The Origin of Nervous Systems and Animal Body Plans

"The evolutionary origin of metazoan cell types such as neurons and muscles is not known."
Jacob Musser and colleagues, Science, 2021¹

"Deconstructing the evolutionary process by which stem group body plans were stepwise formulated will require us to traverse the conceptual pathway to dGRN elegance, beginning where no modern dGRN provides a model."
Eric Davidson, Developmental Biology, 2011²

Summary

An animal is built twice.  Once in history, when the first nerve cells and the first body plans appeared, and once in every embryo, when a single cell becomes a worm, a fly or a sea urchin.  This article asks how well the first building is explained, using what is known about the second.  It covers two steps:  the origin of neurons, and the origin of the developmental programmes that lay out the body plans of the animal phyla.  When the phyla first show up in the rocks is a separate question, answered in The Cambrian Explosion.  How single cells first stuck together is answered in The Origin of Multicellularity.  For each step this article separates what has been observed, what is inferred, and what is assumed.

• What the design side claims. Stephen Meyer's Darwin's Doubt (2013) makes three arguments from development. Mutations that act early in an embryo, the only kind that could change a body plan, always kill or cripple it. The gene networks that lay out the body, which Eric Davidson mapped in the sea urchin, "cannot vary without causing catastrophic effects," and "there are no examples of these deeply entrenched, functionally critical circuits varying at all." And part of the information that shapes an embryo is not in its DNA, so DNA mutations are "simply the wrong tool for the job." Paul Nelson adds that early-acting changes have "never" been tolerated "in any animals that developmental biologists have studied." Young-earth writers say less about this topic than about most; Don Batten of Creation Ministries International argues that Hox mutations make only "monstrosities."
• Neurons: arose once and were lost twice, or arose twice. Evidence from blocks of genes that have stayed together on chromosomes, published in 2023, places comb jellies, not sponges, as the first branch of the animal tree. Comb jellies have nerve nets, but they lack many genes that build neurons in other animals and appear not to use serotonin, acetylcholine or dopamine as signals. So either neurons arose twice, or the first animal had them and sponges and placozoans lost them. Neither has been shown. The parts are older than the cell: a sponge carries a nearly complete set of the proteins that build the receiving side of a synapse, and a placozoan coordinates its feeding with peptide-releasing cells and no synapses. Nobody has shown a stepwise route from such parts to a working nerve net, in a living animal or a fossil, and no design writer found for this article has addressed the question.
• The toolkit is shared, and that was a surprise. The Hox genes that lay out the head-to-tail axis are missing from sponges and comb jellies and present in cnidarians and bilaterians. A squid's eye-starting gene grows extra fly eyes in a fly. Sean Carroll called the shared toolkit "entirely unanticipated." Its distribution fits common ancestry. Design can accommodate it. Neither explains where the programmes that use the toolkit came from.
• Meyer quotes Davidson accurately, and stops short. Davidson does say that interrupting a modern network has catastrophic effects. In the next sentences he calls modern networks "evolutionarily terminal" and says the first networks "must have differed in fundamental respects" and "were likely hierarchically shallow," so that selection could act on them. Davidson's own data are his evidence for that account, not against it. But Meyer's later reply is fair: no such shallow network has ever been observed.
• Networks and early development do change. Davidson's own laboratory found that between a sea urchin and a sea star, "outside of the kernel not one of the genes" compared "has all of the same regulatory inputs." Two sea urchins of one genus, Heliocidaris, differ in how they set the body's back-to-belly axis, in the embryo's main signalling centre, and in early cell fates. Both are alive. Meyer's "no examples" and Nelson's "never" are wrong as stated. The innermost "kernels" do stay the same, as Meyer says.
• No experiment has built a new body plan. The famous Heidelberg screen that Meyer and Nelson cite was a search "for embryonic lethal mutants," so every mutant it kept died by design. Mutations in Hox genes turn one body segment or limb into the copy of another: a four-winged fly, a crustacean with walking legs where its abdomen should be. They rearrange an existing plan. No laboratory has produced a new body plan or a new phylum-level structure, and none of the papers read for this article claims to.
• The cost of rewiring one link has been priced; the number of links has not. The model Meyer leans on for "implausibly long waiting times," by Rick Durrett and Deena Schmidt, finds that in an insect-sized population "a few million years is sufficient" to swap one binding site for another. Nobody has counted how many network links separate two phyla, so the cost of a new body plan is unknown on both sides.
• Information outside DNA is real, and does not settle the question. Planarian worms can carry a hidden two-headed plan through repeated cutting, held not in their genes but in electrical patterns across their cells. Carp nuclei in goldfish eggs grow goldfish numbers of vertebrae. Meyer is right that DNA is not the whole story. But mouse stem cells, with no egg at all, self-organize into embryo models with a head fold, neural tube, somites and a beating heart-like tube. That DNA is not the whole story does not make DNA mutation "the wrong tool."

The verdict.  Five questions get different answers.  Where neurons came from is unfinished on both sides:  the tree is now fairly firm, but whether neurons arose once or twice, and how, is not known.  That animals share one developmental toolkit is best explained by common ancestry, and this is a point against the creationist reading.  The claim that early development and its core networks never vary is wrong as stated:  they vary between living sister species without harm, though the innermost kernels do not.  The claim that no observed process has built a new body plan is right:  every observed change moves, copies, removes or rewires parts of a plan that already exists, and the shallow ancestral networks the mainstream account needs are inferred, not seen.  Information outside DNA is real, but it does not make DNA mutation the wrong tool.

Part 1: The claims, and what each side says

The mainstream claim.  All animals descend from one ancestor.  Neurons arose from earlier cells that signalled by secretion.  The body plans of the phyla arose when the networks of genes that run an embryo were rewired.  Most of the genes were already present.  Charles Marshall, a palaeontologist at Berkeley, put the last point this way in his 2013 review of Meyer's book:  "new phyla were not made by new genes but largely emerged through the rewiring of the gene regulatory networks (GRNs) of already existing genes."³  Eric Davidson of Caltech, who mapped more of these networks than anyone, held that "gross morphological novelty required dramatic alterations in dGRN architecture, always involving multiple regulatory genes, and typically affecting the deployment of whole network subcircuits."²  Here "dGRN" means developmental gene regulatory network.

The intelligent-design case.  Stephen Meyer's Darwin's Doubt (2013, paperback 2014) is the fullest statement.  Its chapter 13 argues from experiments on embryos.  Mutations acting late in development change little.  Mutations acting early could change a body plan, but "developmental biology has shown that mutations affecting body-plan formation expressed early in development inevitably damage the organism."⁴  Meyer sums it up:  "the kind of mutations we need for major evolutionary change we don’t get; the kind we get we don’t need."⁴  His colleague Paul Nelson, a philosopher of biology, states the key premise.  Early-acting mutations "of global effect on animal development" are "those least likely to be tolerated by the embryo and, in fact, never have been tolerated in any animals that developmental biologists have studied."⁴  Nelson repeated it in 2019:  "We have no experimental evidence, from any well-studied model system of evo-devo, that animal embryos tolerate early-acting mutations to body plan-affecting regulatory elements."  He makes one exception:  "losses of structures: wings, eyes, etc."⁵

Meyer's second argument is about the networks.  Drawing on Davidson's sea urchin work, he writes that the networks at the centre of the hierarchy "cannot vary without causing catastrophic effects to the organism," and "there are no examples of these deeply entrenched, functionally critical circuits varying at all."⁴  So "if neither early-acting regulatory genes nor dGRNs can be altered by mutation without destroying existing developmental programs," selection has nothing to work on and "the evolution of animal form will, at that point, terminate."⁴

His third argument, in chapter 14, is that part of the information that shapes an embryo is not in its DNA.  It lies in the layout of the egg's membranes, cytoskeleton and ion channels.  So "DNA sequences can mutate indefinitely and still not produce a new body plan," and "Genetic mutations are simply the wrong tool for the job at hand."⁴  Chapter 16 turns to Hox genes.  Meyer argues that in flies they switch on only after "roughly 6,000 cells have already formed" and the main axes are set, "So Hox genes don’t determine body-plan formation," and that "Mutations in Hox genes can only generate new genetic information in DNA. They do not, and cannot, generate epi genetic information."⁴

Earlier statements of the same case are blunter.  In 2003 Nelson and Jonathan Wells wrote that homeotic genes act as "binary switches between alternate developmental fates, with the information for the resulting structures residing elsewhere," and that surviving homeotic mutants are "life’s proverbial born losers."⁶  A 2010 reply by Meyer, Nelson, Richard Sternberg and Wells said that "saturation mutagenesis" had "been used to produce every possible developmental mutation in fruit flies," and that "none of the observed developmental mutations benefit the organism."⁷

Young-earth writers.  They have written much less on this.  Don Batten of Creation Ministries International wrote in 2006 that "Changes to homeotic genes cause monstrosities (two heads, a leg where an eye should be, etc.); they do not change an amphibian into a reptile, for example."⁸  David DeWitt of Liberty University, writing for Answers in Genesis, takes a softer line, and it differs from Meyer's:  "creationists do not claim that mutations cannot alter the body plan. Rather, they claim that mutations do not lead to an increase in information."⁹  So the two sides of the design camp make different claims.  Meyer says the needed early changes are not viable.  DeWitt grants that body plans can change and puts the problem in the origin of new information.  This article tests each claim as its author makes it.

What neither side has written about.  A search of Answers in Genesis, Answers Research Journal, Creation Ministries International and Journal of Creation, the Institute for Creation Research, the Creation Research Society Quarterly and Evolution News found no young-earth or intelligent-design article on the origin of neurons, the comb jelly genome or nerve net, or Davidson's "kernels" beyond Meyer's chapter.¹⁰  Part 2 therefore weighs the neuron question against the mainstream literature alone.

Part 2: Where neurons came from

What is observed.  Five groups of animals sit at the base of the animal tree.  They are comb jellies (ctenophores), sponges, placozoans (flat animals about a millimetre wide), cnidarians (jellyfish, corals and sea anemones) and bilaterians (animals with a left and right side, from worms to people).  Comb jellies, cnidarians and bilaterians have neurons.  Sponges and placozoans do not.

The comb jelly nervous system is unlike the others.  In 2014 Leonid Moroz and colleagues sequenced the genome of the comb jelly Pleurobrachia and compared it with transcriptomes of ten other comb jellies.  Many genes that set up neurons in other animals "are absent in the ctenophores we sampled."¹¹  Their chemical, staining and drug tests "strongly suggest ctenophores do not use serotonin, acetylcholine, dopamine, noradrenaline, adrenaline, octopamine, histamine or glycine" as signals between cells.  Glutamate is their candidate transmitter.¹¹  In 2023 Pawel Burkhardt's group rebuilt the nerve net of a young comb jelly in three dimensions from electron micrographs.  They "found that neurons in the subepithelial nerve net have a continuous plasma membrane forming a syncytium."¹²  The cells of that net are fused, not joined by synapses as in a jellyfish or a person.

Sponges have no neurons, but they carry many of the parts of a synapse.  Onur Sakarya and colleagues searched the genome of the sponge Amphimedon for 36 families of genes that build the receiving side of a vertebrate synapse.  They found "a nearly complete set of post-synaptic protein homologs."  The main family missing was the receptors themselves:  "the entire ionotropic glutamate receptor family."¹³  In 2021 Jacob Musser and colleagues read the active genes of single cells across a whole freshwater sponge and found 18 cell types.  Among them are "secretory neuroid cells" that wrap the hair-like projections of the sponge's feeding cells.  They read this as a signalling system built from "conserved modules that became incorporated into the pre- and postsynapse in the nervous systems of other animals."¹

Placozoans have no neurons either, and they still coordinate their feeding.  Secretory cells make up "only 4%" of the lower skin of Trichoplax.  Adriano Senatore and colleagues found that a peptide released by these cells makes other cells release it too, which "can evoke a wave of peptide secretion across the entire animal to globally arrest ciliary beating".¹⁴  In 2023 Sebastián Najle and colleagues mapped the cells of four placozoan species.  They "uncovered fourteen peptidergic cell types expressing neuronal-associated components like the pre-synaptic scaffold that derive from progenitor cells with neurogenesis signatures."  The receiving side of the synapse was missing.¹⁵

What is inferred:  the tree.  Whether sponges or comb jellies branched off first was disputed for about fifteen years.  Some genome studies placed comb jellies first, including Nathan Whelan's in 2017.¹⁶  Others, including Davide Pisani's in 2015, Paul Simion's in 2017 and Anthony Redmond and Aoife McLysaght's in 2021, placed sponges first.¹⁷¹⁸¹⁹  In 2023 Darrin Schultz and colleagues settled it by a different kind of evidence:  blocks of genes that have stayed together on the same chromosomes since before animals existed.  Their result puts comb jellies first.²⁰  How that test works, and why sequence data could not settle the question, is told in Phylogenetic Congruence.

On that tree, Schultz and colleagues say, there are two ways neurons could have arisen.  "Either complex neural systems arose more than once," or "neuronal cell types were present in the metazoan ancestor but were lost in the sponge lineage."²⁰  If placozoans sit inside the group with cnidarians and bilaterians, as most trees put them, then placozoans "have lost an ancestral nervous system" as well.²⁰  Genome studies since 2008 have placed them there.²¹

The same genome, read two ways.  Moroz's group reads the missing genes as a sign that comb jellies built their nervous system on their own:  "Extensive parallel evolution of neural organization in ctenophores is the most evident."¹¹  Joseph Ryan's group, describing the genome of another comb jelly, Mnemiopsis, a year earlier, read the same kind of data the other way:  "The set of neural genes in Mnemiopsis is similar to that of sponges."²²  To them the gaps fit losses.  Part of the trouble is definition.  Moroz and Andrea Kohn concluded in 2016 that "we cannot assign the ‘pan-neuronal’ or ‘pan-synaptic’ tag to any single gene across all metazoan lineages."²³  If no gene marks every neuron, the question of how many times neurons arose has no fixed answer.

Counting the events.  The two stories can be scored by counting gains and losses on the tree.²⁴

The count decides nothing on its own.  If gaining a cell type as complex as a neuron is much harder than losing one, the single origin wins.  If the two are about equal, two origins win.  That ratio is an assumption, not a measurement.

What is assumed.  Two assumptions carry the mainstream inference.  The first is that shared parts mean a shared ancestor:  that the sponge's synapse proteins and the placozoan's secretory cells were inherited from the same ancestral cells that became neurons elsewhere.  The second is that living sponges and placozoans tell us what the ancestors were like.  Schultz and colleagues warn against the second:  the nervous systems of living comb jellies, bilaterians and cnidarians "do not represent the stem ancestors of these groups."²⁰  And a loss cannot be watched.  Joseph Ryan and Marta Chiodin, who favour loss, wrote:  "Directly proving that sponges and placozoans lost neural cell types will be extremely difficult if not impossible."²⁵

The first nerve net.  No one knows what it looked like.  Frédérique Varoqueaux and colleagues, in their 2018 study of placozoan peptides, put it plainly:  "how the first neurons evolved from more primitive cells for communication during the rise of animals are not yet understood."²⁶  Najle's group, whose placozoan data give the best parts list, adds that "How ctenophore neurons fit in this scenario remains a major question."¹⁵  The proposals on offer start from secretory cells that signal by releasing peptides, and add synapses later.²⁶  That order fits the placozoan and sponge data.  No one has shown the steps.

Where this leaves the question.  The observations are solid:  the parts of a neuron's signalling system exist in animals with no neurons, and the comb jelly nervous system is built differently from all others.  The tree is now fairly firm.  How many times neurons arose, and by what steps, is inferred from parts lists and is not settled among the mainstream authors themselves.  No young-earth or intelligent-design writer found for this article has published an account of the same data (Part 1).  This question is unfinished on both sides.

Part 3: The shared toolkit

What is observed.  Animals of very different shapes build their bodies with many of the same genes.  The best known are the Hox genes, which give each region along the head-to-tail axis its identity.  Edward Lewis found that the bithorax complex of the fruit fly "contains a minimum of eight genes that seem to code for substances controlling levels of thoracic and abdominal development."²⁷  The distribution of Hox genes across the base of the tree is uneven.  The sponge Amphimedon "lacks Hox genes,"²⁸ and so do the comb jellies sequenced so far.¹¹  The sea anemone Nematostella, a cnidarian, has them and uses them:  in a 2018 study that knocked out the genes with CRISPR and switched them off with RNA, "Four homeobox-containing genes constitute a molecular network that coordinately controls the morphogenesis of radial endodermal segments and the patterning of tentacles."²⁹  Flies, worms and vertebrates have clusters of them.

Much of the kit appeared with the first animals.  Jordi Paps and Peter Holland compared the predicted proteins of 62 genomes and estimated that the first animal had "a remarkable 1189 Novel HG," groups of genes found in no single-celled relative, "a threefold increase compared to novelties in the previous nodes."³⁰  How the first animals came to exist at all, as cells that stick together and divide their labour, is the subject of The Origin of Multicellularity.  This article starts after that step, with animals that already have several cell types.

The same holds for genes that start an eye.  The fly gene eyeless is the counterpart of the mouse and human gene Pax-6.  Switched on in the wrong place, it grows eye structures "on the wings, the legs, and on the antennae."³¹  A squid's Pax-6, put into a fly, does the same:  "squid Pax-6 is able to induce the formation of ectopic eyes in Drosophila."³²  The eyes that grow are fly eyes.  How much this says about whether eyes arose once or many times is argued in Convergent Evolution and Design Reuse.

It was not expected.  Sean Carroll of the University of Wisconsin, one of the founders of evolutionary developmental biology, says so plainly.  Before the toolkit was found, "the prevailing view of the architects and adherents of Modern Synthesis was that the process of random mutation and selection would so alter DNA and protein sequences that only closely related species would bear homologous genes."  Then:  "The fact that such different forms of animals are shaped by very similar sets of tool kit proteins was entirely unanticipated."³³  A creationist can fairly point out that the theory did not predict this.  Carroll's own reading, on the same page, is that it is "entirely new and profound evidence" for "the descent of all forms from one (or a few) common ancestor."³³

What the pattern supports.  The toolkit is not spread at random.  The groups that branch off first, comb jellies and sponges, have no Hox genes.  Cnidarians have a few and use them to pattern their bodies.  Bilaterians have clusters of them and use them along the whole head-to-tail axis.  That nested pattern is what descent from a common ancestor predicts, and it is the pattern a family tree is built from.  A designer could reuse a toolkit across many animals.  But reuse alone does not predict that the sponge and the comb jelly would lack the Hox genes, or that the animals with the simplest axes would have the fewest.  On this point the evidence favours common ancestry.

Where Meyer is right.  Meyer's chapter 16 argues that Hox genes do not themselves build a body plan.  In flies they switch on after the main axes are set.⁴  Nelson and Wells call them "binary switches between alternate developmental fates, with the information for the resulting structures residing elsewhere."⁶  The mainstream literature says much the same thing in its own terms.  Isabelle Peter and Eric Davidson describe the classic four-winged fly this way:  "In the absence of Ubx function in T3, what should be the haltere imaginal disc produces a wing, hence Ed Lewis’ famous 4-winged fly."  Ubx works by switching off the wing-building network in that segment.³⁴  The wing network was already there.  Douglas Erwin and colleagues make the general point:  the differences between a sponge and an arthropod "cannot be due to the absence of these protein-coding gene families," but lie in how the genes are deployed.³⁵

So both sides agree that the toolkit is necessary and not sufficient.  The shared genes show that animals draw on a common stock.  They do not explain where the programmes that use the stock came from.  That question moves to the gene networks, which is where Meyer's main argument lies.

Part 4: Davidson's networks, in Davidson's own words

What a network is.  Every cell in an embryo carries the same genes.  Cells become different because different genes are switched on.  The switches are proteins called transcription factors, which bind to short stretches of DNA beside other genes and turn them on or off.  A developmental gene regulatory network is the wiring diagram of who switches on whom.  The sea urchin's is the best known:  it "contains over 200 experimentally verified regulatory interactions" running from the unfertilised egg to the swimming larva.³⁶  Within such networks Davidson and Erwin named small "kernels."  Peter and Davidson define them:  "These Kernels consist of a few regulatory genes linked by recursive positive regulatory interactions, and they are usually used upstream in specification GRNs and are conserved at large evolutionary distances."³⁴

What Meyer quotes, and what follows it.  Meyer's strongest quotation is from Davidson's 2011 essay, and it is accurate:  "There is always an observable consequence if a dGRN subcircuit is interrupted. Since these consequences are always catastrophically bad, flexibility is minimal, and since the subcircuits are all interconnected, the whole network partakes of the quality that there is only one way for things to work. And indeed the embryos of each species develop in only one way."⁴  So is his quotation that at the top of a network the output "is very impervious to change, except for catastrophic loss of the body part or loss of viability altogether."²  And so is his quotation of Davidson's verdict on classical neo-Darwinism, whose assumptions are "basically counterfactual."²

The sentences after the first quotation change its meaning for the argument.  Davidson goes on:  "Thus we can think of a crown group dGRN as an evolutionarily terminal, finely divided, extremely elegant control system that allows continuing alteration, variation, and evolutionary experimentation only after the body plan per se has formed".  A crown group is a living group and its extinct relatives that share its body plan.  Then:  "The basic control features of the initial dGRNs of the Precambrian and early Cambrian must have differed in fundamental respects from those now being unraveled in our laboratories. The earliest ones were likely hierarchically shallow rather than deep, so that in the beginning adaptive selection could operate on a larger portion of their linkages."²

So Davidson's claim is about today's networks.  He says they resist change because they are deep and finely divided, and that the first networks were not.  His own data are, for him, the evidence that the networks have been built up over time.  Meyer's chapter 13 does not tell the reader this.  It says that Davidson's work "has also shown that such alterations inevitably have catastrophic consequences," and that "Davidson makes clear that no one really knows" how body plans evolved.⁴  Davidson does reject classical neo-Darwinism as a model for body plans.  He does not reject evolution.  He proposes a different evolutionary mechanism, as Charles Marshall's review of Meyer pointed out.³

Meyer's reply, and why it is fair.  In the 2014 epilogue Meyer answers Marshall, and here he quotes Davidson's own phrase.  Speculation about early flexible networks, he writes, deals with a region "where no modern dGRN provides a model."  He adds:  "It is telling that although many evolutionary theorists (like Marshall) have speculated about early labile dGRNs, no one has ever described such a network in any functional detail".  And:  "No developing animal that biologists have observed exhibits the kind of labile developmental gene regulatory network required by the evolution of new body plans."⁴  His underlying point is about method.  Present observation should judge theories of the past, and the shallow early network is posited because the theory needs it, not because anyone has seen one.

That point is correct as far as it goes.  Davidson himself wrote that the path to the modern networks begins "where no modern dGRN provides a model."²  The shallow, evolvable ancestral network is an inference.  It has not been observed in any living animal, and soft network wiring does not fossilise.  What the mainstream can say in reply is that today's networks are not all equally rigid:  outside the kernels they change a great deal between related species, and some quite early steps change too.  That is the subject of Part 5.  It narrows Meyer's claim.  It does not supply the missing ancestral network.

Part 5: Do the networks and early embryos change?

Meyer and Nelson make two factual claims that can be checked against living animals.  The central networks "cannot vary" and there are "no examples" of them varying.  And early-acting changes of global effect have "never" been tolerated.  The comparisons below are between living species.  The time spans are the authors' own estimates on the standard timescale.  On a young-earth timescale they would be shorter, which makes the observed changes faster, not slower.

The kernels are conserved.  Meyer is right about that.  Veronica Hinman and Davidson compared the network that sets up the gut and inner tissues in a sea urchin and a sea star.  A small feedback loop at its core "is found in almost identical form in the starfish," and they wrote that "a detailed element of GRN architecture has been retained since the Cambrian Period in both echinoderm lineages."³⁷  In 2007 they called that core a kernel and suggested why it does not change:  its "recursive wiring, such that each gene has multiple targets within the same kernel, may explain its evolutionary immutability."³⁸

Everything around the kernels changed.  Meyer is wrong about that.  The same 2007 paper reports:  "We show here that the GRN structure upstream and downstream of the conserved network kernel has, by contrast, diverged extensively."  And:  "The surprising fact is that outside of the kernel not one of the genes included in Fig. 5 has all of the same regulatory inputs in the two GRNs."³⁸  Upstream means earlier in the cascade, the part that switches the kernel on.  Their summary:  "The bottom line is the surprising GRN plasticity revealed by the comparison."³⁸

Eric Erkenbrack and Davidson then compared the skeleton-building network of the standard laboratory sea urchin with that of a pencil urchin, a member of the other living subclass.  "Remarkably, most of the canonical design features of the Sp skeletogenic GRN that we examined are either missing or operate differently in Et".  Their conclusion:  "the well-known and elegantly configured regulatory circuitry that underlies skeletogenic specification in modern sea urchins is largely a novel evolutionary invention."³⁹  Both kinds of urchin still build a larval skeleton.  The network changed and the body part did not.  John True and Eric Haag gave this pattern a name in 2001, "developmental system drift," and found it "apparently ubiquitous."⁴⁰

Early development changes between sister species.  The sharpest case is a pair of sea urchins in the genus Heliocidaris.  H. tuberculata has a small egg and a larva that feeds.  H. erythrogramma has a large yolky egg and a larva that never feeds.  Jennifer Israel and colleagues in Gregory Wray's laboratory write that "key patterning mechanisms such as dorsoventral axis specification, the establishment of the primary signaling center of the embryo, and early cell fate specification differ between the two species," and that "some of these modifications involve developmental mechanisms that were previously conserved for 10s–100s of millions of years before changing dramatically and rapidly."  They date the split at 5 million years.³⁶  The same switch to a non-feeding larva "has independently evolved at least 14 times" among sea urchins.³⁶  Both species are healthy animals in the wild.

The pattern is general.  Alex Kalinka and Pavel Tomancak reviewed the molecular, network and life-history evidence in 2012 and concluded that "early animal embryos are not highly conserved."⁴¹

The screen that found only lethals was looking for lethals.  Meyer says the Heidelberg experiments "have shown definitively that early-acting body-plan mutations invariably generate embryonic lethals."⁴  Christiane Nüsslein-Volhard and Eric Wieschaus describe their own work this way:  "In systematic searches for embryonic lethal mutants of Drosophila melanogaster we have identified 15 loci which when mutated alter the segmental pattern of the larva."⁴²  The screen was built to catch mutations that kill embryos.  That every mutant it kept died is how it was designed, not a discovery about what all early mutations do.  Most of its mutations also knocked a gene out completely, in both copies.  Natural differences between species, like those in Heliocidaris, are mostly changes in when and where genes are switched on.  The 2010 claim by Meyer, Nelson, Sternberg and Wells that the screens produced "every possible developmental mutation" overstates what was done.⁷  The 1980 paper reports 15 genes of one class, those that "alter the segmental pattern of the larva," and says they "probably represent the majority of such genes," not of all developmental genes.⁴²

What Nelson can fairly reply.  The Heliocidaris change was not watched as it happened.  It is inferred from comparing two living species.  Nelson has made this point about developmental system drift in general.  He wanted to ask a proponent whether she had "any experimental data in animals showing DSD arising via mutations," and adds:  "I didn't, because I already knew the answer."⁴³  He is right that no one has produced such a change by mutation in the laboratory.  But his premise as Meyer states it is broader:  that early-acting changes of global effect have never been tolerated in any animal studied.  Two living sea urchins that set up their body axes differently show that such changes are tolerated.  And neither Heliocidaris nor the pencil urchin has a new body plan.  The adults are still sea urchins.

Across phyla, the middle of development differs most.  Michal Levin and colleagues compared gene activity through development in ten phyla.  Their data fit "an inverse hourglass model for metazoan body plans" in which the early and late stages are more alike between phyla and the middle stage differs most.  They proposed that "a phylum may be defined as a collection of species whose gene expression at the mid-developmental transition is both highly conserved among them, yet divergent relative to other species."⁴⁴  So the developmental stage that defines a phylum is the one that differs between phyla.  This describes where the differences are.  It does not show how a lineage crossed from one such programme to another.

Where this leaves the claims.  Meyer's "no examples ... varying at all" is wrong:  network wiring upstream and downstream of the kernels, and early axis setting, differ between living relatives.  Meyer's narrower point holds:  the kernels themselves have not been seen to change.  And no observed change in a network has yet produced a new body plan.

Part 6: What mutating the toolkit has done

The laboratory record on Hox genes is long, and it says the same thing in every animal tested.  A mutation in a Hox gene changes which kind of structure a body region makes.  It does not make a new kind of structure.

The four-winged fly.  Meyer describes "A mutation in the regulatory Ultrabithorax gene" that gives a fly a second pair of wings, and says the result "cannot fly because it lacks, among other things, a musculature to support the use of its new wings."⁴  The classic four-winged fly needs three mutations, all in switching regions of the same gene.  Jorge Fernandes and colleagues, working with Edward Lewis, describe flies "homozygous for three mutant alleles (anterobithorax, bithorax3, postbithorax) of the Ultrabithorax gene."  "The adults have two pairs of wings, but the homeotically transformed T3 (HT3) has only rudimentary indirect flight muscles."⁴⁵  Meyer's account of the muscles is close to the source.  Jonathan Wells's 2017 version, that "the mutant wings do not have any flight muscles," overstates it:  the muscles are rudimentary, not absent.⁴⁶  Either way, the extra wings are copies of the wings the fly already makes, grown in place of the small balancing organs called halteres.

Eyes on legs.  Switching on eyeless in the wrong tissue grows eye structures on wings, legs and antennae, and the flies live.³¹  These are fly eyes in the wrong place.  Nelson's 2019 rule allows one exception, losses of structures.⁵  Extra wings and misplaced eyes are not losses, and the flies survive.  But they are not new structures either.  They are duplicates of what the fly already builds.

A crustacean rebuilt limb by limb.  In 2016 Arnaud Martin and colleagues in Nipam Patel's laboratory knocked out six Hox genes one at a time in the amphipod Parhyale, a small crustacean with many kinds of limbs.  The results were striking:  "knockout of Abd-B resulting in an animal with thoracic type legs along what would have been an abdomen, and abd-A disruption generating a simplified body plan characterized by a loss of specialization in both abdominal and thoracic appendages."⁴⁷  Each change swapped one of the animal's existing limb types for another.  The authors argue that shifts in where Hox genes are switched on helped produce the differences between crustacean species, and then add a caution:  "there is no doubt that this will only be part of the explanation."⁴⁷  That fits the comparative work.  Michalis Averof and Patel found in 1997 that differences in where two Hox genes are switched on "correlate well" with which thoracic limbs became feeding limbs in different crustaceans.⁴⁸

The leg-suppression experiment.  Matthew Ronshaugen and William McGinnis compared the Ubx protein of a fly with that of a brine shrimp.  The fly's version suppresses legs;  the shrimp's does not.  They reported that "naturally selected alteration of the Ubx protein is linked to the evolutionary transition to hexapod limb pattern."⁴⁹  "Linked" is the authors' word, and it is weaker than "caused."  Wells points out, fairly, that no one turned a shrimp into an insect:  "McGinnis and his colleagues did not reduce the number of legs in a shrimp."⁴⁶  David DeWitt of Answers in Genesis grants the general point:  "reducing the number of legs may alter the body plan, but it does not explain the origin of legs in the first place."⁹

What the comparisons add.  Benjamin Prud'homme, Nicolas Gompel and Sean Carroll trace the two-winged body of true flies to a change in what Ubx controls.  The hindwing was reduced "under the control of the Ubx protein and did not result from a shift of the expression of Ubx," and "the two-winged dipteran body plan evolved by changes in the regulatory interactions between Ubx and downstream wing-patterning genes."⁵⁰  This is inferred from living species, not reproduced in the laboratory.  It is also a reduction, the kind of change Nelson already allows.

Where this leaves the claims.  Batten's "monstrosities" and Nelson and Wells's "born losers" describe laboratory mutants that knock out a gene or switch it on everywhere.  They do not describe the regulatory changes that comparative studies infer, which happened in healthy lineages.  But the design side's main point survives every case above.  Each experiment moves, copies, removes or simplifies a structure the animal already had.  None of the papers read for this article reports a new kind of organ, a new cell type or a new body plan made by mutation, and none claims to.  The creationist claim that no experiment has produced a new body plan is right.

Part 7: The price of rewiring a network

Meyer's claim.  In his epilogue Meyer argues that the rewiring Marshall proposes "would require multiple coordinated mutations in the cis -regulatory regions adjacent to various genes in these regulatory networks."  Cis-regulatory regions are the stretches of DNA beside a gene where the switch proteins bind.  He adds that rewiring would also need coordinated changes in the genes for the switch proteins themselves, and that making new genes by "multiple coordinated mutations" "would require implausibly long waiting times of the kind discussed in Chapter 12."⁴  In chapter 12 he names the work he relies on, including that of "Durrett and Schmidt (inadvertently)."⁴  Whether new protein-coding genes can arise in time is argued in other articles in this section.  The rewiring step, changing where the switch proteins bind, can be priced here, and the paper Meyer leans on has priced it.

What Durrett and Schmidt computed.  Rick Durrett and Deena Schmidt, mathematicians at Cornell, asked how long a population must wait for two mutations, one that disables an existing binding site and one that creates a new one nearby, when the first is harmful or neutral on its own.  That is a single rewired link.  Their abstract gives both answers:  "Consistent with recent experimental observations for Drosophila, we find that a few million years is sufficient, but for humans with a much smaller effective population size, this type of change would take >100 million years."⁵¹  A single new binding site, made by one mutation from a near-match that is already present, is much quicker:  "in humans, a new transcription factor binding site can be created by a single mutation in an average of 60,000 years."⁵¹  Their conclusion for flies:  the two-step switch takes "several million years, even when we make the conservative assumption that the second mutation is neutral," which "is consistent with the observation of rapid turnover of transcription factor binding sites in Drosophila."⁵¹

What is observed.  Binding sites do turn over quickly.  Dominic Schmidt and colleagues, comparing where two liver switch proteins bind in five vertebrates, note that "the majority of transcription factor (TF) binding events change rapidly between closely related species."⁵²  Diego Villar and colleagues mapped enhancers, the switching regions that act at a distance, in the livers of 20 mammals and found that "Enhancers are rarely conserved across these mammals."⁵³  In sea urchins, early network links changed between two species of one genus in about 5 million years on the standard dating (Part 5).

What the numbers show, and what they do not.  Early marine animals had large populations and short generations, closer to flies than to humans.  That is our assumption, and it is the usual one.  On it, the model Meyer cites prices one rewired link at a few million years or less.  "Implausibly long" does not hold for a single link.

The total cost of a new body plan is another matter.  It depends on how many links must change, and how many of them must change together before any helps.  Nobody has counted how many links separate the networks of two phyla.  For scale only, and labelled as our arithmetic:  a hundred links rewired one after another at a few million years each would take a few hundred million years.  The same hundred changing in parallel, in different genes, would take only as long as the slowest.⁵⁴  Real rewiring would fall between those limits.  So the price of a new body plan turns on numbers nobody has measured, and neither side can yet put a figure on it.  That is a gap in the mainstream account, because the count is what would make it testable.  It is equally a gap in Meyer's, because "implausibly long" needs the same count.

Part 8: Information outside DNA

Meyer's claim, again.  Chapter 14 of Darwin's Doubt argues that an embryo's shape depends on information held in the layout of the cell, not in its DNA.  Meyer gives examples:  the arrangement of the cytoskeleton, patterns in the cell membrane, and the electric fields set up by ion channels.  Of the channels he writes that although they are made of proteins "that may be encoded by DNA ... their pattern in the membrane is not."⁴  From this he concludes that DNA mutations are "the wrong tool for the job," and that changes to the cell structures themselves "are overwhelmingly likely to have harmful or catastrophic consequences."⁴⁴

Evidence that supports him.  Information outside DNA sequence does shape bodies.  In 2005 Yong-Hua Sun and colleagues put nuclei from common carp into goldfish eggs whose own nuclei had been removed.  The seven fish that grew had carp genomes, yet "The somite development process and somite number of nuclear transplants were consistent with the recipient species, goldfish, rather than the nuclear donor species, common carp," and their vertebra counts "belonged to the range of goldfish."⁵⁵  The egg, not the nucleus, set the number of body segments.

Michael Levin's laboratory at Tufts has gone further with planarian flatworms, which regrow a whole body from a fragment.  A brief disturbance of the electrical voltage across the cells of a cut fragment makes some fragments regrow two heads.  The fragments that regrow normally are changed too:  cut again, in plain water, they produce two-headed worms at the same rate.  Durant and colleagues found no difference in their tissues or in their key polarity genes.  "Instead, the altered regenerative bodyplan is stored in seemingly normal planaria via global patterns of cellular resting potential."  Their conclusion is close to Meyer's:  "bioelectric properties can stably override genome-default target morphology."⁵⁶  In another planarian, blocking the channels between cells after the head was cut off made the new head grow in the shape of other species' heads, "from the same genome."⁵⁷  Meyer is right that DNA sequence alone does not fix an animal's shape.

Evidence against the strong form.  If the body plan were written mainly in the egg's own architecture, it could not be built without an egg.  It can be, at least for a while.  In 2022 Gianluca Amadei and colleagues in Magdalena Zernicka-Goetz's laboratory combined three kinds of mouse stem cells, with no egg and no fertilisation.  The cells organised themselves into an embryo model that "displays headfolds with defined forebrain and midbrain regions and develops a beating heart-like structure, a trunk comprising a neural tube and somites, a tail bud containing neuromesodermal progenitors, a gut tube, and primordial germ cells."⁵⁸  The same year Shadi Tarazi and colleagues in Jacob Hanna's laboratory built similar models "starting solely from naive ESCs," that is, from one kind of stem cell.⁵⁹  The models are not embryos and they stop at about day 8.5 of mouse development.  Only "0.1%–0.5%" of the starting clusters in Tarazi's study developed normally to that stage.⁵⁹  But the head-to-tail axis, the neural tube and the segments formed from cultured cells whose internal structures are ordinary, with no egg to supply a pattern.

The logic.  Meyer's argument needs two steps.  The first is that DNA is not the whole story, and the planarian and goldfish results support it.  The second is that the non-DNA information cannot change usefully, so mutation and selection have nothing to work with.  The planarian results cut against that step.  A two-headed worm and a worm with another species' head shape are changes in body layout made without any DNA change, and the worms live.  Neither is a new body plan:  one duplicates a head, and the other copies head shapes that other species already have and fades within weeks.  The goldfish result is also ambiguous in a way Meyer's argument needs it not to be.  A goldfish egg is full of proteins and RNAs made from the mother's own genes, and Sun's study does not separate those from the egg's physical layout.

So the third argument ends where the first two did.  Information outside DNA is real.  It does not show that DNA mutation is the wrong tool, because stem cells can build a body axis without an egg.  And nobody, on either side, has shown how the non-DNA information of a new phylum would have arisen.

Objections and responses

"Comb jellies are now the first branch, so neurons evolved twice.  Convergence is common, so there is nothing left to explain."  The tree does not decide it.  Schultz and colleagues, whose study settled the tree, say there are still "two possible alternatives":  two origins, or one origin with losses in sponges and placozoans.²⁰  And two origins would mean two unexplained origins, not none.  The parts lists in sponges and placozoans show what was available.  They do not show the steps.

"Comb jelly neurons built from different genes show design, not descent."  No design writer found for this article makes that argument, and this article does not make it for them.  A designer could build two nervous systems from different parts, or one from shared parts.  Design alone does not predict which.  Where a prediction can be tested, in the toolkit of Part 3, the pattern is the nested one that common descent predicts.

"Meyer quote-mines Davidson."  Not in the usual sense.  The words he quotes are Davidson's, and the meaning of each sentence is kept.  What chapter 13 leaves out is Davidson's conclusion from the same data, that the first networks were shallow and could be shaped by selection.  Meyer's 2014 epilogue does take up that conclusion and answers it with a fair point:  no such network has been observed (Part 4).

"Heliocidaris is variation within a kind.  It is not a new body plan."  Agreed, and Part 5 says so.  The sea urchin case answers a different claim:  that early-acting changes of global effect have "never" been tolerated and that the central networks "cannot vary."  Two healthy species that set up their body axes differently show that early changes are tolerated.  They do not show that such changes can build a phylum.

"The sea urchin comparisons assume common descent."  The dates do.  The differences do not.  Whatever their history, H. erythrogramma and H. tuberculata now specify their axes and early cell fates differently and both live.  That is enough to answer "never tolerated."

"No laboratory could produce a new phylum.  It would take millions of years, so the absence of an experiment is expected."  True, and it matters for how the claim is stated.  The design argument that no observed process has built a body plan is a claim about what has been shown, and it is correct.  It is not proof that no such process exists.  The mainstream account rests on comparison and inference, as Part 4 says.  Part 7 shows the time per rewired link is not the obstacle.  The unknown is how many links, and how many had to change together.

"Mutations can change body plans, but they add no new information."  This is DeWitt's position.⁹  Whether rewiring adds information depends on how information is measured, which is argued in Measuring Biological Information and When Evolution Creates Information.  Here the data say only this much.  The skeleton network of the laboratory sea urchin is "largely a novel evolutionary invention" compared with its relative's, and both urchins build skeletons.³⁹  New wiring arose, by the authors' reading, and the body part stayed the same.

"The information in the egg came from the mother's genes, so it is genetic after all."  Partly.  Sun's goldfish eggs carried proteins and RNAs made by the mother's genes, and the study cannot separate those from the egg's physical layout.  The planarian results are harder to explain that way.  The two-headed state persisted through repeated regeneration in worms with no difference in their polarity genes, held in electrical patterns that the experimenters could reset (Part 8).

"The synthetic embryos began from embryo-derived stem cells, which already carry the cell's structural information."  They carry the ordinary structures every cell has:  membranes, cytoskeleton, centrosomes.  They do not carry the patterned cortex of an egg.  So the result undercuts the claim that the body plan is laid out in the egg's architecture.  It does not undercut the weaker claim that a cell is needed to read a genome.

"Calling the origin of neurons or body plans 'unexplained' is an argument from ignorance."  This article does not argue from the gaps to design.  It reports which steps have been observed and which are inferred.  Where the mainstream account is strong, as with the shared toolkit, it says so.  Where the design side is right, that no observed change has built a body plan, it says that too.

Issues and unanswered questions

1. How many network links separate two phyla? No one has counted them for any pair of phyla. Without that number neither the mainstream account nor Meyer's "implausibly long waiting times" can be priced for a whole body plan (Part 7).
2. How does a kernel arise? The kernels are conserved, and Hinman and Davidson suggest why they stay the same.³⁸ Neither side has published a stepwise account of how a kernel forms in the first place.
3. What did the first nervous system look like? Whether neurons arose once or twice, and from which cells, is unsettled among the mainstream authors (Part 2). No published count on a fixed gene list says exactly which neural genes comb jellies lack; the totals sit in supplements that were not read for this article.
4. Davidson's other texts. Meyer also quotes Davidson's book The Regulatory Genome (2006, pp. 16 and 195) and a 2009 paper by Davidson and Erwin. Neither could be obtained, so those quotations are not checked here and are not used. The main text of Davidson and Erwin's 2006 Science paper, which first named the kernels, was also not obtained; the definition used here is Peter and Davidson's.
5. Placozoan Hox-like genes. Placozoans are reported to have one Hox-like gene, Trox-2. The primary paper was not read, so the article does not use it.
6. No design-side account of neuron origins. This search found no young-earth or intelligent-design paper on comb jelly or placozoan nervous systems. A design account that predicted the distribution of neural parts across sponges, placozoans and comb jellies would be testable.
7. Fossils. When nervous systems and body plans first appear in the rocks is left to The Cambrian Explosion. Wiring diagrams do not fossilise, so the ancestral shallow networks of Part 4 have no direct fossil test.
8. Page numbers for Davidson 2011. The open copy of Davidson's 2011 essay has no printed page numbers, so its quotations are located here by section. Meyer gives pp. 35–36, 38 and 40.

Verdict

The article asked how well the origin of neurons and of the animal body plans is explained.  Different steps get different answers.

1. Where neurons came from: unfinished on both sides. The tree is now fairly firm, with comb jellies first. The parts of a neuron's signalling system are found in sponges and placozoans, which have no neurons. Whether neurons arose once or twice, and by what steps, is inferred and disputed. No design account of the same data has been published.
2. The shared toolkit: the evidence favours common ancestry. Animals share Hox, Pax-6 and much else, and the parts are spread across the tree in the nested way descent predicts. The finding was "entirely unanticipated," which is a fair point for the creationist. But the creationist arguments that Hox mutants are "monstrosities" do not address the pattern.
3. "Early development and core networks never vary": wrong as stated. The links around the kernels, and early steps such as setting the back-to-belly axis, differ between living sister species that are healthy. The Heidelberg screen that found only lethals was a search for lethals. Meyer is right that the kernels themselves are conserved.
4. "No observed process has built a new body plan": right. Every experiment and every comparison moves, copies, removes or rewires parts of a plan that already exists. The mainstream account depends on ancestral networks that were shallow and easy to change. Davidson inferred them, and no one has observed one. The price of rewiring one link is low, a few million years in a fly-sized population. The number of links a new phylum needs has not been counted.
5. Information outside DNA: real, and not decisive. Egg contents and electrical patterns shape bodies, so Meyer is right that DNA is not the whole story. But stem cells build a body axis without an egg, and non-DNA states can change without harm. So "genetic mutations are simply the wrong tool" does not follow.

The honest summary is that biology has explained a great deal about how animal bodies are built and how their building programmes differ.  It has not shown how a new programme for a new phylum was first put together.

Sources