Di-phosphorylated ERK1/2 is enriched in the 4d cell in O. fusiformis
O. fusiformis has a single ERK1/2 ortholog already expressed at low levels in active oocytes (Figure S1A, B). The expression of erk1/2 is stable during early cleavage divisions and up to 4 hours post fertilisation (hpf; ~32 cell stage), increasing thereafter and plateauing at the larval stage (Figure S1B). A cross-reactive antibody against the active di-phosphorylated form of ERK1/2 (di-P-ERK1/2)19,23 shows weak ubiquitous signal in the zygote and up to 3 hpf (~16 cell stage) (Figure S1C). At 4 hpf, however, the four most animal/apical micromeres exhibit enriched di-P-ERK1/2 signal, which disappears at 5 hpf (coeloblastula) (Figure 2A, Figure S1C). At this stage, a single cell in the vegetal pole shows strong signal for di-P-ERK1/2 (Figure 2A, Figure S1C). This cell is one of the fourth quartet of micromeres, or a 4q blastomere following the conventional spiralian nomenclature31 that uses a number to indicate the quartet and a lowercase or uppercase letter (q for quartet and a to d for the quadrant) to refer to the animal micromeres and the vegetal macromers, respectively. In O. fusiformis, the fourth micromere quartet (4q) starts to form at 4 hpf, when the macromeres of the third quartet (3Q) sharing the vegetal cross furrow divide (Figure 2B). This creates an asymmetry in the timing and order of appearance of the 4q micromeres that is also observed during cleavage of the 3q micromeres, which divide in a counterclockwise order after all 3Q macromeres have cleaved (Figure 2B). By 5 hpf however, the vegetal pole exhibits again a symmetrical arrangement (Figure 2B). At 6 hpf, with the onset of gastrulation29, di-P-ERK1/2 signal is enriched in seven vegetal cells forming a bilaterally symmetrical pattern that includes one 4q micromere and two blastomeres of each of the other three embryonic quadrants (Figure 2A, Figure S1C). At this stage, the macromeres of the fourth quartet (4Q) have cleaved synchronously into the 5Q and 5q cells and all but the 4q micromere enriched in di-P-ERK1/2 have divided. This is the earliest morphological sign of bilateral symmetry in O. fusiformis embryos. The di-P-ERK1/2 positive 4q micromere divides into two large cells only after ingression during early gastrulation at 7 hpf (Figure 2B), with the later progeny occupying a bilateral position on the posterodorsal side of the archenteron once gastrulation completes at 9 hpf (Figure 2B). The behaviour of this 4q micromere resembles the cellular dynamics described for the 4d micromere in other conditional spiral cleaving annelids32-35, which cell lineage analyses identify as the trunk mesodermal precursor in most annelid embryos36-38. Therefore, our data suggests that the di-P-ERK1/2 enriched 4q micromere at 5 and 6 hpf in O. fusiformis is the 4d cell (Figure 2C), resembling the condition observed in many gastropod molluscs18-21,23.
ERK1/2 signalling controls axial polarity in O. fusiformis
To examine the role of the ERK1/2 signalling during O. fusiformis development, we treated embryos with brefeldin A (BFA), an inhibitor of intracellular protein trafficking previously used to block the induction of the organiser in other spiral cleaving embryos28,39; and U0126, a selective inhibitor of MEK1/2 and ERK1/2 di-phosphorylation23,40 (Figure 3A). For both drugs, treatment from fertilisation (~0.5 hpf) to 5 hpf, when di-P-ERK1/2 is enriched in 4d, effectively blocks activation of ERK1/2 (Figure 3B; Figures S2A; Table S2) and causes the loss of bilateral symmetry, posterior structures (e.g., chaetae and hindgut) and larval muscles in a dosage-dependent manner up to 100% of the embryos at a 10 µM concentration (Figure 3C; Figure S2B; Table S3, S4). Compared to control samples, 0.5 to 5 hpf treated embryos lack a fully formed apical tuft and apical organ, showing reduced ectodermal expression of the apical organ marker six3/63 and just a few apical cells positive for the neuronal marker synaptotagmin-1 (syt-1)29 (Figure 3D). In addition, treated embryos lack expression of hindgut (cdx) and trunk mesodermal (twist) markers3, exhibit expanded expression of the oral ectodermal marker gene gsc3 around the single gut opening (Figure 3D; Figure S2C), and retain expression of the midgut endodermal marker GATA4/5/6b3 (Figure S2C). We deem this phenotype as antero-ventrally radialised (or Radial; Figure 3C, E). Therefore, activation of ERK1/2 signalling in the 4d cell at the coeloblastula stage relies on inductive cell-cell communication (impaired by BFA treatment) required to specify and develop posterior and dorsal structures during O. fusiformis embryogenesis.
To dissect the exact timing of induction and activity of ERK1/2 during O. fusiformis cleavage, we treated embryos with 10µM BFA/U0126 in overlapping time windows from fertilisation to early gastrulation (Figure 3E; Table S5). Blocking protein secretion with BFA from fertilisation to the 8-cell stage does not affect normal development. However, BFA treatment between the 8-cell stage and 4 hpf results in larvae with all morphological landmarks of a typical mitraria larva and normal expression of tissue-specific markers, but with a compressed morphology (Figure 3F; Figure S2C). Only treatment with BFA from 4 hpf to 6 hpf, and hence spanning the formation of 4d, causes a radial phenotype, with treatments after 5 hpf being lethal (Figure 3E, Figure S2C). All of this suggests that the intercellular communication event inducing the activation of ERK1/2 in 4d happens between 4 and 5 hpf, right during 4q micromere formation (Figure 1E). Unexpectedly, preventing ERK1/2 di-phosphorylation with U0126 from the 2-cell stage until 4 hpf, when this induction event might begin, also causes a radial phenotype (Figure 3E), with just slight differences between certain timepoints (Figure S2C). Therefore, ERK1/2 activity is essential for normal embryonic patterning and posterodorsal development throughout most spiral cleavage in O. fusiformis. However, the combination of BFA and U0126 phenotypes suggests that ERK1/2 acts autonomously from the 2-cell stage to 4 hpf, while it requires of inductive cell-to-cell communication signals for its enrichment in 4d.
ERK1/2 signalling activates posterodorsal and mesodermal genes
To investigate the mechanisms through which ERK1/2 controls posterodorsal development in O. fusiformis, we next hypothesised that genome-wide profiling of gene expression in BFA and U0126 treated embryos would uncover upstream regulators and downstream targets of ERK1/2 activity. We thus treated embryos with either 10 µM BFA or 10 µM U0126 from fertilisation to 5 hpf (to cover both the autonomous and conditional phases of ERK1/2 activity) and performed RNA-seq transcriptome profiling in treated and controlled embryos collected right after di-P-ERK1/2 enrichment in the 4d cell (coeloblastula; 5.5 hpf) and at the larval stage (Figure 4A; Figure S3A, B). Differential expression analyses revealed 90 and 268 differentially expressed genes (DEGs; log2 (fold change) < -1.5 and FDR-adjusted p-value < 0.05) in BFA treated coeloblastulae and larvae, respectively, and 132 (coeloblastula) and 373 (larva) DEGs after U0126 treatment (Figure 4B). When considering all comparisons and removing redundancies, we detected a total of 628 DEGs, 414 (65.92%) of which were functionally annotated and enriched in gene ontology terms related to regulation of transcription, development, and cell fate specification (Table S6, S7, S8). Most of these DEGs were downregulated (Figure 4B, C; Figure S3C, D) and only three DEGs (cdx, fer3 and foxH) appeared systematically downregulated in both drugs and in the two developmental time points (Figure 4B; Figure S3D). Our approach thus revealed a confident and relatively small set of genes whose expression is strongly dependent on ERK1/2 activity and early inductive signals.
To validate that the DEGs are affected by 4d misspecification, we selected 22 DEGs for further gene expression analyses (Figure 4D, Table S9), including a variety of transcription factors recurrently involved in axial embryonic patterning (e.g., six3/6, gsc, cdx, AP2, foxQ2), mesoderm development (e.g., twist, hand2, foxH) and neurogenesis (e.g., POU4, irxA), Wnt ligands (wnt1, wntA and wnt4), TGF-b modulators (noggin and BAMBI), and Notch signalling components (delta and notch-like) (Table S8). Stage-specific RNA-seq data covering twelve developmental time-points, from the unfertilized oocyte to the mature larva, confirmed that the expression of all candidate genes upregulates at the time of or just after 4d specification and di-P-ERK1/2 enrichment in this cell (Figure 4D). These genes are expressed either in apical/anterior domains (noggin, BAMBI, foxQ2, POU4, six3/6, gsc), the posterior larval tip and chaetae (fer3, lhx1/5, wnt1, notch, msx2-a, irxA, AP2, wnt4, delta), the hindgut (cdx), or mesodermal derivatives (foxH, rhox, wntA, POU3, hand2, twist) (Figure 4D, Figure S4). Analysis of the expression of these genes in control and treated embryos at the coeloblastula (5.5hpf) and larva (24hpf) stages confirmed the expression domains of these genes disappear after treatment with either BFA or U0126 (Figure 4E, Figure S4), thus validating our RNA-seq approach. Altogether, our findings uncover a set of co-regulated genes that act downstream of ERK1/2 signalling and/or early inductive signals, and that are involved in the development of apical, posterodorsal and mesodermal structures in O. fusiformis.
ERK1/2 signalling specifies and patterns the D-quadrant
Our RNA-seq study and candidate gene screening revealed nine genes expressed at the vegetal pole at 5.5 hpf and whose expression was affected by either direct inhibition of ERK1/2 di-phosphorylation (cdx, delta, foxH, wnt1, wntA, rhox, fer3 and AP2) or impairing cell-to-cell inductive signals (gsc) (Figure 4E, Figure S4). None of these genes are expressed at 5 hpf, at the onset of ERK1/2 activity in the 4d micromere (Figure 5A). Instead, the early endodermal marker GATA4/5/6a3, whose expression is unaffected by BFA and U0126 treatment and thus is cell-autonomous, is symmetrically expressed in the gastral plate (including 4d) at this stage (Figure 5A; Figure S5A). At 5.5 hpf, half an hour after the initial activation of ERK1/2 signalling in 4d, the vegetal pole becomes bilaterally symmetrically patterned, but only two of the ERK1/2 dependent genes (cdx and delta) are expressed in the 4d micromere (Figure 5A). The ParaHox gene cdx, which is detected in the hindgut of O. fusiformis at the larval stage3 (Figure 4E; Figure S5B), becomes expressed in 4d, and later on in two cells of the 4d progeny at the gastrula stage (Figure 5B), as expected from the behaviour of 4d at those stages (Figure 1E). Similarly, the Notch ligand delta (Figure S5C, D) is expressed in 4d at 5.5 hpf, but also in most of the descendants of 1d at the animal pole, plus animal micromeres and ectodermal derivatives of the C and D quadrant at the vegetal pole (Figure 5A, C). To investigate how ERK1/2 might control activation of cdx and delta, we used Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) data at 5 hpf and identified transcription factors bound to the accessible regions associated with these genes (Figure 5D). These include transcriptional regulators known to be modulated by ERK1/2 phosphorylation, such as ETS, RUNX and GATA factors17. Therefore, ERK1/2 di-phosphorylation in 4d seems to control the activity of transcriptional regulators that induce posterior fates (cdx) and cell-cell communication genes (delta).
The other six additional ERK1/2-dependent genes pattern the micromeres surrounding 4d at 5.5 hpf, defining mesodermal and posterior ectodermal domains (Figure 5A). The transcription factor foxH (Figure S5E), which regulates mesoderm development during gastrulation in vertebrate embryos41-44, is detected in two micromeres adjacent to 4d (Figure 5A), whose progeny contributes to lateral ectomesoderm in the annelid Urechis caupo35. Later in development, foxH is expressed in a posterior V-shaped pattern of putative mesodermal precursors during axial elongation, fading away in larval stages (Figure S5B). The ligands wnt1 and wntA (Figure S5F), expressed in the posterior region and D-quadrant during development in the annelid Platynereis dumerilii45, are expressed in two bilaterally symmetrical columns of micromeres and wntA is also detected in two additional D-quadrant micromeres (Figure 5A). The homeobox rhox (Figure S5G), which is expressed in male and female primordial germ cells in vertebrates46, is detected in two vegetal micromeres in the D-quadrant (Figure 5A) and thereafter in two small cells inside the blastocele in the gastrula (Figure S4). The transcription factors fer3 and AP2 are expressed in two single micromeres and in a broader posterior ectodermal domain, respectively, becoming restricted to a small expression domain at or near the posterior chaetal sac of the larva (Figure S4; Figure S5B). The larger size of 4d, which stays as a big posterior cell at 5.5 hpf, allows it to establish almost as double direct cell-cell contacts with its surrounding cells than each of the daughter cells of 4a–c, including most of the cells expressing foxH, wnt1, wntA, rhox, and AP2 (Figure 5E). Altogether, these results demonstrate the inductive role of the 4d in defining mesodermal and posterodorsal fates. Moreover, the upregulation of the Notch-ligand delta in 4d after ERK1/2 activation suggests that the organising role of 4d might occur by direct cell-cell communication mediated by the Notch signalling pathway.
The homeobox gene gsc is the only candidate expressed outside the D-quadrant at 5.5 hpf, being detected in a U-shape domain of micromeres of the A, B and C quadrants that occupy the prospective anterolateral blastoporal rim (Figure 5A). Consistent with its location outside the D-quadrant, gsc expression is independent of ERK1/2 activity, but requires of inductive signals, as demonstrated by being downregulated after BFA treatment (Table S3). Accordingly, gsc expression disappears in BFA-treated coeloblastulae, while inhibition of ERK1/2 activity with U0126 expands gsc domains, which becomes detected in all embryonic quadrants in a radial fashion (Figure 5F). Indeed, in ERK1/2 inhibited embryos all 4q micromeres cleave into 4q1 and 4q2, as no 4q becomes the 4d cell (Figure 5F). Therefore, all quadrants adopt antero-ventral fates (gsc), preventing the expression of posterior marker genes such as AP2 (Figure 5F). These results demonstrate that the radial phenotype observed after U0126 is a consequence of misspecifying the 4d cell, and not a result of the D-quadrant becoming specified but not developing further. In addition, this data also demonstrates that specification of the 4d cell through ERK1/2 activity represses anterior fates, as shown by limiting gsc expression.
FGF signalling regulates ERK1/2 di-phosphorylation in 4d
In vertebrate embryos, FGF signalling mediated by ERK1/2 activity regulates the expression of cdx and delta in the developing hindgut and presomitic mesoderm during posterior trunk elongation47. We thus hypothesised that FGF signalling might be the upstream regulator driving ERK1/2 activity in 4d in O. fusiformis (Figure 6A), which is also required to induce the expression of cdx and delta and specify the hindgut and trunk mesodermal progenitor. O. fusiformis has a single FGF receptor (FGFR; Figure S6A) with high amino acid conservation at key functional residues compared to the human FGFR1 ortholog (Figure 7B). While it is transcriptionally upregulated at the gastrula stage, FGFR appears weakly expressed at the gastral plate at the coeloblastula stage in O. fusiformis (Figure S6C, D), as also observed in brachiopods and phoronids48. At this stage, FGF ligands are only weakly expressed (Figure S6C) and not detected by in situ hybridisation. Treatment with 30 µM SU5402, a selective inhibitor of FGFR phosphorylation and activation49 (Figure 6A), prevents di-P-ERK1/2 enrichment at 4d cell in 96% of the treated embryos at 5 hpf (Figure 6B, Figure S6E, Table S2). Indeed, treatment with SU5402 during the specification of the 4d micromere (4 to 6 hpf and 3 to 7 hpf; Figure 6C) results in an anteriorly radialised phenotype, with embryos developing into larvae lacking posterior and reduced mesodermal structures and showing radially expended oral ectodermal fates (gsc) (Figure 6D, Table S3). Conversely, treatment before di-P-ERK1/2 enrichment in 4d (3 to 5 hpf) causes a slightly compressed phenotype (Figure S6F). Therefore, inhibition of FGFR blocks ERK1/2 activation and phenocopies the effect of U0126 and BFA (Figure 3C), suggesting that FGF signalling activity is upstream of and necessary for 4d specification in O. fusiformis.