SOX17 Is Not Required for the Derivation and Maintenance of Mouse Extraembryonic Endoderm Stem Cell Lines

Extraembryonic endoderm stem (XEN) cell lines can be derived and maintained in vitro and reflect the primitive endoderm cell lineage. SOX17 is thought to be required for the derivation and maintenance of mouse XEN cell lines. Here we have re-evaluated this requirement for SOX17. We derived multiple SOX17-deficient XEN cell lines from preimplantation embryos of a SOX17-Cre knockout strain and chemically converted multiple SOX17-deficient embryonic stem cell lines into XEN cell lines by transient culturing with retinoic acid and Activin A. We confirmed the XEN profile of SOX17-deficient cell lines by immunofluorescence with various markers, by NanoString gene expression analyses, and by their contribution to the extraembryonic endoderm of chimeric embryos produced by injecting these cells into blastocysts. Thus, SOX17 is not required for the derivation and maintenance of XEN cell lines.


INTRODUCTION
During mouse embryogenesis, the morula undergoes compaction and gradually differentiates into the trophectoderm (TE) around the outer cell mass and the inner cell mass (ICM) (Johnson and Ziomek, 1981). By embryonic day E3.5, the second cell fate decision takes place involving the segregation of ICM into the pluripotent epiblast (EPI) and primitive endoderm (PrE), which are distributed in a salt-and-pepper pattern. By the late blastocyst stage, PrE forms a layer of cells along the surface of the blastocoel cavity. The epiblast gives rise to the embryo proper, amnion, and extraembryonic mesoderm of the yolk sac.
TE cells give rise to the placenta. PrE forms the two ExEn lineages: visceral endoderm (VE) and parietal endoderm (PE) of the yolk sac (Chazaud et al.,2006;Plusa et al., 2008;Artus and Hadjantonakis, 2012). VE cells and their derivatives play a critical role in organogenesis. VE cells are the first site of hematopoiesis (Toles et al., 1989;McGrath and Pails 2005) and form blood islands and endothelial cells through the expression of Indian hedgehog and vascular endothelial growth factor (VEGF) (Byrd et al., 2002;Damert et al., 2002). At gastrulation, VE cells contribute to forming the gut endoderm tissue of the fetus (Kwon et al., 2008). VE and PE function as an "early placenta" that is responsible for nutrient and waste exchange (Cross et al., 1994).
Extraembryonic endoderm stem (XEN) cell lines can be derived and maintained in vitro (Kunath et al., 2005;Niakan et al., 2013) and reflect the PrE cell lineage. There are four methods to derive mouse XEN cell lines. First, XEN cell lines can be derived directly from blastocysts (Kunath et al., 2005). Second, 4 XEN cell lines can be converted from embryonic stem (ES) cells by forced expression of XEN-specific genes such as Gata6 (Wamaitha et al., 2015), Gata4 (Fujikura et al., 2002), or Sox17 (McDonald et al., 2014 or chemically by transient culturing with retinoic acid (RA) and activin A (Cho et al., 2012). Third, XEN cell lines can be induced from fibroblasts by overexpression of classical OSKM factors (Parenti et al., 2016). Fourth, we have reported the efficient derivation of XEN cell lines from postimplantation embryos (Lin et al., 2016(Lin et al., , 2017. SOX17 is a member of the Sry-related high-mobility group box (Sox) transcription factors and has an essential role in the differentiation of several types of cells (Foster et al., 1994). During mouse embryogenesis, SOX17 is first detected at the 16-32 cell stages coexpressed with Oct4, then in the PrE of blastocysts, and later in VE at E6.0 and in the endoderm at E7.5, where it plays an essential role in organ formation (Kanai-Azuma et al., 2002). Previous studies have revealed its role in the regulation of fetal hematopoiesis (Kim et al., 2007) and vasculogenesis (Matsui et al., 2006;Sakamoto et al., 2007). SOX17 has also been proposed to function as a key regulator of endoderm formation and differentiation, a function that is conserved across vertebrates (Hudson et al.,1997;Alexander et al., 1999;Clements et al., 2000). In mice, genetic inactivation of SOX17 leads to severe defects in the formation of the definitive endoderm (Kanai-Azuma et al., 2002). Overexpression of SOX17 is sufficient to promote ES cells to convert to XEN cells (McDonald et al., 2014). SOX17 is critical for PrE formation, and a lack of SOX17 will significantly decreases the 5 PrE numbers of blastocysts (Artus et al., 2011). XEN cell lines cannot be derived from SOX17 mutant embryos and converted from ES cells (Niakan et al., 2010;Cho et al., 2012). Downregulation of SOX17 by RNA interference will impairs XEN cell maintenance (Lim et al., 2008). Embryonic bodies derived from SOX17 mutant ES cells fail to correctly form the outer ExEn layer (Niakan et al., 2010).

SOX17 mutant ES cells differentiate into PrE cells but fail to differentiate into PE
and VE fates (Shimoda et al., 2007). Here, we re-evaluated the requirement for SOX17 in the derivation and maintenance of XEN cell lines.
The model of sequential expression of PrE lineage-specific genes is Gata6 > Pdgfra > Sox17 > Gata4 > Sox7 (Artus et al., 2010(Artus et al., , 2011. Cells that express SOX17 can be visualized in a gene-targeted knockout mouse strain in which a fusion protein with green fluorescent protein (GFP) and Cre is expressed from the SOX17 locus (Choi et al., 2012) (although this strain is reported to contain and express GFP, and we confirmed the presence of GFP in this targeted insertion in the SOX17 locus, but we cannot detect GFP expression in embryos and cell lines derived from them). In this strain, which we refer to as SOX17-Cre, while SOX17-Cre crosses the gene-targeted Cre reporter strain R26-tauGFP41 (Wen et al., 2011), the GFP reporter is coexpressed with endogenous SOX17 protein and with PrE markers GATA6, GATA4, and DAB2 in preimplantation embryos (Choi et al., 2012;Lin et al., 2016). GFP colocalizes in the same cells with PrE markers GATA6 and GATA4 in blastocysts cultured in vitro and is expressed in the visceral and parietal endoderm of postimplantation embryos (Choi et al., 2012;Lin et al., 2021). XEN cell lines derived from R26-tauGFP41 × SOX17-Cre 6 heterozygous blastocysts display the intrinsic fluorescence of GFP (Lin et al., 2016(Lin et al., , 2021. Thus, in this strain, GFP serves as a robust live marker for PrE and its extraembryonic endoderm derivatives and can be applied in the context of XEN cell line derivation.

Derivation of XEN Cell Lines from SOX17-Deficient Blastocysts
In our previous work (Lin et al., 2017), we obtained a PDGFRA-deficient XEN cell line from blastocyst ICMs. In the experiments, we collected 27 E1.5-E2.5 embryos from SOX17-Cre heterozygous female intercrosses with SOX17-Cre heterozygous male and cultured them in KSOM medium to the blastocyst stage.
We previously noticed that in PDGRFA-GFP ES cell lines, sparse GFP+ cells surrounded rare ES cell colonies (ES cells typically do not express PDGRFA and are thus GFP-) (Lin et al., 2017). The occurrence of these cells is in agreement with observations that SOX17 is expressed in a subset of cells on the outside of otherwise undifferentiated ES cell colonies (Niakan et al., 2010), that ES cells cultured in LIF and 2i contain a few cells expressing GATA6 (Morgani et al., 2013), and that PDGFRA-GFP heterozygous and homozygous ES cells contain a fraction of GFP+ cells (Lo Nigro et al., 2017). It thus appears that some ES cells convert spontaneously to XEN-or XEN-like cells.
A low dose of retinoic acid (RA) and Activin A promotes the chemical conversion of ES cells into XEN cells (termed as cXEN cells) but fails to convert SOX17-deficient ES cells into cXEN cells (Cho et al., 2012;Niakan et al., 2010).
We followed the cXEN conversion protocol of Cho et al. 2012 with a slight change (Lin et al., 2017). The protocol for the conversion of cXEN from ES cells is shown in Fig. 2A. We cultured ES-9, ES-12 and ES-17 (R26-tauGFP41×SOX17-Cre F2 heterozygous) and ES-13 and ES-15 (R26-tauGFP41×SOX17-Cre F2 homozygous) for 48 hr in standard TS cell medium with F4H, to which 0.01 µM RA and 10 ng/ml Activin A were added. Thereafter, all cells were cultured in standard TS cell medium with F4H. XEN-like colonies accumulated on days 6 and 11 (Fig. 2B). We picked XEN-like colonies and passaged the cells into new plates. On day 20, XEN-like cells accumulated but still had a small fraction of ES-like colonies in the plates. Passaging of the mixed cells to single cells by trypsin releases ES-like colonies to many ES-like cells and grows many ES-like colonies, and then the ES-like cells dominate the population. In this case, we used a method that is called the crowded culture method or extended the culture method (Lin et al., 2017). We changed the medium every other day but passaged the cells every 1-2 weeks. As the cultures grew confluent, a fraction of the GFP+ (XEN-like) cells did not adhere tightly to the dishes and were easier to 11 lose during medium changes. It appears that whereas colonies of ES-like cells and differentiating ES cells adhered tightly to the dishes, XEN-like cells became sorted to the outside of these colonies

NanoString Gene Expression Analyses of XEN Cell Lines and ES Cell Lines
Next, we applied the NanoString multiplex platform (Khan et al., 2011, Lin et al. 2016 to compare patterns of gene expression in SOX17-Cre homozygous ES and XEN cell lines. All XEN cell lines had high levels of expression of XEN cell-specific genes, such as Gata4, Gata6, Pdgfra, Sox7, and Dab2, versus low levels of expression or no expression of ES cell-specific genes, such as Sox2, Pou5f1/Oct4, Nanog, and Zfp42/Rex1 (Fig. 3). In SOX17-Cre homozygous XEN cell lines, Sox17 expression was, as expected, absent.

DISCUSSION
We derived 6 SOX17-deficient XEN cell lines: two XEN cell lines (X-ICM-12 and X-ICM-16) and four cXEN cell lines .
Why are SOX17-deficient XEN cell lines not easy to derive from preimplantation embryos (Niakan et al., 2010)? The missing SOX17 signal reduces the number of PrE cells in blastocysts (Artus et al., 2013). These remaining PrE cells still have the ability to support the SOX17-null concepti to develop to E8.5 and E9.5 (Igarashi et al., 2018). The SOX17-null concepti showed high levels of GATA6 expression in PE cells at E8.5 and E9.5 (Igarashi et al., 2018). This result indicated that the model of sequential expression of PrE lineage-specific genes is Gata6 > Sox17 (Artus et al., 2010(Artus et al., , 2011. We observed that GFP+ cells could be maintained in culture and grew slowly to form large colonies. However, in the mixed ES-XEN cultures that we derived from blastocysts, ES cells grew much faster than XEN cells, and ES cells dominated after several passages. 16 Why were we able to chemically convert SOX17-deficient ES cell lines into cXEN cell lines whereas Cho et al., 2012 were not? First, we used a method called the crowded culture method or extended the culture method (Lin et al., 2017). We changed the medium every other day but passaged the cells every 1-2 weeks. The conventional method is to passage cells frequently (Niakan et al., 2013). We observed that SOX17-deficient ES cells were more difficult to convert than SOX17-Cre heterozygous ES cells in TS cell medium with F4H. Second, we collected cells suspended in the culture medium and spun down the medium to enrich for the XEN-like cells after plating into new dishes. We found that XEN cells cultured in TS cell medium were easier to be collected in suspension than in ES medium when colonies became crowded. The conventional method to change medium and passage cells entails removing the culture medium, which would also remove the suspended (XEN-like) cells.

The model of sequential expression of PrE cell lineage-specific genes is
Gata6 > Pdgfra > Sox17 > Gata4 > Sox7 (Artus et al., 2010), which is consistent with the failure in Gata6 mutant embryos to activate sequential expression of Pdgfra, SOX17, and Gata4 in the PrE of blastocysts (Schrode et al., 2014). Gata6 mutants exhibit a complete absence of PrE, while SOX17 or PDGFRA mutants exhibit only a reduced number of PrE cells (Artus et al.,2011(Artus et al., , 2013Schrode et al., 2014). This means that SOX17 or PDGFRA mutants could be partially rescued by other genes or pathways. We speculate that PDGFRA and SOX17 coud have parallel expression, because SOX17-or PDGFRA-deficient cells in blastocysts cannot block each other's expression (Artus et al., 2011(Artus et al., , 2013. We speculate that PDGFRA-deficient XEN cells could be rescued by SOX17 with parallel expression and, conversely, that SOX17 mutant cells could be rescued by PDGFRA in the PrE and in XEN cell derivation and maintenance with parallel expression. The SOX17-deficient XEN cell lines are healthy, grow as well as wild-type and SOX17-Cre heterozygous XEN cell lines, and differ thus far only in SOX17 expression from SOX17-Cre heterozygous XEN cell lines. The rate of chimeras among recovered embryos (37.5%) was similar to that obtained with other genetically marked pre-and post-XEN cell lines (35-39%) (Lin et al., 2016).
Further experiments, such as RNA-seq, may reveal differences in gene expression between SOX17-deficient and wild-type XEN cell lines.

Mouse Strains
The SOX17-Cre strain (Choi et al., 2012) was obtained from MMRRC, strain 036463-UNC, strain name SOX17<tm2(EGFP/cre) Mgn>/Mmnc. Although this strain is reported to contain and express GFP, and we confirmed the presence of GFP in this targeted insertion in the SOX17 locus, but we cannot detect GFP expression in embryos and cell lines derived from them. The R26-tauGFP41 reporter strain (Wen et al., 2011) was obtained from Dr. Uli Boehm, Universität des Saarlandes, Homburg, Germany.

cXEN Cell Conversion from ES Cells with Retinoic Acid and Activin A
The chemical conversion was performed as described (Cho et al., 2012)

Immunofluorescence and Imaging
Cell lines were cultured in 24-well dishes. Cells were fixed in 4%

NanoString Multiplex Gene Expression Analysis
Cells were collected by trypsinization and centrifugation. Cell pellets were dispensed in RNAlater Stabilization Solution (Qiagen) and stored at -80°C for later use. Cell pellets were lysed in RLT Lysis Plus Buffer using a TissueLyser LT (Qiagen) at 40 Hz for 2 min. Extraction of total RNA was performed using the RNeasy Plus Micro kit (Qiagen). The custom nanoString CodeSet "Extra" was used; sequences of relevant capture and reporter probes are in the Supplementary Information. An aliquot of 100 ng was hybridized at 65°C for 18 hr and processed with nCounter (nanoString Technologies). Background subtraction was performed using the maximum count of the negative control. A two-step normalization was performed: (1) the geometric mean of positive controls was used as the normalization factor across samples, and (2) the geometric mean of Actb and Gapdh counts was used as the biological reference normalization factor. A heatmap was generated using the 21 heatmap.2 function in the R package gplots.

AUTHOR CONTRIBUTIONS
J.L. and X.D. designed research, J.L. and A.D performed experiments. J.L. and X.D. analyzed data and wrote the manuscript.