Nature - USA (2020-01-23)

Nature | Vol 577 | 23 January 2020 | 539

lineage segregation occurred—but incompletely—in human 6-d.p.f.
blastocysts independent of sample sources across all samples, and cell
fates appeared fixated in 7–9-d.p.f. embryos (Extended Data Fig. 3d,
e). Thus, we used scRNA-seq in 7–9-d.p.f. 3D embryos to examine the
regulators that segregate TrBs, PrEs and EPIs. We observed EPI-, PrE- and
TrB-specific genes associated with different signalling pathways and
transcription factors (Extended Data Fig. 3f–j, Supplementary Table 1).
Lineage-specific gene comparison between previous results^18 and our
results showed that core lineage transcription factors are maintained
across different samples, while some differences in gene expression
may be contributed by different developmental stages of embryos
(Extended Data Fig. 3k, Supplementary Table 1.6).

AME–EPI separation
In contrast to mouse EPI, the proximal EPI in primates segregates an
additional AME lineage before gastrulation^19. However, human AME
remains unclear in the absence of good molecular markers. As tissue
morphogenesis requires cell–cell adhesion proteins^20 , we observed a
symmetrical distribution of E-cadherin (also known as CDH1) on the cell
membrane of columnar EPIs, while squamous AME displayed no to weak
E-cadherin expression (Extended Data Fig. 4a, b). However, widespread
E-cadherin distributions in human EPIs were concentrated on the apical
site of wedge-shaped mouse EPIs^21. Amnion formation requires signal-
ling from the basement membrane generated by visceral endoderm^21 ,^22.
The layer between PrEs and EPIs formed a laminin-containing basement
membrane, enveloping EPIs but not AME (Extended Data Fig. 4c, d).
The AME weakly expressed OCT4, NANOG and SOX2 (Extended Data
Fig. 4b, f ). At 6 and 8 d.p.f., E-cadherin was ubiquitously distributed
on EPI and TrB cell membranes, whereas laminin enveloped the entire
EPI cluster and was widely expressed in TrBs (Extended Data Fig. 4g–j).

At 10 d.p.f., laminin concentrated around EPIs to form the basement

membrane but was lost in the AME (Extended Data Fig. 4k, l).

We next checked expressions of ERIZN and WGA, which localize to

the apical surfaces of human-pluripotent-stem-cell-derived amnion^6 ,^8.

ERIZN equally contributed to the  apical surfaces of EPIs and AME

(Extended Data Fig. 4m). However, WGA expressed in extra-embryonic

cells, but not in EPIs and AME (Extended Data Fig. 4n), which indicates

differences between 3D-embryo- and human-pluripotent-stem-cell-

derived amnion. Given that obvious separation of AME–EPI occurred at

12 and 14 d.p.f., we analysed scRNA-seq from 12- and 14-d.p.f. EPIs in the

post- and PSA-EPI clusters (Fig. 2b). The t-SNE analysis classified these EPIs

into three clusters, termed AME, intermediate state cells and EPIs, on the

basis of their gene-expression profiles (Extended Data Fig. 4o, p). Com-

pared to EPIs, AME significantly downregulated pluripotency genes and

upregulated genes expressed in the AME of 12–17-d.p.f. monkey embryos

(TFAP2C, MSX2 and BMP4)^23 or self-organized amnion from human pluri-

potent stem cells (TFAP2A and GATA3)^6 ,^8 (Extended Data Fig. 4p). High

expression of hormone genes in AME (Extended Data Fig. 4q–u, Supple-

mentary Table 2) corresponds to the AME of human placentas producing

human chorionic gonadotropin (hCG)^24. These results indicate the AME

is a distinct population with specific phenotypes compared to EPIs.

Forming anterior–posterior polarity and PSA

Primitive streak remains poorly defined and remains an enigmatic struc-

ture in human embryos. One hallmark of primitive streak formation

is the epithelial–mesenchymal transition and upregulated N-cadherin

(CDH2), a mesenchymal marker^25. We found N-cadherin localized in

PrEs before 12 d.p.f. and was activated in some OCT4-expressing cells

outside the EPI or near the AME–EPI junction at 14 d.p.f. (Fig. 3a, b). The

result was confirmed by scRNA-seq data (Fig. 2c, Extended Data Fig. 5l).

7 d.p.f. 5 embryos 65 cells

8 d.p.f. 6 embryos 59 cells

9 d.p.f. 6 embryos 68 cells

10 d.p.f. 6 embryos 77 cells

12 d.p.f. 5 embryos 88 cells

14 d.p.f. 6 embryos 137 cells

cDNAs synthesis

and amplication

by SMART-Seq2

Library construction

and sequencing

Normalization

Quality control

Total 557 cells

555 cells

Lineage trajectory

Global transcriptional change

Lineage-specic genes

Lineage identication

a

b

t-SNE 2

t-SNE 1

20

10

0

–10

–20

–20 –10 0 10 20

ICM (52)

PrE (25)

EPI (126)

CTBs (159)

EVTs (40)

STBs (109) PSA-EPI (44)

555 cells, 1,483 genes

log

(FPKM + 1) 2

6 d.p.f.

7 d.p.f.

8 d.p.f.

9 d.p.f.

10 d.p.f.

12 d.p.f.

14 d.p.f.

6 d.p.f.

7 d.p.f.

8 d.p.f.9 d.p.f.

10 d.p.f.12 d.p.f.

14 d.p.f.

GAPDH

PPIAOCT4

NANOG

SOX2PRDM14

DPPA3GDF3

CDH1

CDH2GATA6

PDGFRAGATA4

SOX17

DUSP4TFAP2C

CK7GATA2

CDX2

TEAD4EOMES

TCEAL4CSH1

HLA-G

MMP2TBXT

Group

Time

Time Time

Group

ICM

EPIPSA-EPI

PrE

CTBs

STBs

EVTs

c

t-SNE 1

d

t-SNE 2

6 d.p.f. 8 embryos 63 cells

–20 –10 0 10 20

10

0

–10

–20

12

10

8

6

4

2

0

Fig. 2 | Lineage delineation by transcriptome using scRNA-seq. a, Schematic
of single-cell collection and transcriptome analyses. b–d, t-SNE analyses
revealed seven clusters, identified as the ICM, EPI, PrE, TrB (including CTBs,

STBs and EVTs) and PSA-EPI based on classical lineage-specific marker

expression (c) and developmental time (d). FPKM, fragments per kilobase of

transcript per million mapped reads. See also Extended Data Fig. 3.