Inherency of Form and Function in Animal Development and Evolution

Just as the structural motifs of animal bodies (e.g., layers, lumens, segments) are inherent to the materials that constitute them, so are the functional capabilities (e.g., contractility, excitability, detoxification) that came to be reflected in differentiated cell types, tissues, and organs. The reasons are not the same, however. Notwithstanding speculations based on the dynamical multistability of gene regulatory networks (GRNs; Kauffman, 1969; Furusawa and Kaneko, 2002), the full range of different cell types compatible with an organism’s genome does not seem to be the global state-space attractors of any unitary physical system. Here I do not mean to suggest that the nonlinear dynamics of GRNs and transitions between adjacent attractors of such systems are not suitable accounts of progression along developmental lineages, e.g., Huang et al. (2007). Such steps take place in functionally delimited domains of the full gene expression state space, with modest combinatorial differences in transcriptional determinants. There are persuasive examples of such phenomena (Mojtahedi et al., 2016; Corson and Siggia, 2017; Verd et al., 2017). However, the differential equation or Boolean representations of GRNs usually employed in the global models have long been criticized on conceptual grounds (Rosen, 1991) and have been further called into question by new findings on transcription factor disorder and condensed state chemistry reviewed in this section (see Niklas et al., 2015).

Instead, I suggest that the inherency of the specialized functions of multicellular organisms, such as animals and plants, resides in the fact that they are exaggerated counterparts of life-sustaining activities of their ancestral unicellular forms. Cellular functions in general have a history that goes back to the transition between the nonliving and living state (Arnellos and Moreno, 2012) but are not entirely shared in all clades. Some differentiated cell types amplify what are universal cell functions, but others draw on more directly ancestral physiology.

In this section, I show how cell types likely arose through the recruitment and partitioning of ready-made ancestral cellular functions by a highly plastic metazoan-specific gene regulatory apparatus. Although based on the “write and read” histone modification language that had evolved in the earliest eukaryotes (Prohaska et al., 2010), this apparatus is unlike that of any extant unicellular relatives. It appeared coincidentally with the metazoans, incorporating elements found in present-day holozoan and more distantly related fungi, as well as some unique elements not found in any other known life-forms. The new gene regulatory mode came to operate in a multicellular context that (as I will show below in relation to the role of the ancestral grainyhead transcription factor) was brought about by some of the same effects. Once metazoan lineage-defining transcription factors appeared (an evolutionary step that is currently enigmatic), partitioning of ancestral functions into differentiated cell types followed in a straightforward and potentially rapid fashion.

To reiterate, although metazoan-type mechanisms of cell differentiation were not inherent in unicellular progenitors, most cellular functions that came to be specialized in differentiated cell types were. Furthermore, once the enhancer-based, developmentally hierarchical means of exaggerating gene activity were in place, the evolution of developmental lineages of initially related, but progressively divergent, cell types was readily achieved.

While the genes of all categories of organisms are regulated by nearby upstream cis-acting promoters, the additional effects of enhancer sequences, which can be located upstream or downstream from their targets, or in introns, are, with the apparent exception of Placozoa, hallmarks of metazoan gene control. Enhancers are novelties of Metazoa or its direct antecedents: they are absent in extant unicellular holozoans (Sebé-Pedrós et al., 2016). There are more enhancers than regulated genes [as many as 50,000 in mammalian genomes, for example (Heinz et al., 2015)], and while their TF-binding motifs are similar to those of promoters (Arenas-Mena, 2017), they function differently. Enhancers mediate the high levels of expression of the characteristic genes of terminally differentiated cells, as well as the precise spatiotemporal generation of cell and tissue types during development due to their responsiveness to patterning cues (Lenhard et al., 2012; Zabidi and Stark, 2016) (see section “Pattern Formation: Inherency in the Integration of Form and Function”).

Enhancers show not only evidence of conservation but also divergent usage over evolution and between species (Chen et al., 2018). They are employed in sponges and all eumetazoans, with an indication of their presence in ctenophores, although not the placozoan T. adhaerens, which appears to rely solely on promoters to generate its few cell types (Sebé-Pedrós et al., 2018). Enhancers regulating stage- and cell type-specific “developmental” and “terminal differentiation” genes are often located at significant linear distances from the respective promoters and interact with a partly different set of TFs from the more proximal enhancers regulating ubiquitously expressed “housekeeping” genes (Zabidi et al., 2015).

The origin of enhancers may reside in the duality of promoter functional architecture in ancestral cells. Unicellular holozoans have inducible promoters that respond to external cues as well as promoters that regulate constitutive genes, with the latter often exhibiting distal cooperative interactions. Arenas-Mena (2017) has proposed a scenario by which inducible promoters were converted in few steps to a distally acting enhancer mode of developmental gene regulation in metazoans.

In metazoans, chromatin loops containing cell-type regulated DNA sequences enter or exit distinct liquid-phase proteinaceous condensates along with other loops containing enhancers and a large multi-subunit obligatory regulatory complex called Mediator, forming “topologically associating domains” (TADs) within the interphase nucleus (Galupa and Heard, 2017; Furlong and Levine, 2018; Plys and Kingston, 2018). Thousands of enhancers can be recruited to a given gene by lineage-determining transcription factors (LDTFs) (Heinz and Glass, 2012; Link et al., 2015) such as the muscle “master regulator” MyoD (Blum and Dynlacht, 2013). This occurs in a highly cooperative and synergistic fashion, with the number of enhancers recruited being responsive to developmental cues external to the differentiating cells.

Transcription in animal cells is initiated by the action of the master coactivators p300 and CBP (Chan and La Thangue, 2001). These highly homologous multidomain proteins contain histone acetyltransferase domains that relax the structure of nucleosomes at promoters and enhancers of all or most transcribed genes, enabling the recruitment of RNA polymerase II and TFs to those sites. Mediator, p300/CBP, and TFs all contain intrinsically disordered regions (IDRs) that assume different 3D structures under different conditions (Liu et al., 2006; Minezaki et al., 2006). The IDRs are required for the formation of nuclear TADs and are the basis for p300/CBP’s role as an interaction hub with a multitude of TFs and coregulators [most prominently nuclear receptors (NRs)] that mediate tissue responses to developmental and physiological cues (Dyson and Wright, 2016). Expression of a specific gene is caused by p300/CBP binding to transactivating protein domains (also abbreviated as TADs) of the TFs and NRs targeting the gene’s promoter, followed (based on evidence from some “super enhancers”) by induction by TF TADs of phase separation of Mediator into activating droplets (Boija et al., 2018).

Extant unicellular holozoans such as Capsaspora owczarzaki contain p300/CBP, and their promoter nucleosomes exhibit p300/CBP-specific histone acetylation marks. This indicates that the gene regulatory role of p300/CBP is phylogenetically older than that of enhancers (Grau-Bové et al., 2017). These coregulators contain bromodomains, which in metazoans promote enhancer-dependent expression of developmental and cell-type specific genes (Wei et al., 2008). Metazoans also express the bromodomain-containing protein Brd4, which co-localizes with LDTFs on active enhancers, where it recruits Mediator and RNA polymerase II (Lee et al., 2017).

Suppressing some genes can be as important as activating others in shaping cell type identity. In metazoan development and terminal cell differentiation, this is accomplished by “silencing” complexes such as the Polycomb-group (PcG) proteins (Khan et al., 2015) and other lysine methyltransferases such as SUV39H (Jih et al., 2017). These inhibit gene function by methylating histone 3 in the nucleosomes on both promoters and coding regions of genes that are inactivated at developmental stages and in terminally differentiated cell types, but not in the nucleosomes of housekeeping genes (El-Sharnouby et al., 2017). Silencing mechanisms based on PcGII proteins evolved in a common ancestor of holozoans and fungi and were retained in yeast and presumably the direct holozoan progenitors of Metazoa (Kingston and Tamkun, 2014; Steffen and Ringrose, 2014), though they were lost in the unicellular holozoan C. owczarzaki (Grau-Bové et al., 2017). Silenced sequences are converted into stably inactive heterochromatin by PcGI proteins, which are metazoan innovations (Grossniklaus and Paro, 2014; Steffen and Ringrose, 2014). Like transcription hubs, silencing complexes are phase separated into biomolecular condensates (Tatavosian et al., 2019), and their association with the latter appears to be regulated by short-lived noncoding RNAs (eRNAs) transcribed from enhancers (Chan et al., 2018).

To summarize, gene activation mechanisms based on p300/CBP activating and PcGII and SUV-type gene silencing mechanisms were conveyed to metazoans from ancestral unicellular holozoans, while enhancers and PcGI gene heterochromatin sequestration mechanisms accompanied the origination of these organisms. Understanding the emergence of cell differentiation, then, requires focusing on an inferred population of holozoan organisms, which, unlike other members of this group, contained enhancers, which are promiscuous and condition dependent in their gene activating activities, and silencing complexes, which since they depend on enhancers, are also subject to microenvironmental input. These organisms also contained β-catenin, a component integral to metazoan cell adhesion (see section “Inherent Forms in Animal Origination and Evolution”) that is also a key element of externally regulated gene expression. This was the biological “ground state” of metazoan evolution.

It is unknown whether multicellularity preceded or followed the appearance of the described metazoan-specific modes of gene regulation in the ancestors of animals. But for a single organism’s cells to diversify into functionally specialized types, multicellular clusters must already exist. At the level of gene regulation, this requires factors that promote cell-cell adhesion and inhibit cell-cell detachment. The metazoan innovations of the classical cadherin cell attachment molecules and their associated cytoskeletal adaptors have been discussed above. This is one side of the multicellularity dyad. Cell division is commonly followed by cell separation, however, and this function must be suppressed for organismal identity to be determinate.

The grainyhead (Grh) family of TFs has a deep evolutionary history in the opisthokonts, i.e., the fungi and holozoans, including all metazoans, where they are intimately involved in epithelial organ development and epithelial barrier repair after tissue damage (Wang and Samakovlis, 2012). In the fungus Neurospora, the TF GRHL has a DNA-binding specificity close to that of animal grainyhead proteins (Pare et al., 2012). Moreover, Neurospora grhl mutants [allelic with conidial separation-2 (csp-2) mutants] are defective in spore dispersal. A conserved role for Grh-like TFs in the development and remodeling of cell-cell adhesion connections thus appears to have been already in place at the origin of the metazoans.

We can speculate on how Grh may have promoted metazoan multicellularity in light of other components we know to have been present at the metazoan ground state by referring to experiments on extant organisms. In the mouse, the Grh gene Grhl2 is crucial for the expression of the classical E-cadherin as well as claudin4, a junctional protein with homologs in sponges (Werth et al., 2010; Riesgo et al., 2014). E-cadherin promoter activity depends on an intronic enhancer, which binds Grhl2, and the absence of the TF leads to a specific decrease in acetylation and other activating histone modifications (Werth et al., 2010). In addition, Grhl2 has a nearly unique suppressive activity on gene activation by p300, exerted at the histone acetylase domain of the master coregulator. It performs this role in the negative regulation of several metalloproteinases, thereby inhibiting EMT and cell detachment events that lead to tubulogenesis (Pifer et al., 2016). Grainyhead TFs therefore have synergistic, reciprocal effects on promoting cohesion of cell clusters. They may have acted constitutively during early metazoan evolution, only to have become subject to downregulation in later-emerging complex forms in which EMT is a mode of development.

As noted, Grh can reduce expression of certain genes, but it can simultaneously increase expression of others. It shares this property with just a minority of TFs, referred to as “pioneer” regulators, which act at upper levels of enhancer hierarchies to prime target regions for the binding of activating or inhibitory cofactors (Jacobs et al., 2018; Mayran and Drouin, 2018; Gehrke et al., 2019). On this basis, it is possible that the repurposing of Grh was an originating step not only in metazoan multicellularity but also in metazoan developmental gene regulation. Once constituted with components inherited from unicellular ancestors (e.g., p300/CBP transactivating hubs, Grh factors, β-catenin, PcGII proteins) or emerging at the unicellular holozoan-metazoan interface (e.g., enhancers; classical cadherins and their cytoskeletal adaptors, including PcGI proteins), a new category of organism, a “proto-metazoan” could proceed to create a multiplicity of differentiated cell types.

Opisthokont progenitors and ancestral holozoans likely exhibited alternative cell phenotypes and states. While some of these may have been antecedents of differentiated cell types in metazoans (Brunet and King, 2017), there are more than 250 cell types that characterize complex animals such as the vertebrates, and they all could not have arisen from ancestral ones. Furthermore, while GRNs for specific cell types (e.g., neurons, skeletal myoblasts) exhibit a fair degree of conservation across animal phyla (Arendt et al., 2016), regulatory mechanisms for the “same” cell types, and even for homologous genes show great variability across, phyla, species, and even in individuals within a species, often making assignment of similarly acting or appearing cell types as homologs ambiguous (Halfon, 2017). It has been suggested, for example, that neurons of ctenophores and cnidarians were independently evolved (Moroz and Kohn, 2016), although this remains controversial. As mentioned above, the mesoderm in triploblastic phyla might be polyphyletic (Burton, 2008), and within the deuterostomes, skeletogenesis is regulated by different TFs in chordates and echinoderms (Rafiq et al., 2012; Burke, 2018; Taylor and Heyland, 2018).

Cell-type homologies based on descent-with-modification exist because early-radiating members of a phylogenetic lineage can transmit their cell-type regulatory modes to their descendants. However, several phenomena, partly overlapping, contribute to “developmental system drift” that can disconnect cell types from their originating GRNs (True and Haag, 2001; Halfon, 2017). These include, as mentioned above, variable enhancer use across species, the fact that enhancers and promoters do not have entirely disjoint roles and can substitute for one another, the fact that developmental TFs have high concentrations of IDP domains and consequently operate at different cis-acting elements in different contexts, including different stages in the establishment of a cell-type lineage (Niklas et al., 2015; Masoudi et al., 2018), and the fact that a TF can act in entirely opposite fashions depending on the context (Murgan et al., 2015; Lukoseviciute et al., 2018).

Cell differentiation proceeds according to relatively routinized programs in present-day metazoans, but it has likely been less programmed earlier in evolution. The pan-metazoan secreted molecule Wnt, acting at the cell surface, typically initiates gene expression changes during early animal development utilizing a conserved (“canonical”) signal transduction pathway. However, the key transcriptional regulatory component of this pathway, β-catenin, which resides at cadherin-cytoskeletal linkages at the inner surface at the cell membrane until it is mobilized for this function, can also be activated at other pathway nodes by environmental influences, such as mechanical stress (Zhang et al., 2016) and hyperoxia (Popova et al., 2012). The most common co-transcriptional activators of β-catenin, the TCF/LEF family of transcription factors, can also be independently modulated, with numerous identified regulators of its transcription including the Wnt pathway itself (Cadigan and Waterman, 2012). Further, both β-catenin and TCF are highly promiscuous regarding the TFs they partner with, due to the intrinsically disordered (protein) TADs at the N- and C-terminal regions of β-catenin (Zhao and Xue, 2016). Moreover, β-catenin can also function independently of TCF/LEF (Doumpas et al., 2019). In ctenophores, deployment of β-catenin and TCF/LEF appears to entirely circumvent Wnt during early development (Pang et al., 2010).

Single-cell transcriptomic analysis of developmental systems shows a remarkably variable set of regulatory modes. In early Drosophila embryogenesis, for example, where the complex interaction of about 1,000 enhancers defines the activity of a few segmentation genes, some genes express across the field of cells in random bursts, others in broad, progressively refined patterns, and still others in sharp synchrony, to eventually achieve precise and reliable levels of expression (Vera et al., 2016; Bothma et al., 2018). Spatiotemporal precision, as discussed in the following section, is probably imposed at a different level of organization from that on which cell type identity is determined.

A highly plastic but ultimately globally controllable transcriptional regulation apparatus that distinguished among housekeeping, developmental, and specialty genes thus arose coordinately with Metazoa and was in operation in basal metazoans like sponges (though apparently not fully in placozoans). The subsequent appearance of 200 or so cell types was the result of recruitment of this metazoan ground state system of gene control for the enhancement and elaboration of preexisting (and thus inherent) unicellular physiological functions. Whether one or more of the cells in a multicellular entity performed any of these intrinsic functions, e.g., motility, extracellular matrix production, light or nociception, oxygen capture, fuel storage, detoxification, and so forth, may or may not have contributed to the survival of the novel differentiated form. This was decided by natural selection (based on possibilities afforded by the new feature) or construction and occupation of previously unexploited niches.

“Proto-cell types,” providing, for example, evolutionary starting points for the evolution of muscle, neurons, connective tissue, blood, glands, and liver, need not have originated via germline encoding (in this scenario, the germline itself may have arisen by this same “plasticity first” process), or in a reproducible spatiotemporal or even quantitatively regulated fashion. That is, the novel cellular phenotypes could have appeared in a cell mass randomly or reversibly. More programmed modes of appearance could have followed through the self-organizational properties of a critical mass cells (see the following section) and by genetic assimilation, once a germline was in place (Newman, 2011).

The most enigmatic ingredient in this scenario for the transition from the metazoan ground state to complex animals with scores or hundreds of cell types was the appearance of LDTFs and related TFs. These initiate the expression of suites of genes that start an embryo or organ primordium on a progressive developmental trajectory leading, in a hierarchical cascade, to one or more terminally differentiated states (Obier and Bonifer, 2016). The LDTFs include such proteins as MyoD, Nkx3-1, neurogenin, Sox9, and PPAR-γ, which specify (with accessory factors), skeletal muscle, heart muscle, neurons, cartilage, and fat, respectively (reviewed in Newman et al., 2009). Included among these are pioneer TFs, like Grh. Some LDTFs act in a mutually antagonistic fashion ensuring, in concert with silencing effectors like PcGI proteins, that mixed-identity cell types are not produced (Sunadome et al., 2014). Although proteins related to certain LDTFs are present in the genomes of unicellular holozoans or yeast, most are novelties of Metazoa (Grau-Bové et al., 2017). They are therefore orphan genes like many of those involved in the morphogenetic innovations described above in “Inherent Forms in Animal Origination and Evolution.” Although there are speculations on the origins of such genes (Bornberg-Bauer et al., 2015; Neme and Tautz, 2016; Schmitz et al., 2018; Werner et al., 2018), there are no generally accepted accounts for the connection between their appearance and their roles in the recruitment of ancestral cellular functions for cell differentiation.

Finally, it is important to recognize that while cell types in mature metazoans are typically functionally stable, wound repair and regeneration require cell type plasticity. The mechanisms described above are reversible under certain conditions, consistent with this need (Wenger et al., 2016; Gehrke et al., 2019), a phenomenon also manifested in the experimental production of stem cells (Sardina et al., 2018).

We have seen how the innovations in form and function that accompanied the transition from unicellular holozoans to the metazoans were fundamentally expressions of latent, inherent properties of this new category of organism. To reiterate, hollow, multilayered, elongated, appendage-bearing diploblastic and triploblastic forms were direct consequences of the mobilization of mesoscale physical forces and effects in liquid and liquid-crystalline cell aggregates. The existence of such multicellular aggregates and the presence within them of cell types specialized to perform exaggerated versions of ancestral unicellular functions were direct consequences of the new type of enhancer- and silencer-based gene regulatory systems that emerged simultaneously with the metazoans. While inherent to metazoans, morphogenesis and differentiation were not present in their unicellular ancestors. New genes proteins, and regulatory motifs (classical cadherins, Wnt, enhancers), several of which with no known predecessors, were needed to create these inherencies.

The capacity to arrange differentiated cell types in reproducible geometries to produce functionally adequate organs and appendages, that is, cellular pattern formation, is still another category of phenomena with inherent and therefore predictable outcomes. Once the basic ingredients, i.e., pan-metazoan toolkit genes and their products, and the physics of permeable, active, and excitable media to which animal tissues are intrinsically susceptible, were in place, many of the shared features of animal functional anatomy emerged with relative facility.

The simplest way cells in a tissue mass can be induced to differentiate in a spatially dependent fashion is by morphogens, secreted diffusible proteins. Wnt, as noted above, is the most fundamental and ubiquitous of these in the metazoans, but there are more than a dozen widely shared ones, including those of the Hedgehog family, such as Shh and Ihh, Bmps, and Fgfs (reviewed in Newman and Bhat, 2009). Any molecule of this type produced in a cluster of ancestral cells could easily have become nonuniformly distributed by leakage from the surface of the mass, or possibly due to poor nutrient supply in the interior leading to lower expression of its gene. Since such conditions would automatically have been reproduced each time a cell cluster was reconstituted, the morphogen and its physically determined spatial arrangement, combined with its concentration-dependent effects in eliciting changes in gene expression or other aspects of cell phenotype (Loh et al., 2016), may have constituted a primitive developmental program. Subsequent natural selection would have reinforced the accuracy and reliability of such pattern-forming modalities if the resulting organisms proved to be survivable. Eventually the transport of morphogens has come to be not only due to diffusion but to active processes mediated by cells along the distribution routes (Sagner and Briscoe, 2017). In the perspective of the present analysis, however, a key feature of these mechanisms is their origins in inherent properties of collectives of metazoan cells.

In addition to morphogen-based global patterning, the phenomenon of lateral inhibition enforces a choice between alternative cell fates at the local level. This is accomplished by early differentiating cells signaling to cells adjacent to them to take on a fate different from them and is typically mediated by the Notch signal transduction pathway, which is found in all metazoan groups with the exception of Placozoa (reviewed in Newman, 2016b). The activity of this pathway depends on the binding of a member of the Notch family of cell surface receptors, with an integral membrane protein ligand from among the Delta, Serrate (or Jagged), and Lag-2 (DSL) class of proteins (Ehebauer et al., 2006). This interaction results in an intracellular portion of Notch translocating to the nucleus where, as a transcriptional co-regulator of dual-action transcription factors of the CSL (CBF1, Su(H), and Lag-1) class (Lai, 2002), changes them from repressors to activators of key target genes. Since Notch’s effects on cell phenotype depend on which of the dual-action factors are present in a given cell type, the pathway does not determine the cell’s fate, but rather causes it to choose a potential fate different from the one it would have it assumed without the juxtacrine (i.e., direct cell-to-cell) signal.

The Notch pathway, like many of the metazoan morphogenesis and differentiation mediators discussed above, is unique to this group, although some components or portions of them can be recognized in earlier emerging forms. Protein modules of Notch receptors are present in the choanoflagellate holozoan Monosiga brevicollis (King et al., 2008) and a Notch receptor, ligands, and an intracellular mediator are present in the filasterean holozoan Ministeria vibrans (Shalchian-Tabrizi et al., 2008). Studies on CSL homologs in yeast indicate that their role in determining alternative cell states is evolutionary deeper than the holozoans (Prevorovsky et al., 2015).

The physical behaviors of the biogeneric materials discussed earlier in this article – rounding up, phase separation, lumen formation, elongation, wetting, condensation – notwithstanding their dependence on living cells – are analogs of those that result from passive arrangements and rearrangements of subunits in nonliving liquids. Here, however, with the effects of morphogen induction and lateral inhibition in eliciting alternative cell states, we see analogs to chemical transformations. Like chemistry, these are active (i.e., energy consuming) rather than passive processes, in which one kind of entity (a molecule or a cell type) is changed into another. The simplest of these effects is unidirectional, a morphogen choosing the fate of a cell from among the organism’s repertoire of types, or a cell imposing a choice on a second cell. When there are reciprocal influences – positive or negative feedbacks, or synchronization – among the components, the systems can become not merely active, but excitable (i.e., energy storing and releasing; self-organizing (Süel et al., 2006; Sinha and Sridhar, 2015)], and more complex patterns can emerge.