A uterus-inspired 3D niche drives embryo development beyond implantation

The mammalian embryo must undergo dramatic morphogenetic changes to invade the uterine endometrium and achieve implantation. Thus, recapitulation of implantation using in vitro systems is crucial for revealing the mechanisms controlling early development and the main problems compromising human fertility. Experimental systems based on two-dimensional (2D) platforms cannot fully recapitulate the in vivo 3D microenvironments of the embryo. Therefore, here we use collagen grafted onto polydimethylsiloxane (PDMS) based on the uterine mechanics and microstructure to establish a uterus-inspired 3D niche (U3N). Our U3N enables mouse embryos to form egg cylinders at high rate and reach the developmental stages of heartbeat. Moreover, a unique interface forms between the embryo and collagen, showing the invasion of trophoblasts into collagen res, which simulate the developmental process of implantation. Our ndings highlight embryo-substrate interaction as a key characteristic of post-implantation development in vitro and as an important design parameter of 3D conditions for embryo culture.


Introduction
Implantation of the embryo into the uterus represents one of the key events of mammalian development 1 . Trophectoderm (TE) cells of the blastocyst make the rst contact with the uterine epithelium for implantation. The contact part of the embryo is de ned as mural trophoblast and the opposite part far from the uterus is called as polar trophoblast. When mural trophoblast attached to uterus, the epithelial cells of which become receptive for embryos with the stimulation of progesterone and estrogen.
Subsequently, the mural trophoblast cells develop into trophoblast giant cells which secrete matrix metalloproteinases to rebuild the extracellular matrix (ECM) for the invasion of embryos into uteri on 4.5 days post coitum (E4.5). The TE of embryos differentiate into extraembryonic ectoderm (ExE) and epiblast cells (EPI) of the embryo develop into pro-amniotic cavity after implantation. The mature amniotic cavity is formed from EPI and ExE in egg cylinder on the stage of E5.5-E6.5 followed with a series of developmental events including gastrulation, rst appearing of allantoic, amniotic fold, rst brachial arch etc. The incorrection of any of these events will lead to abnormality of the embryo development and even pregnancy failure 2 . Yet, the development of post-implantation embryos remains poorly understood because the embryo becomes closely tied to the mother and is embraced by the maternal uterine tissue of the decidua, which poses a challenge to the direct observation of embryonic organogenesis in utero 3,4 . Culturing whole-embryo cultures on top of a matrix which mimic the uterus serves as an excellent model to study embryo morphogenesis and organogenesis [5][6][7][8][9][10] .
Early studies have cultured mouse embryos on top of a bre network of bovine eye 11 from blastocyst stage to 5.0 days post coitum (E5.0). Since then, a series of in vitro culture (IVC) systems have tried to simulate the conditions of womb via different approaches. These include tissue culture plates with various coatings such as collagen, laminin or bronectin 12,13 ; co-culture with primary cells isolated from the uterus 14 ; co-culture with somatic cells, such as broblasts and Vero cells 15 ; small culture dishes coated with polyacrylamide hydrogel on the bottom of the glass plate 16 ; and matrigel basement membrane matrix 17,18 . These studies have found that specialised materials such as collagen allowed better embryo development 19 . However, rare studies focus on the development beyond implantation and the accurate simulation of uterine environment remains a massive challenge because there are great changes in terms of cell migration and tissue morphology for the embryo and the uterus together, in which there are also complicated physical and chemical factors involved.
Biological materials offer the possibility to manipulate mechanical hardness, elasticity and chemical modi cations to affect key cell behaviours, such as cell migration, proliferation and differentiation [20][21][22][23][24] . The mechanical properties of matrix could also in uence embryo development. The modulus for the oviduct and uterine epithelium is approximately 100-1000 Pa, with a difference of almost six orders of magnitude from the conventional polystyrene culture plate, which is about 1 GPa 25,26 . There is an urgent need to better understand the relevant mechanisms and design optimal materials inspired by uterine environments to accomplish better embryo culture in vitro.
In this study, we established a uterus-inspired 3D niche (U3N) with a layer of collagen on a PDMS substrate, which closely mimic the physical modulus and porous structures of the uterus matrix. Adjacent radial microvilli of the embryo rearrange the collagen bres at the interface and the implantation-related genes were also detected, suggesting this co-culture system is similar to the implantation in vivo. to create The U3N system drives embryo development beyond implantation to E8.5 in vitro (Fig. 1a, b) and the success of this approach was evidenced by the formation of key structures, including the head, the blood island of the yolk sac and the heartbeat. Furthermore, there was a higher ratio of the formation of egg cylinder and heartbeat for the embryos in U3N system.

Results
Construction of uterus-inspired 3D niche for better early embryonic development In order to reproduce the mechanical and structural environment of the endometrium, we rst measured the modulus of the endometrium and the uterine horn in adult mice ( Supplementary Fig. 1a) and found the modulus ranges of the endometrium and the uterine myometrium are 2.37 ± 0.81 kPa and 0.41 ± 0.16 MPa, respectively. We then measured the modulus of 7.0 mg/ml collagen and PDMS with mixing ratio 10:1 (base to crosslinker) and they were 0.83 ± 0.13 kPa and 0.92 ± 0.10 MPa respectively and the commonly used Petri dish is 1 GPa as we know. Modulus of endometrium and collage were both at kPa levels, while myometrium and PDMS were both at MPa levels, suggesting a combination of collagen and PDMS can mimic the mouse uterus well (Fig. 1b).
The embryo-uterus synchrony is critical for successful implantation and subsequent development 27 . To better mimic the uterine, we carefully studied the macro and micro structure of the uterus matrix and found there were special patterns. The diameter of the bre (D) and the size of the pore cross-sections (S) of the uterus matrix were 0.09 ± 0.02 µm and 0.72 ± 0.21 µm 2 (Fig. 1d). Then, we made porous hydrogel from collagen with speci c concentration (7.5 mg/mL) collagen with D = 0.12 ± 0.02 µm and S = 0.69 ± 0.17 µm 2 , which was similar to the uterus matrix (Fig. 1d). Based on the above two points, we designed a U3N by grafting a layer of collagen (CO; 7.5 mg/ml; 46 ± 10 µm in thickness) onto the PDMS (PD; 496 ± 156 µm in thickness) and form the PDCO to the uterus during pregnancy ( Fig. 1b and Supplementary Fig.  1).
We harvest E3.5 embryos from pregnant female mice and transferred them onto petri dish, PD or PDCO ( Fig. 2a and Supplementary Fig. 2). This co-cutlre system was maintained in vitro for over 10 days ( Supplementary Fig. 2a). Compared with embryos cultured on PD (16 ± 8%, n= 64 embryos), those grown on PDCO (36 ± 6%, n= 71 embryos) exhibited a higher ratio of two cavities (pre-promamniotic cavity and TE cavity) on day 3 of co-culture ( Fig. 2b and Supplementary Fig. 2b). Additionally, there was a higher rate of formation of the egg cylinders (47 ± 15% for PDCO and 32 ± 11% for PD) and an enhanced developmental e ciency of the embryos with heartbeat (11 ± 5% for PDCO and 0% for PD) in the embryos on PDCO (Fig. 2b) compared with those on PD or CO alone ( Supplementary Fig. 2). Furthermore, the embryos cultivated to E4.5 on PDCO were transferred into the uteri of pseudo-pregnant mice and then the uteri were collected on 6 days after transplantation. We could observe successful decidua formation (36.7%, n = 24 embryos), which was nevertheless not noticed in the PD group ( Fig. 2c) (n = 29 embryos).
The result con rms the necessity of establishing a uterus-like environment from both the macro (modulus of the uterus horn and endometrium) and micro (microstructure of the endometrium) perspectives.
Cell lineage segregation is crucial for early embryo development and is a criterion to de ne the quality of an embryo. For the reason, we performed single-cell RNA sequencing on embryos cultured on PD or PDCO and then analysis con rmed that the cell population of embryos grown on PDCO on day 2 of IVC is closer to E4.5 embryos in vivo. Although embryos grown on PD and PDCO had cell populations of trophoblast (TE), primitive endoderm (PrE), epiblast (EPI) and inner cell mass (ICM) (Fig. 2d-f and Supplementary Fig.   3), which was consistent with previous ndings 28 . EPI gives rise to almost all of the fetal tissues and the statistics showed embryos in the PDCO group had more EPI and fewer TE cells than those in the PD group (Fig. 2f).
To further investigate the relationship between embryo development and the 3D substrate, we optimised U3N by adjusting the key parameters of collagen in terms of the thickness and concentration, based on which we designed a series of experimental groups and the schematic diagram and the abbreviation of the groups were shown in Supplementary Fig. 4. We rst tested the effects of the thickness of collagen on embryo development. Our results suggested that higher rates of embryo with two cavities lead to a trend towards an increased rate of egg cylinder embryos on day 4 of IVC, although this was not statistically signi cant for the PDCO-t group, or a modi ed PDCO with the middle thickness of the top layer of collagen ( Supplementary Fig. 5a, b). We then xed the thicknesses of both collagen (46 ± 10µm) and PDMS (496 ± 156µm), but modi ed the concentration of collagen (5.0 mg/ml, 7.5 mg/ml, 10.0 mg/ml) and generated PDCO-L (5.0 mg/ml), PDCO-M (7.5 mg/ml) and and PDCO-H (10.0 mg/ml) ( Supplementary Fig. 4). We observed that the formation rate of egg cylinder for PDCO-M, with a moderate level of collagen concentration (7.5 mg/mL), was higher than those of other groups with either higher or lower concentrations at the same thickness of collagen ( Supplementary Fig. 5c, d).
The above results all showed that the concentrations and the thicknesses of the collagen in the U3N system somehow determine the growth of the embryo. And the collagen at a concentration of 7.5 mg/mL and a thickness of 45.6 µm can best support embryo growth. In order to clarify why this is the truth, we collected the PDCO-t and PDCO-T with different collagen thicknesses and carefully examined gel porosity in terms of the surface and the section using scanning electron microscopy (SEM), adopting freeze-fracturing for internal analysis. The collagen precisely represented the uterus matrix in terms of the bre diameter and pore area in the front and cross-sections at a concentration of 7.5 mg/mL and with a thickness of 45.6 µm (Supplementary Fig. 6). From this, we concluded that the modi ed PDCO with a moderate thickness of collagen (PDCO-t) allowed better embryonic development.
Interactions between peri-implantation mouse embryos and uterus-inspired 3D reconstructed environments Epithelium-trophectoderm adhesion and interaction is the initiated stage of implantation. However, there were few reports describing the maternal-fetal interface using the ex-vivo long-term culture system [29][30][31][32] . We investigated whether there was invasion of embryos into the U3N system through confocal imaging of the embryos and the substrates. The embryos exhibited a at morphology with a prominent nucleus on the surface of the PD after being attached, a signi cantly lower proportion of CDX2-positive cells aggregated (Fig. 3a). By contrast, the attached embryos on PDCO showed a dome-shaped morphology, with uniform nuclei and rearranged collagen bres around the contact interface. types of hormone such as Apln, Sct, Tspan33 and Actn1 (Supplementary Fig. 9a). The expression of 12 implantation-related genes was also detected, such as Emp2, Itgb4, Cdh1and Igf2 ( Supplementary Fig.  9b).
To assess the real connection between embryos and materials during attachment and invasion, SEM imaging was performed and displayed the formation of radial microvilli where the embryos attached. A peculiar dynamic traction of collagen bres was also found (Fig. 3b). By contrast, embryos did not grow out visible microvilli with slippy morphology on PD (n = 5) (Fig. 3b). We use TS cells to mimic TE cells, it was attempted to use the PDCO culture system to support mouse embryo stem cells (ESCs) and TE stem cells (TSCs). The result showed that, relative to ESCs, TSCs and their derivatives showed an obviously similar microvilli structure to what we observed for the embryos on PDCO (Fig. 3c), which demonstrates that TE plays a prominent role in surface interactions when embryos contact collagen bres. Furthermore, trophectoderms detailed characterized on day 2 of IVC to see the effects of 3D condition on embryo development in terms of staining of differentiation marker, mechanic properties and RNA expression analysis because trophectoderm is the distinct part of the embryo to be attached to the substrate.
We found that the appearance of embryos was different on the day 2 of IVC, the embryos on PD had a relatively larger ag spreading out than the ones on PDCO (Fig. 4a, b). Immuno uorescent imaging con rmed that the spreading areas induced by the interaction between the embryos and substrates were dominated by trophectoderm differentiation (TFAP2C positive, Fig. 4c) 35,36 . Cells with lower modulus were demonstrated with higher motility [37][38][39] . Atomic force microscopic (AFM) measurement discovered TE-derive cells with a higher modulus on PD than the ones on PDCO (Fig. 4d-f), which represented one of the reasons why cells could invade collagen. The differentially expressed genes of embryos on PD and PDCO showed that upregulated genes were enriched in negative regulation of the intrinsic apoptotic signalling pathway and the oxidation-reduction process (Fig. 4g). Another gene expression comparison indicated that the upregulated genes in embryos cultured on PDCO were mainly clustered on embryo implantation, such as Fkpb4, which is located in the decidualised stromal cells around the implantation site 40 ; embryo organogenesis, such as Wnt7b, which is involved in placental development and angiogenesis 41 ; Grhl2, which is related to forebrain development and cell proliferation 42,43 ; Coro1b, which is correlated with F-actin and regulates cell migration 44,45 ; and S1pr1, which improves the movement of smooth muscle cells, the progesterone receptor signalling pathway, cellular response to retinoic acid, negative regulation of intrinsic apoptotic signalling pathway and oxidation-reduction process, among other things (Fig. 4g). These results showed that the embryos on PD and PDCO were different in morphology, cell modulus, and gene expression, which initiated on the day 2 of IVC. Taken together, these data substantiated the in uence of subtracted on the trophectoderms, which may further dominate the embryonic development.

Development of embryos supported by PDCO beyond gastrulation
In order to clarify the stage of development for embryos cultured on PDCO, we rst evaluate the embryo morphology at distinct points during culture using standard microscopy according to comparison with the normal embryos in vivo. The embryos on PDCO could be cultured to the stage of E8.5 with typical structure at each point (Fig 5a). The patterns of morphogenetic cell lineage formation at each chronological had been also demonstrated in the cartoon gures (Fig 5b). Morphologically, red haematopoietic cells were visible in the yolk sac of embryos on day 8-10 of IVC ( Supplementary Fig.  10a), and these may have been nucleated red blood cells. At the same time, an embryonic heartbeat was distinctively visible behind the pharyngeal of the body (Supplementary Fig. 10b and Supplementary Movie 3). Furthermore, we stained the embryos with OCT4 for EPI, FOXA2 for PrE and CDX2 for TE on day 4 of IVC. The results of confocal imaging indicated that the embryos supported by PDCO exhibited the structure of typical egg cylinder on day 4 of IVC (Fig. 5a). Normally, embryos undergo a delay when cultured in vitro compared to those in vivo 46,47 . The embryos still experienced a delay in PDCO system, whereas which allowed embryo developing into the pre-gastrulation stage 1 day earlier compared with the other culture systems (Figs. 6a). When IVC reached the fth day, the embryos manifested the asymmetric structures of amnion cavities, which implied that embryos had developed into early gastrulation stage and were similar to the E6.5 embryos in vivo (Figs. 6b). To further con rm the effects of long-term culture for embryos, the U3N system was also superior not only with regard to the morphology of embryos (Fig. 6b) but also for transcriptomic expression. Unsupervised clustering analysis of RNA-Seq data showed that comparable transcriptome characteristics were observed among the embryos on day 8-10 of IVC and the embryos of E7.5 and E8.5 (Fig. 6c). To assess cell fates over a period of development, we tested speci c lineage markers of gastrulation and different organogenesis, including neurulation, yolk sac haematopoiesis and cardiogenesis. The embryos on day 8-10 of IVC also expressed more haematopoiesis-related genes, such as Wdr5, Farsa, Rap1a, Myo1c, Bud31, Mrpl38, Calm2 and Hspa4, than E8.5 embryos (Fig. 6d). Transcriptome analysis also con rmed that the IVC embryos expressed cardiogenesis-related genes, such as Acta2, Efhd2, Tpm1, Adk, Vdac2, Selenow, Pkp2 and Fermt2 (Fig. 6e). Markers of nerve, primitive streak and notochord-related genes were also detected in the embryos on PDCO (Supplementary Figs. 10c, d  and 11).
This work presents the application of 3D conditions in mediating embryo growth in vitro to the stage of organogenesis. The uterus-inspired 3D niche, U3N was created through combination PDMS and collagen and the optimization of the thickness and concentration collagen. We found that TE volume expansion was restricted over the attached substrate (XY pane) and exceeded into the collagen gel vertically, subsequently leading to high e ciency of embryo development. In contrast, there was larger extension area of TE when embryos cultured on PDMS with collagen coating (PDCO-C) but there was no vertical invasion, which resulted in lower e ciency of embryonic development. The invasion of the embryo through bres, which authentically mimics early mouse embryonic development in vivo. It is known that TE-derived cells secrete matrix metalloproteinases and cathepsins to digest ECM and maternal cells, and embryos invade the uterus by remodelling ECM in vivo 33 . In vitro, we found that during embryo implantation, TE-derived cells were in close contact with collagen, and the collagen bres were remodelled in a way that maybe similar to embryo implantation in vivo.
Simultaneously, the modulus differences among TE-derived cells on PD and PDCO (Fig. 4f) indicated a synergy between cells and cell substrates for stiffness 48 . It was observed that the pores in PDCO were larger than those in PDMS, facilitating the penetration of embryos embedded on PDCO (Fig. 3a). Therefore, micro/nano structure in uences the movement of cells in a direction perpendicular to the surface or within the 3D structure; this result was also supported by single-cell RNA-seq pro ling revealed Hippo-pathway genes with higher expression (Supplementary Fig. 10e), which also in uence the expression of CDX2 49 , a critical gene for early embryo development and the differentiation of TE-derived cells 50,51 . Our ndings also provides a solution to a conundrum of culturing embryos in vitro , which can determine the in uence of the TGC formation on the later development of embryos 18,52 . Drastic changes in terms of embryonic structure and cell variation make it challenging to prolong normal development and to portray the precise status of the post-implantation embryo in vitro. Our 3D uterus-like structure enables new forms of research on the development of the yolk sac, and early embryonic organs, which can broaden knowledge of embryogenesis during post-gastrulation. In summary, the U3N exhibits an exquisite ability to promote embryo development in vitro in a way that produces a great rate of embryos reaching egg cylinder and heartbeat. Acceding to the opportunities for further investigation, we propose that the present platform could be useful tool for early developmental biology with novel mechanisms of cell fate determination, advanced transplantation therapy and tissue fabrication engineering.