Systematic comparison of developmental GRNs explains how novelty is incorporated in early development

The impressive array of morphological diversity among animal taxa represents the product of millions of years of evolution. Morphology is the output of development, therefore phenotypic evolution arises from changes to the topology of the gene regulatory networks (GRNs) that control the highly coordinated process of embryogenesis1. While genetic variation can arise anywhere in the genome and affect any part of an individual GRN, the need to form a viable embryo provides a constraint on the types of variation that pass the filter of selection. A particular challenge in understanding the origins of animal diversity lies in determining how GRNs incorporate novelty while preserving the overall stability of the network, and hence, embryonic viability. Here we assemble a comprehensive GRN, consisting of 42 genes (nodes) and 84 interactions (edges), for the model of endomesoderm specification in the sea star from zygote through gastrulation that corresponds to the GRN for sea urchin development of equivalent territories and stages2. Using these detailed models we make the first systems-level comparison of early development and examine how novelty is incorporated into GRNs. We show how the GRN is resilient to the introduction of a transcription factor, pmar1, the inclusion of which leads to a switch between two stable modes of Delta-Notch signaling. Signaling pathways can function in multiple modes and we propose that GRN changes that lead to switches between modes may be a common evolutionary mechanism for changes in embryogenesis. Our data additionally proposes a model in which evolutionarily conserved network motifs, or kernels, may function throughout development to stabilize these signaling transitions.

territories and stages 2 . Using these detailed models we make the first systems-level comparison of 1 early development and examine how novelty is incorporated into GRNs. We show how the GRN 2 is resilient to the introduction of a transcription factor, pmar1, the inclusion of which leads to a 3 switch between two stable modes of Delta-Notch signaling. Signaling pathways can function in 4 multiple modes and we propose that GRN changes that lead to switches between modes may be a 5 common evolutionary mechanism for changes in embryogenesis. Our data additionally proposes 6 a model in which evolutionarily conserved network motifs, or kernels, may function throughout 7 development to stabilize these signaling transitions. 8 9 The regulatory program that controls development is unidirectional and hierarchical. It initiates 10 with early asymmetries that activate highly coordinated cascades of gene regulatory interactions 11 known as a gene regulatory network (GRN). GRNs function to orchestrate the intricate cellular and 12 morphogenic events that comprise embryogenesis 1,3 , and their topologies must be structured in ways  The GRN for the specification of sea urchin endomesoderm is the most comprehensive, 19 experimentally derived GRN known to date 2,10-11 . It explains how vegetal-most micromeres express 20 signaling molecules, including Delta, needed to specify the adjacent macromere cells to 21 endomesoderm, how micromeres ingress as mesenchyme, and are finally specified to form a 22 biomineralized skeleton. This GRN initiates with the maternally directed nuclearization of β-catenin 23 which activates the paired homeodomain TF pmar1 12 . Pmar1 represses the expression of hesC. The

24
HesC TF is a repressor of genes encoding many of the TFs needed to specify micromere fate (i.e. alx1, ets1, tbr, and tel) including the delta gene. The activation of Pmar1 therefore indirectly leads to the 1 expression of many of the regulatory genes within the vegetal pole, micromere territory in what has 2 been termed the double-negative gate 13 . The Pmar1 TF appears to be a novel duplication of the phb 3 gene, and is found only in sea urchins 14 , and the Pmar1 repression of hesc, i.e. the double-negative  Understanding the impact of integrating this novel subcircuit into early development demands a 9 systems-level approach: not one limited to local properties around the new circuit, but an 10 understanding of how the network as a whole responds to the change. Therefore, we assembled a 11 detailed GRN for sea star endomesoderm specification through gastrulation (Extended Data Figure 1).

12
An interactive, temporal model, including primary and published data, is hosted on a web server 13 (grns.biotapestry.org/PmEndomes) which allows for further and more fine-grained exploration. This 14 GRN was produced using the same experimental approaches as those used to generate the sea urchin 15 network 20 to allow for a meaningful comparison; i.e., whole-mount in situ hybridization (WMISH) to  In contrast to the sea urchin 13 ( Figure 1F), sea star blastula stage embryos co-express the delta 23 and hesC transcripts throughout the endomesoderm-fated vegetal pole ( Figure 1A), but their 24 expression is partitioned into adjacent cells by midway through gastrulation ( Figure 1B). Indeed, in all known species of echinoderms that lack an identifiable pmar1 gene, hesC remains expressed within the 1 delta + territory of the blastula mesoderm 15,17,21 . The related Phb genes have recently been shown to act 2 as positive inputs into endomesodermally expressed hesc in sea star and cidaroid sea urchin 14 . We   The summary view of these results permits a global comparison of these echinoderm GRN 22 topologies ( Figure 3A), which are the synthesis of over a decade of work including the present study.

23
Immediately apparent are the several distinct subcircuits found in common between these networks.

24
Common modules include ets1, erg, hex, and tgif in the mesoderm 38,39 ; otx, gatae, bra, and foxa in the endoderm 40,41 ; and pax6, eya, six3, dach, and six1/2 in the coelom (Extended Data Figure 8, 37 ). In 1 contrast, entire subcircuits present in sea urchins, i.e. dri, foxb, and vefgr that direct batteries of 2 skeletogenic differentiation genes in the sea urchin micromeres, are entirely absent from the sea star 3 network. This comparison also reveals that similarly regulated subcircuits are highly positively cross 4 regulated, in keeping with the previous definition of network kernel 42 . It was previously suggested that 5 such kernels would be found in early development, as they function downstream of maternally derived 6 and transient signals, to stabilize gene expression needed to specify distinct embryonic territories 43 . A 7 stabilizing function is thought to be derived from the intra-circuit positive regulatory feedback. Here 8 we show that these kernels appear throughout the GRN and are not limited to only early development. 9 Preliminary experimental analyses from other species of echinoderms 15,44-46 suggest these kernels are 10 present in multiple species and thus represent a genuinely conserved, rather than convergent, feature of 11 GRNs. The mechanistic basis for the evolutionary stability of these subcircuits remains unclear and it 12 will be important to define additional such network motifs to begin to understand whether the observed 13 stability is a cause or a consequence of the observed highly recursive regulatory wiring of these  Indeed, despite this transition in Delta-Notch signaling, we find that the GRNs in both taxa converge to a conserved regulatory subcircuit that directs the fate of coelomic mesoderm. Therefore, in contrast to 1 previous expectations, evolutionarily stable network kernels that were proposed to function to 2 lockdown early developmental regulatory states 3 , are not restricted according to network hierarchy. We 3 propose instead that these network subcircuits act as stabilizing features throughout development, 4 functioning as developmental checkpoints through which embryogenesis must transit ( Figure 3B).