Self-organization of a human organizer by combined Wnt and Nodal signalling

Abstract

In amniotes, the development of the primitive streak and its accompanying 'organizer' define the first stages of gastrulation. Although these structures have been characterized in detail in model organisms, the human primitive streak and organizer remain a mystery. When stimulated with BMP4, micropatterned colonies of human embryonic stem cells self-organize to generate early embryonic germ layers ¹ . Here we show that, in the same type of colonies, Wnt signalling is sufficient to induce a primitive streak, and stimulation with Wnt and Activin is sufficient to induce an organizer, as characterized by embryo-like sharp boundary formation, markers of epithelial-to-mesenchymal transition and expression of the organizer-specific transcription factor GSC. Moreover, when grafted into chick embryos, human stem cell colonies treated with Wnt and Activin induce and contribute autonomously to a secondary axis while inducing a neural fate in the host. This fulfils the most stringent functional criteria for an organizer, and its discovery represents a milestone in human embryology.

Extended Data Figure 1. Controls for investigating hESC PS initiation hierarchy

(a) Micropatterned colonies stimulated with IWP2, SB, or blank media, fixed and stained for germ layer molecular markers after 48hrs. This experiment was repeated at least n=3 times independently with similar results. (b) qPCR for BMP4 of unpatterened small colonies stimulated for 4hrs conditions arraigned on the x-axis. As consistent with model hierarchy, there is no significant induction of BMP4 by ACTIVIN, WNT3A, or itself. Error bars represent the standard deviation of n=3 biologically independent replicates and the measure of the center represents the mean. (c) Quantification of Figure 1D. In this and in all other analysis unless stated otherwise, nuclei were segmented using DAPI and intensity of immunofluorescence signal for each marker was normalized to the DAPI intensity. Single cell expression data was binned radially and averaged. The final radial profile represents the average of n=25 colonies and errors bars represent the standard deviation.

Extended Data Figure 2. PS germ layer quantification and EMT timing

(a) Quantification of Fig. 2A. The radial profile represents the average of n=25 colonies and errors bars represent the standard deviation. (b) Micropatterned colonies stimulated with BMP4, WNT3A, WNT3A+SB, or WNT3A+ACTIVIN and fixed and stained for primitive streak molecular markers SNAIL, E-CAD, and N-CAD after 12, 24, 36, or 48hrs. Note that WNT3A and WNT3A+ACTIVIN stimulated colonies turn on EMT markers faster than BMP4 or WNT3A+SB stimulated colonies, and have mostly downregulated SNAIL by 48hrs. This experiment was repeated at least n=3 times independently with similar results.

Extended Data Figure 3. Further micropattern fate characterization

(a–b) Micropatterned colonies stimulated with BMP4, WNT3A, WNT3A+SB, or WNT3A+ACTIVIN and fixed and stained for EOMES at 24 and 48 hours (a) or PITX2 at 48 hours (b). EOMES is highest in WNT3A and WNT3A+ACTIVIN treated conditions and is also dynamic, with its highest expression at 24 hours, coinciding with the onset of PS markers (Extended Data Fig. 2B). PITX2 is not highly expressed in any condition. This experiment was repeated at least n=3 times independently with similar results. (c) Quantification of Fig. 2J–K and Extended Data Fig. 3A–B. The radial profile represents the average of n=25 colonies and errors bars represent the standard deviation.

Extended Data Figure 4. Further organizer characterization

(a) 1000μm and 500μm diameter micropatterned colonies stimulated with WNT3A+ACTIVIN and fixed and stained for GSC and BRA at 24 hours. Note that as observed by Warmflash et al., for BMP induction, shrinking the colony size results in removal of center micropattern fate region, thus resulting here in a higher proportion of GSC expressing cells. This experiment was repeated at least n=3 times independently with similar results. (b) Quantification of (a). The radial profile represents the average of n=25 colonies and errors bars represent the standard deviation. (c) Scatterplot of single-cell expression of GSC vs BRA. Note that at 24 hours most cells co-express BRA and GSC, but that by 48 hours GSC expression is increased and BRA expression is decreased. Because of this we grafted micropatterns at 24 hours as well as at 48 hours post-stimulation, reasoning that earlier coexpression of BRA and GSC would result in greater graft contribution to axial mesoderm structures. (d) qPCRs of additional organizer markers, taken from RNA collected from 500μm diameter micropatterns stimulated with either BMP4, WNT3A, WNT3A+SB, or WNT3A+ACTIVIN for 24 or 48 hrs. With the exception of NOGGIN, the characteristic organizer secreted inhibitors DKK1, CER1, CHORDIN, LEFTY1, and LEFTY2, are all most highly expressed in WNT3A+ACTIVIN conditions. The high NOGGIN induction by BMP4 in hESCs has been noted before, and may represent a human-mouse species difference. NODAL, which in mouse is restricted to the organizer later in gastrulation, is also most highly expressed in WNT3A+ACTIVIN conditions. Error bars represent the standard deviation of n=3 biologically independent replicates and the measure of the center represents the mean.

Extended Data Figure 5. Generation and validation of RUES2-GLR cell line

(a) Sequencing of the targeted alleles of SOX17, BRA and SOX2 genes. No indels were detected. (b) The RUES2-GLR line maintain pluripotency normally, as assessed by the staining of typical pluripotency markers (OCT4, NANOG and SOX2). Scale bar is 100μm. This experiment was repeated at least n=2 times independently with similar results. (c) The RUES2-GLR line was karyotypically normal.

Extended Data Figure 6. Functional validation of RUES2-GLR cell line

(a) Specificity of the germ layer reporters. When induced to differentiate to individual germ fates, only the specific reporter was turned on. SOX2-mCitrine reporter was expressed during pluripotency and 3 days after neural (ectoderm) differentiation, BRA-mCerulean was turned on after 3 days of mesodermal differentiation and SOX17-tdTomato reporter was active after 3 days of endodermal differentiation. Scale bar is 100μm. (b) Snapshots of a time-lapse imaging of the RUES2-GLR line after 50ng/mL BMP4 in micropatterns, showing how differentiation starts from the edges and extends inwards. Scale bar is 100μm. (c) RUES2-GLR line reproducibly generated the typical self-organized concentric rings of germ layers when induced to differentiate with a step presentation of 50ng/mL BMP4 in micropatterns. Scale bar is 200μm. All of these experiments were repeated at least n=3 times independently with similar results.

Extended Data Figure 7. Control chick grafts

(a) Representative grafts for control conditions. With the exception of the BMP4 control condition, grafted hESC colonies were static, with the colonies either growing or dying in place. With BMP4, often the colonies were elongated, possibly due to hESC migration. In all control conditions, however, there was never induction of SOX2 in the host cells. Note that in the case of the WNT3A+SB graft shown, two colonies were grafted into two different locations. Scale bar is 500μm. Experiments were repeated at least n=3 times independently with similar results. (b) Confocal cross-sections showing co-expression of SOX17 (tdTomato) and FOXA2 or OTX2 in human cells that contribute to the secondary axes induced by a 24hr WNT3A+ACTIVIN stimulated hESC micropattern. Scale bar is 20μm. Experiments were repeated at least n=3 times independently with similar results.

Extended Data Figure 8. Further characterisation of the induced secondary axis

(a) Examples of classifying the notochord-like feature (NLF) based on morphology. For z=+19μm, one can discern the NLF as a tighter and brighter rod of cells running north-south that is also distinct and somewhat separated from the surrounding chick epiblast. For z=+46μm, one sees that paired elongated cells stick out ahead of the other cells in a continuation of the originally identified NLF. Other cells belonging to the NLF between z=+46μm and z=+19μm are obscured at these slices or out of focus, but can be easily identified slice-by-slice at the other z positions. Scale bar is 100μm. (b) Snapshots of Extended Data Video 1. Clockwise from top-left: yellow shows co-Sox17:tdTomato (blue) with human (red) cells; cross-section shows that chick and human cells arrange themselves into germ layers properly, and that they flank the central notochord-like feature indicated by the arrow (cyan); a proportion of human mesoderm cells contribute to part of the notochord-like structure, while the cyan-coloured cells without HNA (red) shows that the remainder of the NLF is composed of host cells. (c) Examples of donor hESC graft contributing to the induced notochord-like feature, imaged live 27hrs (left) and 23hrs (right) post-graft. Scale bars are 200μm (left) and 100μm (right). Similar notochord-like features were observed in at least n=10 independent biological replicates.

Figure 1. hESCs obey BMP->Wnt->Nodal PS signalling initiation hierarchy

(a) Model of proposed PS signalling initiation hierarchy in hESCs, along with indication at which step the inhibitors SB and IWP2 act. Like in mouse, BMP acts on WNT, and then WNT acts on NODAL. There is also positive feedback between WNT and NODAL. (b) RNA-seq expression of all known WNT ligands in pluripotency and after 4hrs of BMP4 in hESCs on 500μm diameter micropatterns. The results show that overall WNT transcription is low in pluripotency and that WNT3 is the only strong and direct WNT ligand target of BMP4 stimulation. Data is from previously published data set GEO accession number GSE77057. (c) qPCR of WNT3 and NODAL in small colonies of hESCs after 4hrs stimulation with each condition shown on x-axis. The data is consistent with the predictions of the model. Error bars represent the standard deviation of n=3 biologically independent replicates and the measure of the center represents the mean. (d) Pie sections are of representative 1000μm diameter micropatterned colonies stimulated with BMP4, BMP4+IWP2, or BMP4+SB and fixed and stained for germ layer molecular markers after 48hrs. All micropattern experiments were performed on at least n=3 separate occasions with similar results, and unless mentioned otherwise, all other micropatterns are 1000μm in diameter. Staining is quantified in Extended Data Fig. 1C.

Figure 2. WNT is necessary and sufficient to induce PS markers and morphology

(a) Micropatterned colonies stimulated with WNT3A, WNT3A+ACTIVIN, WNT3A+SB, or ACTIVIN and fixed and stained for germ layer molecular markers after 48hrs. Staining is quantified in Extended Data Fig. 2A. (b) Micropatterned colonies stimulated with WNT3A, WNT3A+ACTIVIN, WNT3A+SB, or ACTIVIN and fixed and stained for pluripotency marker NANOG after 48hrs. (c) Sharp boundaries are visible even in phase contrast of WNT3A, WNT3A+SB, and WNT3A+ACTIVIN micropatterns at 52hrs. (d–f) Micropatterns stained with DAPI (blue; d and f), COLLAGEN IV (white; d and e), BRA (red; f), and/or SOX2 (green; f) after 52hrs of WNT3A+SB. Note how COLLAGEN IV traces a sharp continuous circle between the periphery and the interior of colony and forms a dividing line between the epiblast-like SOX2+ interior and differentiating BRA+ primitive streak exterior. In addition, our EMT data shows that the BRA- and SOX2- cells express SNAIL. (g) and (h) Micropatterns stained with phalloidin (white; g) or DAPI (blue; h), SOX17 (red; h), and SOX2 (green; h) after 52hrs of WNT3A. Cross section shows an inner epithelial SOX2+ epiblast region folded back in on itself. Outside and above the fold is the mesenchymal SOX17 region. Such geometry seems to permit the mesenchymal SOX17 cells to migrate along the epiblast-like region in basal-to-basal contact. Corroborating this, no SOX17 cells were ever observed further inwards beyond the leading edge of the fold where the basal presenting surface ends. (i) Micropatterned colonies stimulated with BMP4, WNT3A, WNT3A+SB, or WNT3A+ACTIVIN and fixed and stained for EMT markers SNAIL, E-CAD, and N-CAD after 48hrs. Note that WNT3A and WNT3A+ACTIVIN stimulated colonies show less SNAIL at 48hrs because they started their EMT earlier, at 24hrs (see Extended Data Fig. 2B for 12, 24, and 36hr timepoints). (j–k) Micropatterned colonies stimulated with WNT3A, WNT3A+ACTIVIN, WNT3A+SB, or BMP4 and fixed and stained after 48hrs for marker set characteristic of anterior PS (OTX2 and FOXA2; j) or organizer specific transcription factor GSC (k). Staining is quantified in Extended Data Fig. 3A. All experiments were repeated at least n=3 times independently with similar results.

Figure 3. Human organizer induces secondary axis in chick embryo

(a) Schematic showing strategy for graft hESC-chick graft. (b–g) Secondary axis induced by 24hr stimulated RUES2-GLR colony into HH stage 3 chick; (b–d) SOX17-tdTomato (red) live marker at 0, 24, and 38hrs post-graft; (f–g) SOX2 (green) is prominent in the tip of the secondary axis 48 hrs post-graft, and does not overlap with the hESCs (e and g, red, Human Nuclear Antigen). Scale bar is 200μm. (h–k) Another example of a 24hr stimulated hESC micropattern inducing a secondary axis in a chick host, 27hrs post-graft; (h) live image of SOX17-tdTomato hESC cells (red); (i and k) fixed stains for Human Nuclear Antigen (HNA, red) and SOX2 (j and k, green). Scale bars are 500μm (h) and 200μm (i–k). (l–r) Example of secondary axis induction from a 24 hr stimulated hESC micropattern with more complete self-organizing structures, 27hrs post-graft; (l) live image of SOX17-tdTomato hESC cells (red); (m) confocal slice of secondary axis for DAPI (grey), HNA (red), and BRA (green); (n–r) confocal cross-section of indicated region in (m), with the same channels plus SOX17 (blue). Note in the merged image (r) how the secondary axis is layered, with epiblast chick cells on top of a layer of human BRA cells which in turn are on top of a layer of human SOX17 cells, exactly how the epiblast, mesoderm, and endoderm layers would arrange themselves in a gastrulating mouse or chick embryo. Scale bars are 500μm (l), 100μm (m), and 50μm (n–r). (s) In situ for chicken SOX3 shows expression in the host chick throughout the neural tube and head, as well as in the induced secondary axis. (t) OTX2 is expressed in the host forebrain but is absent in the graft induced tissue (indicated by arrow). (u) HOXB1 is expressed in the host and the graft induced secondary axis. (v) GBX2 is expressed in the host and the graft induced secondary axis. (w–x) Zoom of region indicated in (v): (w) shows secondary axis and tdTomato-hESCs (red) after fixation; (x) shows GBX2 expression after in situ. The arrow shows the location of the graft hESCs before and after. All experiments were performed at least n=3 times with similar results, for exact numbers and measure of reproducibility please see Table 1.