Background

In humans and other placental mammals, the fertilized egg gives rise to both the fetus and the placenta. Placentation in the first trimester can impact fetal growth, and abnormal placentation can lead to more pronounced effects complicating pregnancy including intrauterine growth restriction (IUGR) which results in very low birth weight infants, a sexually dimorphic outcome [1,2,3,4]. The process of placentation occurs throughout the first trimester of pregnancy, whereby the outer cells of the blastocyst (the trophoblast cells) invade the maternal tissue and develop into the placenta. It is a highly regulated state of active cell proliferation, cell migration, and cell differentiation [5]. To sustain this growth, placentation is in a high state of transcriptional activity, as evidenced by its marked hypomethylated state [5,6,7,8,9]. Placentation requires multiple factors, including maternal immune tolerance, various growth factors, fetal-maternal communication via chemical signaling, and a receptive maternal decidua that allows extravillous trophoblast cells to invade the maternal circulatory system and access maternal nutrients throughout pregnancy [5, 10,11,12,13].

Fetal growth, development, and well-being is impacted by fetal sex [14, 15]. Females are slightly smaller as early as 8–12 weeks and continue to be smaller throughout gestation [16, 17]. Fetal survival rates are also sexually dimorphic in early pregnancy. Male fetuses have a 2.00–2.50-fold higher risk for spontaneous abortions in the first trimester and 1.25-fold higher risk before 23 weeks (mid-second trimester) [18]. Despite studies describing the placental transcriptome at term and differences attributed to fetal sex [19,20,21], there are no studies looking at the first trimester placenta, which is a critical time during development that may contribute to sexually dimorphic outcomes and is a time when pregnancy complications likely manifest [7, 22]. In addition, it is during this time that origins of common adult diseases such as hypertension, coronary heart disease, type 2 diabetes, and depression may develop, which are also sexually dimorphic [23, 24].

Thus, our goal is to identify the late first trimester placenta transcriptome and identify gene expression differences due to fetal sex. This work provides a valuable reference of the late first trimester placental transcriptome at the earliest time point of placental development that can be studied in ongoing human pregnancies. It can be used to develop a sex-specific placenta atlas unique to humans, important because the human placenta has unique features not found in animals models [25]. In addition, findings from these studies may be the initial step in identifying potential mechanisms in placentation that lead to sexually dimorphic outcomes in fetal growth and development as well as long-term adult health, and can ultimately be used to develop non-invasive diagnostic testing of fetal well-being [26].

Methods

Chorionic villi sample procurement

First trimester placenta was obtained from discarded chorionic villi at 10.5–13.5 weeks from chorionic villus sampling (CVS) used for genetic testing, with informed written consent. All protocols were performed in accordance with the institutional review board’s guidelines at the Cedars-Sinai Medical Center under IRB Protocols Pro00006806 (Prenatal repository) and Pro00008600 (differential gene expression of early placenta). Remaining CVS tissue that was not used for genetic testing was cleaned to separate maternal-derived decidua tissue and fetal-derived chorionic villi. Tissue samples (5–15 mg) were kept on ice and submerged in 250 μl RNAlater RNA Stabilization Reagent (QIAGEN, Hilden, Germany) within 30 min, then stored at − 80 °C in Cedar-Sinai Medical Center’s Prenatal Repository until further processing. All subjects were Caucasian, spontaneous singleton conceptions that resulted in a live birth. There were 17 females and 22 males. All karyotypes were either 46,XX or 46,XY.

Subject demographics analysis

Subject demographics including maternal age, race, ethnicity, pre-pregnancy BMI, as well as fetal gestational age at CVS, crown-rump length (CRL), karyotype analysis, gestational age at delivery, and birthweight were analyzed for n = 39 subjects. Dating by last menstrual period, when available, and early initial ultrasonography was completed using crown-rump length (CRL) measurements to confirm gestational age. Gestational age and CRL measured at the time of CVS were used for analysis. t test was used for continuous variables and Fisher’s exact test was used for categorical variables. Eight samples lacked complete clinical data for birthweight, gestational age, or crown-rump length. Omitting those samples (leaving n = 31), the results were unchanged.

RNA extraction from chorionic villi tissue

CVS were processed as previously described [27]. Briefly, CVS tissue was thawed on ice with 600 μl of RTL Plus lysis buffer (QIAGEN) and 1% β-mercaptoethanol added to each sample. Cells were homogenized by passing at least 10 times through single-use needles with decreasing gauge (22G, 25G then 27G) attached to a 1 mL sterile, RNase-free syringe. The homogenates were loaded onto AllPrep spin columns and the remainder of the protocol was performed following manufacturer instructions. DNA and RNA were extracted using the commercial AllPrep DNA/RNA Mini Kit (QIAGEN) for RNA-seq. RNA was eluted with 30 μl of the kit-provided RNase-free water at room temperature, then the elution was passed through the column a second time to increase RNA yield. DNA was eluted with 100 μl of the kit-provided EB buffer warmed to 70 °C. Equal numbers of male and female samples were processed each time to reduce batch effects. The average RNA integrity number (RIN) score for RNA samples was 7.70.

RNA-sequencing and statistical analysis

RNA-Seq libraries were constructed from 200 ng of total RNA using Illumina TruSeq Stranded Total RNA with Ribo-Zero Gold sample prep kits (Illumina, Carlsbad, CA, USA). Constructed libraries contained RNAs > 200 nt (both unpolyadenylated and polyadenylated) and were depleted of cytoplasmic and mitochondrial rRNAs. An average of 22.53 million 2 × 75 bp paired-end reads were generated for each sample on an Illumina NextSeq 500. Reads were aligned to the human reference genome (build GRCh38 with Ensembl release version 87 and UCSC release 24) using STAR [28] and we counted reads mapping to Ensembl genes using the featureCounts software in the Subread package [29]. We then used the DESeq2 Bioconductor package [30, 31] in the R statistical computing environment [32] to normalize count data, estimate dispersion, and fit a negative binomial model for each gene. The Benjamini-Hochberg false discovery rate (FDR) procedure was used to estimate the adjusted P values for Ensembl gene IDs. Significant genes were defined as below 5% FDR.

Quality control to verify sample identity

To ensure samples did not contain maternal contamination, RNA sequencing results were screened for decidua-specific genes (SCARA5, PIGF, SLPI) and samples excluded if any decidua-specific genes were expressed. Fetal sex, identified on cytogenetic analysis, was also verified after RNA-sequencing using the individual baseMean and fragments per kilobase per million mapped fragments (FPKM) levels of XIST (a female-specific gene) and highly expressed Y chromosome genes (DDX3Y, KDM5D, UTY, and ZFY).

Expression threshold selection

In order to identify the transcription profile of the late first trimester human placenta of ongoing pregnancies, an expression threshold of FPKM > 1.281 was selected. After DESeq2 analysis, the mean FPKM from female samples was found for all Y-linked genes. The maximum value was 1.2808 for CD24P4, an outlier 13 standard deviations higher than the next value, 0.3968 for PSMA6P1. Signal for CD24P4 was likely sequencing noise from CD24, a highly expressed chromosome 6 gene with 99.43% identity to CD24P4 (BLASTN E value = 2.0 × 10−90) found using the BLAST tool from the National Center for Biotechnology Information (NCBI) [33]. The higher threshold was selected to ensure a reliable transcriptome atlas. Genes on all chromosomes were considered expressed if they reached mean FPKM > 1.281 in females, males, or both.

Pathway and gene enrichment analysis

To investigate the biological significance of our gene subset, data were analyzed with Ingenuity Pathways Analysis (IPA) software (QIAGEN, Redwood City, CA, USA, www.qiagenbioinformatics.com/IPA), as previously described [34, 35]. Genes were partitioned into quartiles after sorting by maximum average FPKM (male or female). Gene groups were analyzed for enrichment in gene ontology, canonical pathways, and upstream regulators. A right-tailed Fisher’s exact test was used to calculate P values of enriched categories. The chromosome location of upstream regulators was found by cross-referencing gene names from IPA with the db2db tool from biological DataBase network (Frederick National Laboratory for Cancer Research, https://biodbnet-abcc.ncifcrf.gov/db/db2db.php) [36]. The top 25% of expressed genes were input into IPA for upstream analysis and 98.0% (3493 of 3563) were accepted. Upstream regulators were ranked by most significant P value (Fisher’s exact test). Full names and descriptions were pulled from the National Institute of Health NCBI databases (Gene, PubChem) and miRbase: the microRNA database [33, 37]. For each upstream regulator, the percent of accepted input genes that it affects is shown as a percentage (e.g., 195/3493 × 100% for D-glucose) in column “% Genes Regulated.”

Identification of placental specific or enriched transcripts

In order to identify highly expressed genes with placenta tissue-specificity, we performed an in silico analysis of tissue specificity on genes with FPKM > 128. Protein-coding genes were ranked by highest FPKM value across all samples in the first trimester. We referenced the expressed sequence tag (EST) profile breakdown by healthy body sites (typically 45 sites) of protein-coding genes using the NCBI UniGene database (https://www.ncbi.nlm.nih.gov/unigene) accessed 15 Aug 2017 [33]. To determine placenta enrichment, the transcripts per million (TPM) of protein-coding genes in different normal body sites were compared. Genes with no placenta expression in UniGene (TPM = 0) were validated to be FPKM < 2 in term placenta using RNA-seq data from NCBI Gene Expression Omnibus (GEO) database, accession GSE73016 [38].

Venn diagram of gene ontology for sexually dimorphic genes

The Venn diagram for differentially expressed genes (DEGs) was created with modifications from Bellott et al. [39]. Additional functional categories were added in an effort to maximize the number of sexually dimorphic genes represented. The NCBI Gene database (https://www.ncbi.nlm.nih.gov/gene) and UniProt [40] were used for an initial ontology scan then PubMed was cross-referenced to verify gene ontology (https://www.ncbi.nlm.nih.gov/pubmed/). Almost all DEGs (43/58; 74.2%) had known or predicted function.

Overview of X chromosome inactivation status and X/Y gene homology

The human genome annotation file (GRCh38) was downloaded from Ensembl [41]. The annotations for the 25 differentially expressed X-linked genes (FDR < 0.05) were extracted from the human genome annotation file and input to NCBI Genome Decoration Page (https://www.ncbi.nlm.nih.gov/genome/tools/gdp) to generate the ideogram of the X chromosome showing the locations of these genes. The consensus calls for X inactivation status were adapted from Balaton et al. [42], which compiled a comprehensive list of X inactivation statuses from several studies. Genes that were reported to be subject to X inactivation in most or all of these studies were categorized as X inactivated genes (color-coded red). Genes that were reported to be escaping X inactivation in most or all of these studies were categorized here as escaping genes (color-coded blue), including one gene (CHM) that had an even split of X chromosome inactivation (XCI) status calls between studies. On the human X chromosome, 25% of X-linked genes located outside of pseudo-autosomal regions have Y counterparts [42], which are either functional (“Y homologs”) or non-functional (“Y pseudogenes”) [39, 43]. Table 7 listed all X/Y gene pairs that contain functional Y homologs, but not non-functional Y pseudogenes. All X-linked genes listed in Table 7 are protein-coding genes.

Results

Identification of a transcriptome signature from late first trimester human placenta

To identify the transcriptional profile of the late first trimester human placenta, RNA sequencing was performed on chorionic villi from 39 singleton pregnancies (17 female and 22 male samples). RNA-sequencing resulted in approximately 22.53 million 2 × 75 bp pair-ended reads per sample. Genes were annotated using human reference build GRCh38. Out of 57,543 human genes, a total of 47,778 genes had non-zero baseMean values indicating possible detection in chorionic villi. Principal components analysis showed clusters separated by sex (Additional file 1A). To determine whether each gene was expressed or not in chorionic villi regardless of sex, the FPKM values were used to identify a normative first trimester placental signature. An FPKM > 1.281 threshold was selected to remove sequencing noise, higher than the typical arbitrary threshold of FPKM > 1. This higher threshold was selected to remove all background sequencing noise which produced low levels of Y chromosome signal from cytogenetically confirmed female samples (no Y chromosome). A total of 14,250 genes were determined to comprise the normative first-trimester placenta transcriptome: 13,319 present in both males and females, and the remaining 931 genes above the expression threshold in one sex but not the other. The log2FPKM of expressed genes, as an average of all samples, ranged from − 3.7356 to 16.3245 (Fig. 1a).

Fig. 1
figure 1

Normative late first trimester placenta transcriptome. a Box and whisker plot showing log2FPKM distribution of expressed genes. All: average log2FPKM of 14,250 expressed genes over all 39 samples. F, M: all genes over the FPKM cutoff in either female or male samples, respectively. The whiskers range is the interquartile range ± 50%, with outliers shown as blue circles. Median is red line. b Chromosome distribution of 14,250 genes expressed in chorionic villi, all biotypes (gray bars). Expressed genes are shows as a percentage of total genes in each chromosome (green diamonds). c Biotype categories of all expressed genes and each FPKM quartile are shown. The proportion of each biotype is shown as a percentage and labeled if ≥ 3%

Full size image

The chromosomal distribution of the expressed genes (protein-coding, long non-coding, pseudogenes, and short non-coding) are presented in Fig. 1b. The highest number of expressed genes is on chromosome 1 followed by one of the shortest chromosomes, chromosome 19 (Fig. 1b). Since gene distribution is not proportional to chromosome size, we looked at the proportion of expressed genes in the placenta as a percentage of total genes for each chromosome using human genome assembly GRCh38.p7 [41]. Even after controlling for gene distribution, 34.1% (973/2852) of genes on chromosome 19 were expressed, the highest of any other chromosome. The Y chromosome had the lowest proportion of expressed genes with only 2.8% (16 genes) of all 568 Y chromosome genes annotated in human genome assembly GRCh38.p7. Of these 16 Y-linked genes, there were 8 protein-coding, 4 long non-coding, and 4 pseudogenes.

Expressed genes in the late first trimester human placenta

The biotypes of all expressed transcripts of the normative late first trimester placenta transcriptome consist of 77.0% protein-coding genes, 9.8% long non-coding genes, 6.5% pseudogenes, and 6.7% short non-coding genes (Fig. 1c). Expressed genes were sorted by highest-to-lowest FPKM values using the female average or male average for each gene (whichever was higher) and separated into quartiles of approximately 3562 genes each. This method best represented the high expression of sex-specific genes such as Y-linked genes absent in females, which would otherwise drop from the highest quartile to the lowest quartile if the average FPKM of all 39 samples were used. The biotype distributions of the top two quartiles were similar in the first trimester placenta. Within the highest quartile (top 25% expressed genes), 94.4% (3364) were protein-coding genes, 1.3% (48) were long non-coding, 1.1% (40) were pseudogenes, and 3.1% (111) were short non-coding (Fig. 1c, Additional file 2). Of the short non-coding genes in the highest quartile, 32.4% (36 of 111) were microRNA-encoding genes, of which 66.7% (24 of 36) were located on chromosome 19. Within the second highest quartile (51-75% of sorted genes), 90.8% (3234) were protein-coding, 3.1% (111) were long non-coding, 2.7% (97) were pseudogenes, and 3.4% (120) were short non-coding genes (Fig. 1c). The proportion of non-coding genes only reached > 10% in the lower two quartiles (Fig. 1c).

Ingenuity Pathways Analysis (IPA) for gene enrichment analysis was performed for expressed genes in the first trimester placenta, partitioned into quartiles (Table 1). In the most highly expressed quartile, the enriched canonical pathways were EIF2 signaling, regulation of eIF4 and p70S6K signaling, protein ubiquitination, mitochondrial dysfunction, and sirtuin signaling. These pathways are essential for protein synthesis, cell growth, and energy metabolism. Within the second highest quartile, the most enriched canonical pathways were related to hormonal signaling and growth regulation, including estrogen, JAK/STAT, ceramide, insulin, and STAT3 signaling.

Table 1 Enriched canonical pathways for expressed genes in the late first trimester placenta, by FPKM quartiles
Full size table

Among the top 25% most expressed genes, the most enriched molecular and cellular functions were cell death and survival, protein synthesis, cellular development, cell growth and proliferation, and cellular movement (Table 2). The top five enriched categories for physiological systems development and function were organismal survival, cardiovascular systems development and function, connective tissue development and function, tissue development, and organismal development (Table 2). Upstream analysis in IPA identified the top 40 gene regulators upstream of the top 25% expressed genes (Additional file 3). Most upstream regulators were associated with either essential cell regulation, cell growth, or hormonal signaling. Transcriptional regulators such as MYC (P = 1.70 × 10−70), p53 (P = 1.63 × 10−65), MYCN (P = 3.17 × 10−63), and HNF4A (P = 2.10 × 10−58) were among the most significant upstream regulators (Additional file 3). Regulators involved in cell growth include rapamycin-insensitive companion of mTOR (RICTOR), transforming growth factor beta 1 (TGFB1), D-glucose, epidermal growth factor receptor (EGFR), as well as epidermal growth factor (EGF). In addition, hormones such as beta-estradiol, its receptor (ESR1), as well as the progesterone receptor (PGR) were also significant upstream regulators. Together, these upstream regulators control essential cell regulation, cell growth, and hormonal signaling, consistent with the canonical pathways identified (Additional file 3).

Table 2 Enriched gene ontology for top 25% of expressed genes
Full size table

In order to identify highly expressed genes with placenta-enriched tissue expression, we selected protein-coding genes with FPKM > 128, two orders of magnitude above the expression threshold followed by an in silico assessment that compared them to EST profiles of healthy human tissue in the NCBI UniGene database (http://www.ncbi.nlm.nih.gov/unigene/) (Table 3). There were 175 protein-coding genes with FPKM > 128 in the first trimester placenta, with the highest expression from KISS1 (metastasis-suppressor), CGA (glycoprotein hormones alpha subunit), TFPI2 (tissue factor pathway inhibitor 2), CSH1 and CSH2 (chorionic somatomammotropin hormones), CGB8 and CGB5 (chorionic gonadotropin beta subunits), SPP1 (secreted phosphoprotein 1), TMSB10 (thymosin beta 10), and EEF1A1 (eukaryotic translation elongation factor 1 alpha 1). Of these 175 protein-coding genes, 35 (20%) were known to be placenta-enriched, with UniGene-documented transcripts per million (TPM) highest in placenta compared to other tissues (Table 3). Two genes, XAGE3 and HIST1H2BI, were exclusively expressed in the placenta, with HIST1H2BI showing TPM = 0 in all healthy tissue on UniGene (Table 3). Sixteen genes (CGB3, CGB5, CGB8, HIST1H1B, HIST1H1E, HIST1H2BO, HIST1H3B, HIST1H3C, HIST1H3F, HIST1H4B, HIST1H4D, HIST1H4F, PSG3, PSG6, PSG9, and XAGE2) were almost specific to the placenta, with expression in 6 other tissues or less, most of embryonic or vascular system origin. Thirteen genes (EBI3, HIST1H1B, HIST1H1D, HIST1H1E, HIST1H2BI, HIST1H2BO, HIST1H3B, HIST1H3C, HIST1H3F, HIST1H4B, HIST1H4C, HIST1H4D, and HIST1H4F) had no UniGene-documented placenta tissue expression, although some were expressed in many other tissues (Table 3).

Table 3 Highest expressed protein-coding genes in late first trimester placenta
Full size table

To validate this potentially novel first trimester placenta expression, we cross-referenced RNA-sequencing data of term placentas (NCBI GEO Accession GSE73016) and found that EBI3 and HIST1H2BO were expressed in term placenta [38]. We also compared the top 175 protein-coding genes to a microarray study that compared early first trimester placenta (45–59 days) versus C-section delivered term placentas (NCBI GEO Accession GSE9984) and found 23 genes that are differentially expressed between early pregnancy and delivery [44]. Of the top 175 genes, 5 were significantly upregulated in first trimester (CD24, COL6A2, ENO1, HMGA1, KRT7) and 18 genes were significantly upregulated in term placenta (ADAM12, CYP19A1, EBI3, FBLN1, GDF15, HSPB8, KISS1, PAPPA, PAPPA2, PSG1, PSG3, PSG4, PSG6, PSG9, S100A9, S100P, SDC1, SLC2A1) [44]. The XAGE2, XAGE3, CGB family, and histone-encoding genes were not identified as gestationally different. Overall, there remained 11 histone-encoding genes highly expressed in late first trimester placenta that showed no UniGene-documented placenta expression (Table 3) nor expression in term placenta [38, 45], suggesting they are unique to first trimester placenta and potentially critical for early placentation.

Sex differences exist in placenta gene expression

Among the 39 first trimester placenta evaluated, there were 17 females and 22 males. There was no statistically significant difference in maternal age, maternal race, maternal ethnicity, pre-pregnancy body mass index (BMI), gestational age at CVS, or gestational age at delivery (Table 4). Mean CRL was 58.8 ± 12.8 mm and 51.8 ± 10.2 mm (P = 0.0714) for females and males, respectively. Thus, there was no difference in fetal size in the late first trimester among the sexes, based on CRL, adjusted or unadjusted. There were also no statistically significant differences in second trimester parameters. However, there was a significant difference in the birth weight among the sexes, with males being heavier at 3715 ± 407 g versus 3176 ± 405 g for females (P = 0.0006) (Table 4). Even in a multivariate linear regression after adjustment for maternal age, gestational age at delivery and pre-pregnancy BMI, male infants were 401 g heavier (P = 0.003, 95% CI 146–657) than female infants at birth (Table 4). This sexual dimorphism in birth weight is consistent with existing literature [15].

Table 4 Subject demographics analysis
Full size table

Principal components analysis demonstrated that male and female samples separate into two clusters along a diagonal between the first and second principal components (Additional file 1A). Using DESeq2 on the unfiltered RNA-seq data, volcano and MA plots display the separation of RNA-seq results, with the largest fold-changes in expression coming from Y-linked genes as expected then X-linked genes (Additional file 1B, C). We identified 112 genes significantly different between males and females after adjusting for multiple comparisons and filtering by Benjamini-Hochberg false discovery rate, FDR < 0.05 (Additional file 1 B and Additional file 4). Cell markers for different trophoblasts cell types were not significantly different, suggesting that male and female CVS samples contained similar cell types (Additional file 5). Of 112 genes, 58 genes also meet our FPKM > 1.281 selection criteria, leaving 35 expressed genes significantly upregulated in females and 23 expressed genes significantly upregulated in males (Fig. 2a, Tables 5 and 6).

Fig. 2
figure 2

Significantly differentially expressed genes in placenta. a Chromosome and biotype distribution of the 58 DEGs, 35 upregulated in females (F) and 23 upregulated in males (M). “Auto” = autosomal chromosomes. b Ideogram of chromosome X showing the 25 significantly differentially expressed X-linked genes between first trimester male and female placentas (FDR < 0.05). Left: 3 genes upregulated in males. Right: 22 genes upregulated in females. The consensus calls for X inactivation status were adapted from Balaton et al. [42]

Full size image
Table 5 Genes significantly upregulated in females (downregulated in males)
Full size table
Table 6 Genes significantly upregulated in males (downregulated in females)
Full size table

As expected, most of the 58 differentially expressed genes (DEGs) are from the sex chromosomes (40/58; 69.0%). Twenty-five genes (43.1% of all DEGs) come from the X chromosome (18 protein-coding, 6 long non-coding, and 1 short non-coding), with 22 upregulated in females (Fig. 2b). Of all the X-linked genes upregulated in females, the majority, 13/22 (59.1%), were previously described to escape X chromosome inactivation (XCI). Of the remaining 9 genes, 4 underwent XCI and 5 had unknown XCI status [42, 46, 47]. Several (18/40; 31.0%) of the sex chromosome genes differentially expressed among the sexes in the late first trimester were identified to be sex different in term placenta tissue (Additional file 6), suggesting that they are important throughout gestation [20]. Differentially expressed autosome genes in the late first trimester identified here were unique and did not overlap with term placenta (Table 5 and 6, Additional file 6) nor with previous microarray studies comparing early first trimester terminations versus term placenta [20, 44].

We hypothesized that homologous X/Y gene pairs would be highly represented among differentially expressed sex-linked genes. First, we examined X-linked genes. Currently, there are 19 X-linked genes with known Y homologs outside of the pseudo-autosomal regions where the X and Y chromosomes may recombine, PAR1 and PAR2 [43]. Six of the 19 genes (31.6%) are upregulated in late first trimester female placentas and escape XCI in other tissue [42]. Furthermore, the corresponding Y homologs of these six X-linked genes are upregulated in late first trimester male placentas (Table 7).

Table 7 Chromosome X-linked genes with Y homologs
Full size table

Next, we examined Y-linked genes. Only 16 Y-linked genes were expressed in the first trimester placenta and 15 were significantly differentially expressed (FDR < 0.05). Over half (8/15; 53.3%) of differentially expressed Y-linked genes are ancestral (DDX3Y, EIF1AY, KDM5D, RPS4Y1, TXLNGY, USP9Y, UTY, ZFY), all protein-coding except TXLNGY. These ancestral Y-linked genes are derived from the autosome-like proto-sex chromosomes of the last common therian ancestor of marsupial and placental mammals [39, 48, 49]. The X-linked homologs of 6 of these 8 genes were significantly upregulated in females, all except TXLNG (FDR = 0.468) and USP9X (FDR = 0.771). Males also expressed PCDH11Y, a human-specific Y-linked gene that arose from a relatively recent duplicative transposition from the X chromosome, 6 million years ago [49, 50]. PCDH11X, the X-linked homolog of PCDH11Y, was not differentially expressed between males and females in the first trimester placenta (FDR = 0.969). The remaining Y-linked genes were non-coding (Table 6).

For each of the 58 DEGs, gene ontologies were described and categorized into regulatory function groups when known (Fig. 3). Males and females both had upregulated genes involved in chromatin modification, transcription, splicing, translation, signal transduction, metabolic regulation, cell death and autophagy regulation, and ubiquitination. Some redundant functions were due to balanced expression of homologous X/Y genes. However, some of these categories also had an X-linked gene upregulated in males (ARMCX3) and differentially expressed autosome genes predominantly upregulated in females. Females also had more upregulated genes associated with signal transduction, DNA replication, cell cycle regulation, and regulation of various metabolites (e.g., hormone regulation by STS, dephosphorylation of pseudouridine 5′-phosphate by PUDP). Males had various upregulated genes from the Y chromosome without known function (Fig. 3b). Three sex-linked DEGs also had differentially expressed antisense genes upregulated in the same direction (Fig. 3), forming sense/antisense pairs of unknown, potentially regulatory function.

Fig. 3
figure 3

Functions of significantly differentially expressed genes in placenta. Venn diagrams of function categories for DEGs with known gene ontologies [39, 84, 85, 93, 97, 98, 100,101,102,103,104,105,106, 114,115,116,117,118,119,120,121,122,123,124,125,126]. Autosome genes are underlined. Sense/antisense pairs of DEGs are linked with dashed lines, with antisense genes outside the circles. *Homologous X/Y genes. **Genes with an X-linked homolog which is not differentially expressed. Functions predicted from sequence only are denoted with “(?)” next to the gene name. Genes without known or predicted function were omitted: ANOS2P, ARMCX6, FRG1JP, IQCJ-SCHIP1-AS1, LINC00278, LINC00643, PSMA6P1, RP13-36G14.4, RPL23AP11, TTTY14, TTTY15, and TXLNGY. a Genes upregulated in females. b Genes upregulated in males

Full size image

Sex differences predominate over gestational age differences in CVS samples

We next sought to identify age-specific gene expression between early and late chorionic villus sampling (CVS) collection times. We performed subanalysis with week 11 samples (10.5-11.5 weeks at CVS) and week 13 samples (12.5–13.5 weeks at CVS) and found that samples separated by sex, but not gestational age (Additional files 7, 8, and 9). Therefore, we focused our subanalysis only on sex differences within subgroups. At week 11, there were 30 DEGs between male (n = 9) and female (n = 5) CVS samples (Additional file 10). Three new autosome genes (STAT6, SLC4A1, LAMB3) were significantly upregulated in males (FDR < 0.05). The remaining 17 sex-linked DEGs were consistent with the larger 39 sample analysis: 4 female-upregulated X-linked genes (XIST, KDM6A, EIF1AX, and NUDT10) and 13 male-upregulated Y-linked genes all had FDR < 0.05 in the 39 sample analysis (Additional files 4 and 10). At week 13, there were 20 DEGs between males (n = 4) and females (n = 4) (Additional file 10). Two new autosome genes were male-upregulated (SERPINB7 and PLTP). Sex-linked DEGs were consistent with 39 sample results: female-upregulated X-linked genes (XIST and KDM5C), male-upregulated X-linked genes (ARMCX3 and ARMCX3), and 14 Y-linked genes (all except PSMA6P1).

Discussion

Normative transcriptome and highest expressed genes consistent with a time of highly regulated rapid growth

The first trimester placenta is in a state of high transcriptional activity [7, 51]. We identified 14,250 genes expressed as part of the normative late first trimester placenta transcriptome. Chromosome 19 has the highest proportion of expression genes among all human chromosomes [52, 53], and this is reflected in the placental transcriptome as well. Chromosome Y was least expressed. However, the placenta transcriptome was enriched for ancestral Y-linked genes that have survived on the chromosome Y since its evolution from an autosome precursor, with 47.1% (8 of 17 total ancestral genes present in humans) expressed in late first trimester [39]. Expressed Y-linked genes were located in previously identified euchromatin regions, where DNA is less condensed and more accessible for transcription [54], consistent with the placental state of high transcriptional activity. Most (10/16; 62.5%) of expressed Y-linked genes have current or previous homology with the X chromosome. All eight protein-coding genes and one pseudogene (TXLNGY) have X homologs [54]. Another pseudogene (ANOS2P) is a “degraded homolog” of X-linked ANOS1 gene, inactivated by a frameshift [55]. Notably, the testis-determining factor (SRY gene) is not expressed in the first trimester placenta, although it is expressed in male fetal cells as early as pre-implantation [56]. Together, this suggests that Y chromosome expression in the first trimester placenta may be supplementing the X chromosome in a pseudo-autosomal fashion, necessary for placental functions that may be distinct from fetal sex development.

The biotypes of expressed genes were skewed across FPKM distribution, with the highest expressed genes predominantly protein-coding genes. Short non-coding genes were expected to be underrepresented because the RNA extraction method is optimized for transcripts 200 nucleotides or above. Nevertheless, the two most expressed genes by FPKM were both short non-coding genes, 7SK and RN7SL2. Human 7SK small nuclear RNA is abundant in animal tissue and critical for transcription elongation [57]. RN7SL2 encodes 7SL RNA, an essential component of the signal recognition particle responsible for targeting proteins to the endoplasmic reticulum, a cell organelle important for protein secretion [58, 59]. Since placentation is a time of rapid cell proliferation and essential fetal-maternal communication, the high abundance of 7SK and RN7SL2 may be important to reduce growth restriction and facilitate protein secretion.

Gene enrichment analysis and upstream analysis of the top two quartiles of expressed genes showed a consistent pattern of expression in the placenta. Among the top 25% of genes, protein synthesis and cell growth pathways and functions were significantly enriched. Mitochondrial function and metabolic pathways were also significantly enriched, including the sirtuin signaling pathway involved in nutrient sensing and cellular metabolism [60]. Second quartile genes were enriched for hormone signaling pathways, and their binding partners (e.g., estradiol hormone) were found upstream of genes from the highest expressed quartile. Pregnancy maintenance is dependent on the interplay between hormone signaling and growth signals. The maintenance of the corpus luteum and its production of hormone during the first 8 weeks depends on trophectoderm-originated human chorionic gonadotropin (hCG) hormone [61], after which the placental production of progesterone and estrogen by the syncytiotrophoblast is sufficient to maintain pregnancy [62]. This is consistent with our findings demonstrating placental enrichment of hormone signaling canonical pathways with upstream regulation through estrogen. Other hormones such as insulin-like growth factors are also critical for regulating metabolism and fetal growth, especially in early pregnancy [63].

Finally, enriched pathways also show evidence of placental immune regulation activity. The JAK/STAT and STAT3 signaling pathway interacts with various cytokines, with mutations in pathway genes associated with both autoimmune and immunodeficient diseases [64]. Ceramide signaling in late first trimester placenta may be important for placental barrier integrity and maternal-fetal interactions, as ceramides are lipids important for skin’s barrier properties, yet excess ceramides are associated with pregnancy-induced hypertension and preeclampsia [65, 66].

Several highly expressed protein-coding genes are placenta-enriched

Since placentation is a time of rapid growth, it was expected that essential and ubiquitously expressed genes such as ribosomal proteins would be among the highest expressed genes in the placenta. However, the highest expressed genes were not all housekeeping genes. Several had known and important functions in trophoblast growth and placental development. For example, KISS1 (highest expressed protein-coding gene, third highest overall) is frequently found in overgrown tissue and tumors, yet seems to be protective against metastasis [67, 68]. In placentation, KISS1 expression limits trophoblast invasion and migration by increasing cell adhesion to collagen, ensuring normal growth [69]. The second highest protein-coding gene was glycoprotein hormone alpha polypeptide (CGA), required for production of hCG, an important hormone produced by trophoblast cells days after conception and the biomarker for pregnancy tests [70,71,72]. To identify genes with placenta-specific or placenta-enriched expression, we searched for tissue expression profiles of the highest expressed protein-coding genes in the NCBI UniGene EST database for healthy tissue. Non-coding genes were omitted due to limited tissue expression data. Our in silico search showed that the highest expressed protein-coding genes in late first trimester placenta are not trivially expressed in all tissue types. Several genes have very limited tissue expression, placenta expression much higher than in other tissue, or both. Of the top 175 protein-coding genes in the first trimester, X antigen family member 3 (XAGE3, also known as PLAC6) was the only gene known to be solely expressed in the placenta, out of 45 healthy body sites with possible transcript information for XAGE3 on UniGene. XAGE3 is part of a family of cancer/testis-associated antigens, although it is not expressed in either tumor or testis tissue, and its function is currently unknown [73]. BLAST of XAGE3 showed high conservation among primates, poor conservation with non-primates (58% identity or less), and no significant hits in non-placental mammals, reinforcing the uniqueness of XAGE3.

Eleven histone-encoding genes highly expressed in first trimester placenta were not found in UniGene’s EST profile for placenta, have not been previously described in term placenta, and are also not among highly expressed genes in term placenta transcriptome studies [38, 44, 45], showing that the first trimester placenta transcriptome is different from term placenta. Histones are DNA-binding proteins responsible for packaging DNA into nucleosomes, and differences in histones and histone modifications are associated with differences in DNA accessibility for transcription. The high expression of these histone-encoding genes in the first trimester placenta, but not later in gestation, suggests that the first trimester chromatin structure is unique and functions to maintain a high transcriptionally active state. Eight of these histone genes (HIST1H1B, HIST1H1E, HIST1H3B, HIST1H3C, HIST1H3F, HIST1H4B, HIST1H4D, and HIST1H4F) had very limited tissue expression in non-placental tissue, primarily appearing in embryonic, vascular, skin, and connective tissue. Although these genes are not significantly sex-different in placenta, all eight are significantly upregulated in response to testosterone in mouse embryonic stem cells, suggesting an additional mechanism by which the prenatal hormone environment may affect placental and fetal gene expression [74]. A ninth gene, HIST1H2BI, had no significant UniGene-documented expression in any normal tissue, showing EST count of 1/1092688 (effectively TPM = 0) in fetal brain, and no EST counts in any other normal or abnormal tissue. Its abundance in first trimester placenta (mean FPKM = 187, TPM = 564) suggests that its tissue and developmental stage expression is heavily regulated. Together, these results suggest that first trimester placenta RNA transcription is tightly controlled and possibly contains a novel histone profile that is different from both term placenta and adult tissue.

Sex differences exist in late first trimester placenta gene expression

Out of 58 DEGs in the late first trimester placenta, over a third (22/58; 37.9%) were X-linked genes upregulated in female samples. X chromosome inactivation in females is initiated during early embryonic development, when one X chromosome is randomly silenced to achieve dosage compensation between XX females and XY males. About 12% of X-linked genes always escape X inactivation and about 8% additional X-linked genes variably escape X inactivation in humans [75]. In our first trimester placenta results, over half (13/22; 59.1%) of X-linked genes upregulated in females are reported to escape XCI based on previous studies [42]. This overrepresentation of genes known to escape X inactivation in female first trimester placenta may be the result of a necessary double dose effect, similar to a double dose effect described with DAX [76]. This suggests that the increased expression from genes that escape XCI may be due to expression from both chromosomes.

XCI studies in early and term human placentas have been controversial. In earlier studies, XCI in the first trimester human placenta was reported either random in some studies or preferentially paternal in others [77,78,79,80]. In term placentas, relatively large patches of placental cells show either maternal or paternal inactivated X and the chorionic villous trees are clonally derived from only one or a few precursor cells [81, 82]. Our first trimester placenta samples consist of individual chorionic villous trees and thus might represent clonality. However, since each chorionic villus is composed of heterogenous cell populations (including trophoblast cells, mesenchymal cells, fetal endovascular cells) and different cell types have different gene expression patterns as already shown in term placenta [20, 83], the clonality of allelic specific expression in individual villi of the first trimester placenta remains to be determined.

Furthermore, the sexual dimorphic expression patterns of X-linked genes in the first trimester placentas from this study resemble what was identified by a recent transcriptome analysis in adult human tissues [84]. Sexually dimorphic expression patterns that are highly conserved from early placenta into adulthood may indicate important systemic sex differences. Among the 22 genes upregulated in females in first trimester placenta, 14 (63.6%) were also upregulated in females across multiple non-placental adult human tissues [84]. Almost all (11/14; 78.6%) were identified to escape XCI [42], with the exception of XIST which is X-inactivated, and EIF1AX-AS1 and RP13-216E22.4 whose XCI statuses are unknown. This suggests that the upregulation of these X-linked genes in females may be due to expression from both chromosomes, beginning in the first trimester and continuing into adulthood. These genes may be critical markers of sex differences in fetal growth and development that transcends into adulthood determining sex differences in adult diseases, known as fetal origins of adult diseases.

However, not all sexually dimorphic genes matched sex differences in adult tissues [84]. There were eight X-linked genes that were upregulated in females only in the late first trimester placentas: BRCC3, CHM, HDAC8, LINC00630, MIR6895, OFD1, RP13-36G14.4, and YIPF6. These genes play important roles in transcriptional regulation, DNA damage response, and vesicle-mediated transport, suggesting biological processes that may be sex-biased and critical for early placenta development. LINC00630 promotes cell proliferation by promoting the protein stability of histone deacetylate HDAC1 [85]. Non-coding gene MIR6895 is antisense to histone demethylase-encoding KDM5C, and thus may also regulate placental chromatin status. Among males, 13 Y-linked genes were consistent with that in the adult tissues, but pseudogenes ANOS2P and PSMA6P1 exhibited male-upregulation only in the first trimester placenta. Similar to other pseudogenes, they may act as miRNA decoys which may be important in early development [86, 87]. Three X-linked genes ARMCX3 (armadillo repeat containing, X-linked 3), ARMCX6 (armadillo repeat containing, X-linked 6), and ARHGEF9 (Cdc42 guanine nucleotide exchange factor 9) were also upregulated in male placenta, but not adult male tissues [84]. The three upregulated X-linked genes in males typically undergo X inactivation in other tissue, indicating that one active allele is likely sufficient for their function [42]. It remains to be determined whether the increased gene levels of these three X-linked genes in males is due to upstream Y-linked genes or if the second X chromosome downregulates these genes in females. Though there is limited research, male-upregulated expression of X-linked genes has been reported previously [84, 88]. Alex3 (encoded by ARMCX3) regulates migration and invasion in tumor cells, functions which are also critical for placentation and subsequent fetal outcomes, including fetal growth [89].

Autosome genes comprised almost a third of DEGs (18/58; 31.0%) in late first trimester placenta. In silico upstream analysis of the autosome DEGs did not yield any regulators on the sex chromosomes nor hormonal regulators that may explain the differences in expression between males and females, though this may be due to limitations of currently available data. Although many of the sex chromosome DEGs have been found in previous studies of sex differences, none of the autosome DEGs are previously identified as sex different in term placenta [20]. One autosome gene (RASSF6 on chromosome 4) was significantly upregulated in females in both late first trimester placenta and adult breast tissue [84]. Since both placenta and adult breast tissue are responsive to sex hormones, this overlap in RASSF6 makes biological sense. Overall, sex differences in autosome gene expression exist, but appear to be more subject to age- and tissue-specific variability than sex chromosome DEGs.

Distribution of sexual dimorphisms along parallel pathways (X/Y homologs)

Over a fourth (6/22; 27.3%) of X-linked genes upregulated in females have functional homologs on the Y chromosome, including EIF1AX, DDX3X, KDM5C, KDM6A, RPS4X, and ZFX. However, not all X-linked genes with Y homologs are sexually dimorphic. In fact, when examining significantly different expressed Y-linked genes in male, three Y-linked genes (PCDH11Y, TXLNGY, and USP9Y) had functional X-homologs that displayed balanced expression between the sexes. USP9Y was once thought to be essential for spermatogenesis, but men with complete deletions of USP9Y can be fertile [90, 91]. Polymorphisms in the USP9Y gene affect serum lipid profiles and coronary heart disease risk, showing that this Y-linked gene has a non-gametogenesis function [92]. The expression of three Y-linked genes in male placenta without balancing upregulation of X-linked homologs in female placentas suggests that these genes may have additional important functions for placental development that may have long-term health implications for sexually dimorphic diseases.

With pathways analysis of the 58 DEGs, we found few predicted whole pathway differences between males and females (not shown). This may be due to autosome-like expression of X/Y gene pairs such as EIF1AX/EIF1AY and RPS4X/RPS4Y1. Although the individual protein-coding genes show sexually dimorphic expression, the canonical pathways are not differentially “activated” or “inactivated,” but instead function in parallel in the same direction. We saw this dimorphic-but-parallel expression in chromatin modification, transcription, splicing, and translation pathways.

Functional significance of sexually dimorphic genes

Sexually dimorphic genes are likely functionally relevant for placenta biology. Females had a greater number of upregulated genes encoding DNA binding proteins, including genes associated with chromatin modification (including X inactivation) and genes encoding components of centrioles and cohesin complexes (OFD1 and SMC1A, respectively) important for segregation of replicated DNA during cell division [93, 94]. Metabolism-associated upregulated genes in males were primarily involved in response to nutrient deficit. MTRNR2L8 (1.61-fold upregulated in males) encodes Humanin-like 8, a small peptide homologous and sometimes identical (due to a polymorphic site) to the mitochondrially encoded Humanin peptide which promotes cell survival in ATP-deficient environments [95, 96]. Humanin promotes insulin sensitivity which may contribute to increased fetal growth seen in males [97]. HMGCS2 (1.60-fold upregulated in males) encodes an enzyme that promotes autophagy by catalyzing the first step in ketogenesis, a pathway that derives energy from lipids when carbohydrates are depleted [98]. This is consistent with previous pregnancy studies that find greater risk for nutrient deficit in males [99]. In contrast, metabolism-associated genes upregulated in females are involved in post-transcriptional modification of RNA (PUDP) and hormone biosynthesis (STS) [100, 101]. Cell adhesion, ciliogenesis, and cell-cell communication genes (OFD1, OSBL3, PCDH11Y, TBC1D32) were also differentially expressed, suggesting sex differences in how placenta cells interact with their environment [102,103,104,105]. ITGB8 (encodes integrin-β8) promotes tumor angiogenesis and invasiveness in glioblastoma [105], functions necessary for normal first trimester development when placental cells invade maternal tissue and access maternal blood.

Differentially expressed transcription factors in first trimester may explain sex differences in later pregnancy complications. For example, ZNF300 (1.58-fold upregulated in females) encodes a transcription factor protein that shares DNA binding sites with Early growth response 1 (Egr1), a zinc finger transcription factor that is elevated in placentas of pregnancies that later develop preeclampsia, which is more common in pregnancies with a male fetus [15, 106, 107]. Increased expression of ZNF300 in females may function as a competitor of EGR1 binding sites, leading to reduced rates of preeclampsia. Further studies of ZNF300 and other sex different genes are needed to understand their biological roles in placental development.

Y-linked and chromatin modification genes are consistently sex different

Gestational age-specific sex differences exist, even between 11 and 13 weeks gestation. We found that a subset of Y-linked genes are consistently expressed in late first trimester placenta regardless of gestational age and into adulthood, whereas X-linked genes and autosome genes were more variable. However, X-linked genes that affect chromatin modification (XIST, KDM6A, KDM5C) are conserved throughout gestation and into adulthood [84].

Sex differences of fetal weight at birth

In our cohort, male infants had higher birth weights compared to female infants, consistent with known sex differences at birth [108, 109]. Sex differences in embryonic and fetal growth rates have been long recognized [110], but the underlying mechanisms remain poorly understood. Recent studies show associations between placental biomarkers and fetal growth [111,112,113]. Since sex differences in gene expression may be implicated in placental function, altering growth and development which can translate into sexual dimorphism of disease in adulthood, further studies are necessary to better understand the functional roles of these genes in early pregnancy.

Limitations

All subjects in this study were Caucasian. Caucasians make up the largest demographic for chorionic villus sampling tissue that is available through our prenatal biorepository. It remains to be seen if sex differences in placenta transcripts vary among different races. Additionally, the normative placenta transcriptome described for the late first trimester here may be underrepresenting short RNAs. We found that 6.7% of the overall transcribed genes were short non-coding RNAs. The RNA extraction method used prior to RNA-seq is optimized for RNA of lengths > 200 nt, which loses 5S rRNA, tRNAs, mature miRNAs, and several other (though not all) short RNAs. Furthermore, although we compared differential expression from the late first trimester to previously published data looking at term placenta, results need to be interpreted with caution since term placenta studies were performed using microarrays with smaller sample sizes and not RNA-seq.

Conclusions

This is the largest study to date that identifies the late first trimester human placenta transcriptome, highlighting similarities and differences among the sexes. This is the earliest time point that ongoing human pregnancies that result in a live birth can be studied. We identified a group of 11 histone-encoding genes which were highly expressed in first trimester placenta, but have not been previously reported in term placenta, suggesting they may have temporally specific functions important for early pregnancy regardless of sex. When examining our cohort of patients for sex differences, we found 58 significantly differentially expressed genes with an overrepresentation of genes known to escape X inactivation. Many of the sexually dimorphic genes fell into a group of X/Y gene pairs, suggesting that dosage compensation plays a role in sex differences. Gene ontology analysis of differentially expressed genes suggested molecular signaling differences that may affect pregnancy outcomes, but further study of these genes in placenta models is needed. This is the first step into the characterization of normal placental function that can lead to the development of diagnostic tests for the normal and dysfunctional placenta, as well as identification of sex differences that start in utero and may translate into sex differences of adult diseases. This study can be used to develop a sex-specific placenta atlas which will be unique to humans at this early time point in placental development.