reSeArCH Letter
reported to be signatures of EPI during the implantation process in
monkeys^10 (Fig. 2a and Supplementary Tables 4–6). Immunostaining
showed that OTX2, a pluripotency marker for EPI development during
implantation^11 , was expressed in a fraction of GATA6-expressing PE
cells, but not in OCT4-expressing EPI cells, verifying our single-cell
RNA-sequencing data (Fig. 2b and Extended Data Fig. 2d). Human
and monkey embryos shared partial signatures of these three lineages
and related derivatives^10 , although human embryos also expressed
several unique signature genes, such as UTF1 in EPI, GPC3 in PE and
CYP19A1 in TE (Extended Data Fig. 3).
Principal component analysis and pseudo-time analysis
revealed that all three lineages presented their own developmental con-
tinuity, suggesting that there are stepwise implantation routes (Extended
Data Fig. 4a, b). Further analysis of stage-specific gene-expression
patterns for each lineage indicated that the embryo started preparing for
mother–fetal interactions during implantation (for example, the expres-
sion of embryonic morphogenesis and pregnancy-associated genes;
Fig. 2c–e and Extended Data Fig. 4c–f). Notably, the t-SNE results
showed that TE gradually formed two separate subgroups around days
10–12. The differential expression of HCGB family genes correspondedto the subgrouping of TE by days 12–14 (Extended Data Fig. 4g).
We determined that these subgroups consisted of cytotrophoblasts
(which specifically expressed ITGA6) and syncytiotrophoblasts (which
specifically expressed CGB family genes)^3 ,^12. Cytotrophoblasts also
expressed regulators of the TE and placenta, such as FABP5 and FGFR1,
whereas syncytiotrophoblasts expressed hormone-related genes of the
placenta (for example, PSG3 and PSG6) and a number of newly iden-
tified genes (for example, TCL6 and TBX3) (Extended Data Fig. 5).
Furthermore, we found that both meiosis- and mitosis-derived copy
number variations (CNVs) were widely present in the cultured embryos
during implantation^13 ,^14 (Extended Data Fig. 6 and Supplementary
Methods). These aneuploid cells still clustered with the corresponding
euploid cells in the t-SNE map, which suggests that the differentiation
of the major lineages was generally not distorted by mild CNVs at the
early stage of implantation (Fig. 3a).
Inactivation of the X chromosome is important for the dosage
balance of X-linked genes between females (XX) and males (XY),
whereas upregulation of the expression of genes on the X chromosome
is critical for the dosage balance between X-linked genes and autosomal
genes^15 –^19. The expression of X-linked genes should be equivalent to
that of autosomal genes, which is achieved through upregulation of
genes on the X chromosome in both male and female cells^20. Further
measurements showed that the ratio of X chromosome to autosomes
in male cells was near 2:2 (but not 1:2), indicating that the upregulation
of the X chromosome had already started and the expression levels
of genes on the only copy of the X chromosome in a male cell had
already become comparable to those on two copies of the autosomes.
Comparatively, the ratio of X chromosomes to autosomes in female
cells appeared to be above 1 and slightly higher than the ratio found in
male cells during implantation^21 (Fig. 3b and Extended Data Fig. 7).
We therefore suggest that both the upregulation and inactivation of
the X chromosome had started, but was not fully completed in female
cells at day 12. That is, one of the X chromosomes had randomly been
inactivated and expression of genes on this X chromosome had been
downregulated, whereas the expression of genes on the other, active,
X chromosome had increased by nearly twofold. Subsequently, 238
representative single cells for all three major lineages were selected for
full-length cDNA high-depth sequencing to separate parental alleles
within each individual cell (Supplementary Tables 1, 7). The genomes
of the parental donors were also sequenced to call single-nucleotide
polymorphisms (Supplementary Methods). At day 6, the majority of
the cells still expressed X-linked genes in a balanced way from pater-
nal and maternal alleles. However, at later stages many cells showed
gradual accumulation of paternal- or maternal-allele-biased expression
patterns, clearly indicating the initiation of random X chromosome
inactivation during the early implantation period (Fig. 3c, d).
DNA methylation has critical roles in the epigenetic regulation
of the development of mammalian embryos. Nevertheless, the
lineage-specific DNA methylation dynamics around implantation
remain largely unknown. Using the single-cell Trio-seq2 strategy^5 and
another round of embryonic culture followed by the collection of single
cells, 2,544 individual cells from 17 embryos were first analysed for
lineage identification (Extended Data Fig. 8, Supplementary Methods
and Supplementary Tables 8–10). We subsequently selected 371
lineage-specific individual cells for post-bisulfite adaptor-tagging DNA
methylome sequencing (Extended Data Fig. 8b–d). Furthermore, 130
euploid cells were retained for subsequent analyses after removal of
aneuploid cells (Extended Data Fig. 9). Principal component analysis
of DNA methylome data showed that these 130 cells formed 4 major
clusters (Extended Data Fig. 8g, h), with a combination of the EPI, PE
and TE at the blastocyst stage (day 6) as a single cluster, and the EPI,
PE and TE beyond the blastocyst stage as another 3 separate clusters,
suggesting that all of the 3 lineages showed considerable changes in
DNA methylation soon after implantation.
Next, we explored the dynamics of DNA methylation of each lineage.
In general, EPI, PE and TE experienced strong genome re-methylation
during implantation (Fig. 4a and Extended Data Fig. 10a). The median−20020−40 −20 020
t-SNE1t-SNE2Aneuploid
EuploidabcdLineage
EPI
PE
TE
6810 120.91.01.1Male810120.8FemaleChr.X/A ratioParental origin
EPIMaternal Paternal
Lineage PE TE ysTEMaternal – paternal alleleChromosome−50−25025(^50) Day 6 Day 8Day 10 Day 12
(^6810) Chr.X Chr.1 Chr.X Chr.1 Chr.X Chr.1 Chr.X Chr.1
Day
Day Day
12
Biallelic ratio
Chr.1
Chr.2
Chr.3
0
0.5
1.0
1.5
Fig. 3 | CNVs and unsynchronized X chromosome inactivation among
different lineages during implantation. a, CNV information projected
onto the t-SNE plot based on the transcription-factor regulatory network.
In total, 3,184 single cells were included; 1,027 cells were aneuploid and
2,157 cells were euploid. b, The ratio of total expression levels of genes
located on the X chromosome (Chr.X) and the same number of genes
located on autosomes (A). In total, 3,184 cells were included; male,
n = 1,016 cells; female, n = 2,168 cells (Supplementary Table 1). Black
lines indicate median values, the boxes range from the 25th to 75th
percentiles and the whiskers correspond to 1.5× the interquartile range
(IQR; the distance between the first and third quantiles). c, The proportion
of bi-allelic expression of chromosome-X-linked genes compared to the
expression of autosome-linked genes (Supplementary Methods). In total,
150 cells were included; day 6, n = 24 cells; day 8, n = 30 cells; day 10,
n = 70 cells; day 12, n = 26 cells. Black lines indicate median values, the
boxes range from the 25th to 75th percentiles and the whiskers correspond
to 1.5× the IQR. d, The percentage differences between maternal and
paternal alleles (maternal allele (%) − paternal allele (%)) for each cell.
According to the heterozygous single-nucleotide polymorphisms, reads
were traced to their parental origins. For each chromosome, the ratio of
reads from paternal or maternal alleles was calculated (Supplementary
Methods). In total, 150 cells were included. For each developmental stage,
the mean value for maternal or paternal alleles (maternal (y > 0), paternal
(y < 0)) were calculated for chromosomes X and 1. Mean values are shown
as pink and blue centre lines for maternal and paternal alleles, respectively.
662 | NAtUre | VOL 572 | 29 AUGUSt 2019