to inform on disease subtypes and severity are also critical in enabling researchers to design proper integrative meta-analysis studies. Traditionally, much focus was on protein-coding mRNA transcripts that result in functional changes. However, growing evidence implicate possible roles for non-coding RNA (e.g. long non-coding RNA, miRNA, circular RNA) in the placenta ( Cox et al. , 2015 ). Differential placental expression of non-coding RNAs has been investigated in pre-eclampsia ( Gunel et al. , 2017 ; Hu et al. , 2018a ; Lykoudi et al. , 2018 ; Zhou et al. , 2018 ), GDM ( Li et al. , 2015 ; Wang et al. , 2019b ), IUGR or SGA pregnancies ( Wen et al. , 2017 ; Ostling et al. , 2019 ) and early pregnancy loss ( Hosseini et al. , 2018 ). The use of RNA sequencing technologies to further characterize these non-coding RNAs provides a tantalizing approach to identify and develop new biomarkers and therapeutic targets for pregnancy complications. As such, researchers may want to consider the different RNA extraction methods available to enable capture of all placental RNA species, both coding and non-coding, for analysis in future studies. There is a paucity of genome-wide transcriptome studies in many aspects of pregnancy ( Supplementary Table SII ). The pre-eclamptic placenta is disproportionately profiled as compared with other common pathologies of pregnancy including preterm labour, IUGR, GDM and stillbirth ( Table II ). Although an estimated 2.6 million stillbirths occur annually worldwide ( Blencowe et al. , 2016 ), of which ∼40% are unexplained ( Reinebrant et al. , 2018 ), no study has yet profiled placentas from this devastating pregnancy complication, although such studies could yield useful insights as to why a foetus dies in utero , especially in cases of unexplained stillbirth. Transcript profiling of abnormal placental development, such as placenta praevia, placenta accreta and molar pregnancy could also highlight the potential causative factors behind their pathogenesis ( Desterke et al. , 2018 ), possibly enabling the discovery of new approaches to treat or prevent the recurrence of these pathologies in future pregnancies. Additionally, even with much ongoing research effort to minimize infectious diseases, pregnant women, being relatively immunosuppressed, remain at great risk of infections, such as malaria ( Dellicour et al. , 2010 ) and have shown increased susceptibility to recent global health emergencies, such as the swine flu pandemic and Ebola outbreak ( Kourtis et al. , 2014 ; Silasi et al. , 2015 ). We propose that placental transcriptome studies be used to improve understanding of how maternal infections affect the placenta, so as to identify the mechanistic pathways that can be targeted to reduce the transmission of harm to the growing foetus. Furthermore, given the increasingly recognized importance of the maternal nutritional, mental and emotional states, the rise in women exposed to harmful environmental and chemical exposures during their pregnancies and the profound impacts these can have on the pregnancy and the future health of the child ( Hoirisch-Clapauch et al. , 2015 ; Lewis et al. , 2015 ; Unger et al. , 2016 ; Chen et al. , 2018 ; Henschke, 2019 ; Varshavsky et al. , 2019 ), their effects on the placenta are all deserving of further investigation, so as to increase ways of promoting benefits of some lifestyles while minimizing adversity. Pre-existing medical conditions (e.g. autoimmune and endocrine diseases, thrombophilia) are strongly associated with an aberrant hormonal, metabolic and inflammatory milieu that is detrimental to placentation, and thus, such pregnancies are predisposed to significantly higher rates of complications with poorer neonatal outcomes ( Ali et al. , 2016 ; Vannuccini et al. , 2016 ; Meakin et al. , 2017 ; De Leo and Pearce, 2018 ; De Carolis et al. , 2019 ; Mitriuc et al. , 2019 ; Stepien and Huttner, 2019 ). Placental transcriptome analysis could thus reveal how such conditions heighten a woman’s susceptibility to obstetric complications, which may lead to new treatments to prevent defective placental function and improve pregnancy outcomes in affected women. Pregnancy-specific factors can also have profound consequences on pregnancy outcomes. Given major societal changes, more women are using ART to conceive and/or entering pregnancy at an older age, both of which are associated with placental dysfunction and more obstetric problems including IUGR and stillbirth ( Nelissen et al. , 2014 ; Lean et al. , 2017 ). Placental profiling may help reveal whether higher rates of obstetric complications observed are due to underlying subfertility or ART as suggested by animal studies ( de Waal et al. , 2015 ), and enable appropriate strategies to be developed for mitigating harms. Frequently excluded from transcriptome studies, profiling placentas from multiple pregnancies may also demonstrate the mechanisms involved in the inherently elevated risk of obstetric complications ( Witteveen et al. , 2016 ), which could be targeted to improve outcomes for the mother and her children. Moreover, further studies of twin placental transcriptomes with disconcordant intrauterine growth, whereby the healthy twin can serve as a well-matched control ( Roh et al. , 2005 ; Wen et al. , 2017 ), may help elucidate novel mechanisms of foetal growth that can be capitalized upon to improve IUGR outcomes. Integrating placental transcriptome data with other datasets including other ‘omics’ and longitudinal data will enhance knowledge into healthy placental development and disease mechanisms. A study comparing the pre-eclamptic placental transcriptome to the blood transcriptome of cardiovascular disease identified significant overlap between the two, and provided novel insights into possible shared relationships between pregnancy complications and subsequent health in the mother or child postnatally ( Sitras et al. , 2015 ). Placental transcript profiling in birth cohorts with comprehensive longitudinal follow-up of the children may also potentially uncover new placental programming mechanisms that can influence extrauterine life in the longer term. Indeed, previously identified placental eQTLs were predictive of birthweight and subsequent childhood obesity in a cohort study, highlighting the role of placental gene expression in modulating postnatal outcomes, and such genes could serve as potential molecular targets for interventions ( Peng et al. , 2018 ). Besides profiling more placentas from additional birth cohorts, it is of great interest to examine currently available birth cohort-related placental transcriptome datasets for any possible associations with childhood outcomes ( Binder et al. , 2015 ; Cox et al. , 2019 ). In doing so, we may be able to develop a catalogue of placental biomarkers predictive of future health, which could be used to identify offspring at high risk of subsequent poor health. Thus, the placenta may serve as a unique window into the future extrauterine life of the offspring and provide an opportunity to intervene and change the health trajectories of those exposed to an adverse intrauterine environment.

Many maternal sociodemographic factors, lifestyle and environmental exposures have been linked with pregnancy complications and offspring health adversity. A better understanding of how these different exposures affect the placenta will provide a greater insight into the pathophysiology and biological pathways leading to complications and abnormal programming of offspring health. As such, studies in this research theme examined placentas from pregnancies with different exposures from assisted reproduction and various clinical trial interventions to pre-existing maternal conditions ( Table II ). Alarmingly, a growing number of women are entering pregnancy in an obese state worldwide. Being obese markedly increases the risks of pregnancy complications, such as pre-eclampsia, GDM and preterm delivery ( Sureshchandra et al. , 2018 ). Transcript profiling of placentas from obese women may uncover the underlying basis of their heightened risk of pregnancy complications and the molecular mechanisms involved in programming offspring health. Five placental transcriptome studies (n = 242, 44% affected) have considered maternal obesity ( Table III: Obesity). Two studies similarly found enrichment of differentially expressed genes involved in angiogenesis, lipid metabolism and the immune/inflammatory response ( Saben et al. , 2014a ; Altmae et al. , 2017 ). A different study identified perturbed placental nutrient transport as another consequence of exposure to the maternal obesity milieu ( Sureshchandra et al. , 2018 ), which could alter foetal growth and postnatal health trajectories. Indeed, a large study of 183 placentas examining maternal pre-pregnancy BMI as a continuous variable demonstrated that gene clusters enriched for maternal immune dysregulation were positively associated with maternal BMI and negatively associated with low birth weight, providing evidence of a molecular basis to the relationship between the two ( Cox et al. , 2019 ). Therefore, the placenta exposed to maternal obesity is characterized by immune dysregulation, which can have widespread effects on placental function and foetal growth and development. Tobacco smoke is an external environmental exposure that is detrimental to pregnancy. Global profiling of smoke-exposed placentas in four studies (n = 197, 30% affected, Table II: Smoking) showed harmful and dysregulating effects of tobacco smoke on different aspects of placental growth and metabolism ( Huuskonen et al. , 2008 ; Bruchova et al. , 2010 ; Votavova et al. , 2011 ; Votavova et al. , 2012 ). Comparison of smoke-induced transcriptomic changes in placenta by active or passive smoking showed a substantial overlap between both groups compared with the non-smoking group in biological processes, such as lipid metabolism, oxidative stress and blood coagulation ( Votavova et al. , 2012 ), suggesting that these common molecular placental mechanisms are involved in transmitting the harm of smoking to the growing foetus be it active or passive. As such, these placental pathways could be potential intervention targets to modify the pregnancy outcomes of those exposed to any type of tobacco smoke during gestation. Various nutritional or drug interventions are being explored in clinical trials to improve pregnancy outcomes. While treatments may show promising results in vitro on placental cell cultures and in vivo in animal models, many have not shown the desired effect upon testing in most clinical trials. Transcript profiling of placentas from women treated as part of a clinical trial during pregnancy may thus provide valuable insights into the underlying molecular mechanisms affected, inter-individual variability in effects and possible explanation for trial findings. We identified three studies that analysed a total of 136 placentas from clinical trials ( Table II: Clinical trials). A study conducted on placentas collected from women supplemented with a low or high dose of choline from the start of the third trimester, with the intention of improving placental vascular function and reducing the risk of developing pre-eclampsia, showed widespread effects on the placental transcriptome, particularly on processes involved in vascular regulation ( Jiang et al. , 2013 ). A promising finding was significantly reduced expression of the pre-eclampsia-associated FLT1 gene, which can induce systemic vascular dysfunction at high protein concentrations, thus providing a molecular basis for the utility of choline supplementation during pregnancy ( Jiang et al. , 2013 ). Omega-3 fatty acid supplementation altered placental expression of genes involved in cell-cycle regulation in a sexually dimorphic manner with greater changes occurring in pregnancies with female foetuses, which correlated with offspring birthweight and birthweight centiles ( Sedlmeier et al. , 2014 ). This study highlights the placental response to omega-3 supplementation and the potential mechanisms involved in modulating foetal growth and postnatal development ( Sedlmeier et al. , 2014 ). A third study examined placentas collected from obese women who were treated with metformin or a placebo ( Chiswick et al. , 2016 ), with the aim of determining if there were any changes in genes regulating foetal growth or metabolism. However, while the transcriptome dataset is publically available alongside complementary methylome data, the study findings remain unpublished. Nevertheless, this dataset serves as a valuable resource to understand the effects of metformin on the placenta and acts as a possible reference for studies involving metformin in treatment of GDM. With the global rise of ART, more pregnancies are now being conceived by IVF, which is associated with poorer pregnancy outcomes ( Nelissen et al. , 2014 ). Three studies have examined the IVF placental transcriptome in the first and third trimesters (n = 169 placentas, 28% exposed, Table III:   In vitro fertilization). To determine which differentially expressed genes are related more specifically to IVF, the largest study of 141 first-trimester chorionic villus samples included a non-IVF ART group alongside spontaneous conceptions for comparison and identified CACNA1I , which codes for a calcium channel subunit, as one such gene ( Lee et al. , 2019 ). Further comparison between just the IVF and non-IVF ART groups showed differential expression of SLC18A2 , CCL21 , FXYD2 , PAEP and DNER , which supports the notion that IVF also has distinct effects on the placenta compared with other types of ART ( Lee et al. , 2019 ). However, as ART are used primarily by couples who struggle to conceive naturally, discovered alterations may be due to the underlying parental factors contributing to subfertility rather than a result of ART used. Since infertility causes are so varied, such as being due to structural defects of the reproductive tract, ovulatory dysfunction, endometriosis, childhood cancer chemotherapy or unexplained maternal or paternal factors, stratifying by causes of infertility may help discriminate the unique gene signatures for infertility as compared with those that are consequential of ART in future studies. Antenatal maternal mental health is of rising importance as cumulative evidence suggests a potent impact on pregnancy and childhood outcomes. Currently, only one study (n = 20 placentas, 50% affected, Table III: Antenatal depression) has examined the effects of maternal depression and antidepressant treatment on the placental transcriptome ( Olivier et al. , 2014 ). Most differentially expressed genes compared with controls showed limited overlap between the untreated and medically treated depression ( Olivier et al. , 2014 ). This suggests that not only does depression itself affect the placental transcriptome, but that depression and antidepressants have largely independent effects on the placenta and that those treated with antidepressants should be evaluated separately in future studies.

H.E.J.Y. contributed to the study design, performed the literature searches and wrote and edited the manuscript. S.Y.C. conceived the study and provided critical revision of the manuscript for intellectual content.

Placental transcript profiling presents enormous potential to enhance understanding of healthy placental development and function, highlight the possible underlying causal and consequential mechanisms of pregnancy complications ( Fig. 3 ), and predict and improve the health outcomes of mothers and offspring from compromised pregnancies. Challenges in obtaining sufficient numbers of quality samples with clear clinical characteristics will need to be overcome to drive the field forward. Current data may also be capitalized upon by performing meta-analyses to increase statistical power, although interpretation of findings will need to carefully account for limitations, such as inconsistent clinical criteria between studies and gestational age matching of cases and controls. Furthermore, additional resources should be dedicated to analyse placentas from the understudied areas, which will enable the dynamic complexities of the placental transcriptome to be more fully appreciated. Factors that influence the placental transcriptome and the associations of altered placental molecular pathways with pregnancy complications or exposures. Dynamic transcriptional regulation of the placental interface throughout gestation is multi-factorial (arrows). Differential regulation of specific molecular pathways identified in multiple placental transcriptome studies may highlight the potential underlying causal mechanisms involved in pregnancy complications or represent the placental mechanisms affected by pregnancy exposures (connecting lines). GDM, gestational diabetes mellitus; IHCP, intrahepatic cholestasis of pregnancy; IUGR/SGA, intrauterine growth restriction/small for gestational age; LGA, large for gestational age; RM, recurrent miscarriage.

Two methods used to obtain genome-wide placental transcript profiles are microarrays and RNA sequencing. These high-throughput technologies generate large amounts of data and offer a means to analyse the placenta in an unbiased manner. Microarrays utilize short oligonucleotide probes embedded on a chip, which when hybridized to specific RNA or DNA sequences present in the sample emits fluorescence ( Cox et al. , 2015 ), allowing simultaneous quantitation of many gene transcripts. Commercially available chips from Affymetrix, Agilent and Illumina are commonly used in placental research, although a few studies have custom-made theirs. The main drawback of microarray chips is that a gene-specific probe must be present for a gene to be detected. As the earliest placental microarray studies were performed around the time the first human genome was sequenced, not all transcripts were detectable by gene probes available at that time. Moreover, RNA biology was not as well understood as it is now with the current knowledge of non-coding RNA and RNA gene silencing. Additionally, microarray probes are species-specific. Nevertheless, given established bioinformatics pipelines and readily available statistical tools to analyse microarray data, with publically available datasets for comparison from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) and the European Bioinformatics Institute (EBI) ArrayExpress repositories, microarrays continue to be widely used ( Cox et al. , 2015 ). Next-generation RNA sequencing is more sensitive than conventional microarray and generates a fuller picture of the placental transcriptome. The Illumina Hiseq and Genome Analyzer II systems are the most popular platforms in placental research and are increasingly used as the price of sequencing falls. Sequencing can detect rare and novel RNA transcripts, identify single-nucleotide variants in both coding and non-coding RNA of all lengths, and is not species-dependent. Furthermore, recent development of microfluidics technology with RNA sequencing allows transcripts of individual cells to be determined, which was not previously possible ( Hu et al. , 2018b ). However, single-cell RNA sequencing of the syncytiotrophoblast, which forms the placental cellular barrier, remains a challenge since its large size and multi-nucleated nature does not permit its isolation with microfluidics technology. The targeted sequencing depth or number of reads sequenced per sample is dependent on experimental aims and design. For instance, the ENCODE guidelines recommend a minimum sequencing depth of 30 million reads for bulk RNA sequencing that is commonly used for differential gene expression analysis, while 10 000 to 50 000 reads per cell in single-cell sequencing are sufficient to classify cells in an unbiased manner ( Haque et al. , 2017 ). Another question-dependent experimental design consideration is whether to sequence library preparations from ribosomal RNA (rRNA)-depleted total RNA (including non-coding RNA) or those enriched for mRNA transcripts by poly A+ selection. Given the vast potential of RNA sequencing, it is unsurprising that its use is growing exponentially in placental research.

The key advantage of culturing in vitro explants or isolated cells from the placenta is that we can determine the precise effects of altering a single factor in well-controlled experimental conditions, without having to contend with other inter-individual variability in inter-placental comparisons. Transcript profiling of cultured placental explants, primary cells and cell lines have thus also contributed to increased understanding of placental function, with many studies profiling the placental response to various agents and treatments including hypoxia, irradiation, growth factors, cytokines and infectious agents ( Table III ). Such in vitro models have limitations including issues with cell purity of isolated primary cells, loss of 3D cytoarchitecture and absence of cell–cell interaction in single-cell type cultures. Although placental explant and organoid cultures may overcome some of these limitations, culture conditions may not fully reflect in vivo conditions and could lead to spurious findings as a result of altered responses to the exposure of interest. Hence, transcriptome findings of in vitro models should be interpreted with some caution and further verified in ex vivo studies. Most in vitro studies utilize primary trophoblast cultures from term placenta or extended trophoblast cell lines. Given the common use of immortalized cell lines in placental research, widely-used trophoblast-like cell lines were profiled to ascertain their genome-wide phenotype and how representative they were of primary trophoblast ( Burleigh et al. , 2007 ; Bilban et al. , 2010 ; Apps et al. , 2011 ; Takao et al. , 2011 ). These studies found little overlap in the profiles between primary cells and cell lines, and recommended caution in use of cell lines in placental research ( Burleigh et al. , 2007 ; Bilban et al. , 2010 ; Apps et al. , 2011 ). Another reason for genome-wide profiling of cultured trophoblast is to characterize the differentiation process from cytotrophoblast to syncytiotrophoblast ( Shankar et al. , 2015 ; Rouault et al. , 2016 ; Yabe et al. , 2016 ; Zheng et al. , 2016 ; Robinson et al. , 2017 ; Azar et al. , 2018 ; Gauster et al. , 2018 ). Pairing genome-wide transcript profiling with targeted gene manipulation using small RNA or plasmid technologies allows gene regulatory effects to be determined alongside consequential effects on trophoblast cell function ( Rigourd et al. , 2008 ; Tauber et al. , 2010 ; Xie et al. , 2014 ; Than et al. , 2018 ), despite the potential limitations in methodology discussed earlier. Additionally, several studies have examined non-trophoblast cell types in culture. For example, to expand knowledge on the genetic regulation of placental vascularity, Augsten et al. (2011) performed a microarray on primary term placental endothelial cells treated with high-density lipoprotein to assess how foetal lipoprotein, which contains a considerably higher proportion of apoE than that of adults, alters placental vessel function. Another microarray study identified genes involved in the immunoregulatory and pro-angiogenic function of first-trimester decidual endothelial cells relative to another endothelial cell type from skin ( Agostinis et al. , 2019 ), providing an insight into the unique role of these maternal endothelial cells in modulating immune tolerance of the foetus at the interface. Transcriptome studies of non-trophoblastic cells remain few, and if conducted and interpreted with the caution discussed previously, further investigations utilizing cultures of these cell types may potentially enhance understanding of the complex processes occurring in pregnancy. Placental-derived cultures are also used to mimic in vivo exposures. For instance, several datasets reflect the response profile of placental explants, trophoblast cell lines, primary decidual and chorion organoid cultures exposed to infectious agents – Coxiella burnetii , Trypanosoma cruzi and Zika virus ( Ben Amara et al. , 2010 ; Weisblum et al. , 2017 ; Castillo et al. , 2018 ), which are further informed by related studies that examine the effects of immunomodulatory cytokines on placental function ( Ibrahim et al. , 2016 ; Verma et al. , 2018 ; Yockey et al. , 2018 ). One study characterized the molecular signatures of the response to insulin in trophoblast cells cultured from first-trimester placentas of lean and obese women, revealing that prior exposure to obesity blunted trophoblast sensitivity to insulin ( Lassance et al. , 2015 ). Notwithstanding the technical limitations that may impact transcriptome findings, placental-derived cultures can add another dimension in expanding knowledge of placental function, which may not be fully apparent from study of tissues and immediately isolated cells.

Click here for additional data file.