What can invertebrates tell us about back pain?


Editor's Introduction

Development of the annelid axochord: Insights into notochord evolution

annotated by

The spinal column is a major source of structural support in vertebrates. The cartilaginous discs between each vertebra develop from an embryonic structure called the notochord. But how did the notochord evolve? The experiments in this paper reveal deep similarity between our cartilaginous notochord and the muscular annelid axochord. It turns out we have more in common with segmented worms than meets the eye.

Paper Details

Original title
Development of the annelid axochord: Insights into notochord evolution
Original publication date
Vol. 345, Issue 6202, pp. 1365-1368
Issue name


The origin of chordates has been debated for more than a century, with one key issue being the emergence of the notochord. In vertebrates, the notochord develops by convergence and extension of the chordamesoderm, a population of midline cells of unique molecular identity. We identify a population of mesodermal cells in a developing invertebrate, the marine annelid Platynereis dumerilii, that converges and extends toward the midline and expresses a notochord-specific combination of genes. These cells differentiate into a longitudinal muscle, the axochord, that is positioned between central nervous system and axial blood vessel and secretes a strong collagenous extracellular matrix. Ancestral state reconstruction suggests that contractile mesodermal midline cells existed in bilaterian ancestors. We propose that these cells, via vacuolization and stiffening, gave rise to the chordate notochord.


Defining the properties and characteristics of the last common ancestor of bilaterian animals, Urbilateria, is a key question of the evolution and development field (1). In an attempt to infer a possible urbilaterian precursor for the chordate notochord (23), we reasoned that this structure should occupy a similar position with regard to overall morphology and molecular topography during development and in the adult body plan of living descendants (Fig. 1, A and B); that it should express, during its development, a suite of genes that have proven specific and indispensable for notochord formation in the chordates; and that it should be of widespread occurrence in bilaterian body plans (Fig. 1C). We focused our search on the model annelid Platynereis dumerilii, which is amenable to molecular studies and has retained more ancestral features than Drosophila melanogaster or Caenorhabditis elegans (4). By looking for cell populations that would resemble the vertebrate chordamesoderm (a population of mesodermal midline cells that converge medially to give rise to the notochord; red in Fig. 1A), we identified segmental pairs of mesodermal cells on the non–bone morphogenetic protein (BMP) body side (5) that stood out by early and continuous expression of colA1, encoding collagen type A (Fig. 2, A to D). SiMView light sheet microscopy (6) revealed that these cells moved underneath the neuroectoderm toward the midline until they contacted their bilateral counterpart (movie S1 and Fig. 2E). Subsequently, these cells narrowed and elongated without a net increase in cell surface (fig. S1), and additional adaxial mesodermal cells were observed to intercalate between the elongating pairs (Fig. 2E), reminiscent of the processes by which the chordamesoderm converges and extends (table S1) (7). Lineage tracking by targeted photoconversion of the fluorescent protein kikGR confirmed the origin of these cells from the mesodermal bands (fig. S2).


Fig. 1.  Comparison of notochord and axochord development, gene expression, and anatomy. (A) Notochord development schematized for zebrafish at 9 hours post fertilization (hpf) or 90% epiboly, 14 hpf/neural keel, and 30 hpf/organogenesis stages. (B) Axochord development schematized for Platynereis at 34 hpf, 72 hpf, and 2 months of development. Top left images are in similar orientation with regard to the animal (an)–vegetal (veg) axis. Bold dashed lines represent lines of convergence of neuroectodermal and mesodermal cells, opposite to BMP signaling (blue arrows). The non-BMP body side will be dorsal (d) in vertebrate and ventral (v) in annelid, reflecting inversion of body posture in early chordate evolution (35). Thin black arrows indicate convergence and extension. Thin black dashed lines indicate positions of transverse sections (bottom left). Red, notochord (nch) or axochord (ach); orange, mesoderm (mes); purple, neuroectodermal midline; yellow, medial interneuron column; faint yellow, motor neuron column; gray, sensory interneuron column; blue, epidermis; green, endoderm (end)/gut. Transcription factors defining the respective tissues are written in corresponding colors. Bold black arrows indicate developmental progression. neuroect, neuroectoderm; coe, coelom; dao, dorsal aorta; pmy, primary myocyte; som, somite; vao, ventral aorta; vom, ventral oblique muscle; vlm, ventral longitudinal muscle. (C) Simplified phylogenetic tree. Black boxes illustrate the proposed evolutionary transition from ventral midline contractile cells to notochord.

Panels A and B

A and B are cross-sectional representations of midline development in zebrafish (A) and annelid (B). 

Over time, the paired structures of the zebrafish come together to form structures at the midline. This is characteristic of chordates. For example, humans have a single spinal chord and a single spinal column to support the body. This is not so for annelids. Note how the axochord meets at the midline but does not form a unified structure. One distinction to note is that the nerve bundle is dorsal in zebrafish and ventral in annelids. 

The color coding highlights tissues and regions where genes are expressed during development. 

Panel C

C is an evolutionary tree. Read from bottom to top, it depicts critical events in the evolution of the mesodermal midline beginning many millions of years ago.

The unique location, large size, and specific arrangement of the Platynereis mesodermal midline cells allowed their unambiguous identification after whole-mount in situ hybridization (WMISH) and thus expression profiling by confocal imaging. To test a possible homology of these cells with the chordamesoderm, we chose a chordamesoderm-specific gene set according to the following criteria: (i) specificity—their combined expression uniquely defines the chordamesoderm; (ii) conservation—their chordamesoderm expression is conserved in at least three of four vertebrate species; and (iii) function—they have proven essential for chordamesoderm development or signaling. We thus investigated expression of seven transcription factors (brachyuryfoxAfoxD,twistnotsoxD, and soxE), the signaling molecules noggin and hedgehog [chordin appears absent from annelid genomes (8)], and the guidance factors netrin and slit (table S2 for references). Transcripts for all but one [the notgene (2)] were detected (figs. S3 to S5) in accordance with previously reported brachyury expression (9), and their coexpression confirmed by double WMISH (Fig. 2, F to L). Although none of the genes were exclusively expressed in the annelid mesodermal midline, their combined coexpression was unique to these cells (implying that mesodermal midline in annelids and chordamesoderm in vertebrates are more similar to each other than to any other tissue). It is unlikely that the molecular similarity between annelid and vertebrate mesodermal midline is due to independent co-option of a conserved gene cassette, because this would require either that this cassette was active elsewhere in the body (which is not the case) or that multiple identical independent events of co-option occurred (which is unparsimonious). As in vertebrates, the mesodermal midline resembles the neuroectodermal midline, which expresses foxDfoxAnetrin, slit, and noggin (figs. S6 and S7) but not brachyury or twist. However, unlike in chicken (10), the annelid mesodermal and ectodermal midline populations are not directly related by lineage (fig. S2). Last, the Platynereis mesodermal midline is devoid of paraxis, which is exclusively expressed in laterally adjacent mesoderm (fig. S8), in line with its vertebrate ortholog demarcating paraxial mesoderm (11). In vertebrates, this segregation depends on canonical Wnt signaling, with β-catenin–positive cells preferentially adopting a paraxial fate and position (12). Consistently, we observed nuclear localization of β-catenin in the more-lateral mesoderm only, and β-catenin overactivation converted the mesodermal midline toward a more lateral fate and position (fig. S8).


Fig. 2.  The molecular fingerprint of Platynereis mesodermal midline cells.  (A to C) Bright-field images and (D) confocal z-projection ofcolA1 WMISH. Dotted white circle and line indicate foregut and midline, respectively. DAPI, 4′,6-diamidino-2-phenylindole. (E toE’’’) Snapshots of SiMView time lapse of a live larva with flourescently labeled nuclei showing ventrally converging (green dots) and intercalating (red dots) axochordal cells. Time interval, 2 to 3 hours. (F to K) Confocal z-projections of double WMISH 3-dpf larvae, ventral views, anterior up. (L) Explanatory scheme. vom also weakly express colA and foxD. Foregut also expresses bra and foxA.

Panels A -D

Expression of colA1, which encodes collagen type A, first using traditional bright-field microscopy and again using WMISH.   

The authors looked for colA1 expression becuase cells secrete collagen to form an extracellular matrix. This mesh of fibers reinforces tissues like bone, cartilage, tendon, and skin. 

Panel E

Panel E is a series of images over time.  The pairs of collagen expressing cells migrating toward each others until they touched at the midline. 

Panels F - K

transcription factors, signaling molecules, and guidance factors that co-localize with colA1

Panel L

A diagram showing the expressed transcription factors, signaling molecules, and guidance factors 

We next compared the developmental fate of annelid and vertebrate mesodermal midline cells. Phalloidin staining and expression analysis of muscle markers (fig. S9) revealed that, after elongation, the Platynereis mesodermal midline cells differentiate into the previously described “medial ventral longitudinal muscle” (13) (Fig. 3A). Given the ropelike appearance and axial position of this muscle, we propose to call it “axochord.” A muscular nature of a putative invertebrate counterpart of the chordate notochord is consistent with the observation that in the most basal chordate, amphioxus, the notochord is composed of specialized muscle cells (14) and expresses the same muscle markers (15). We further observed segmental sets of transverse muscles connecting to the axochord (“ventral oblique muscles”) (13) (Fig. 3, A and B, and fig. S3). Scanning electron microscopy revealed that, in adult worms, the axochord is deeply embedded in the fibrous sheath of the ventral nerve cord (16) and remains connected to the transverse muscles (Fig. 3, C and D). Immunostainings confirmed its axial position between neuropil and blood vessel (fig. S12; similar to the notochord; Fig. 1, A and B). Axochord contractility was evident from live imaging (fig. S9, E to G, and movie S2) and occurred in alternation with the transverse muscles (movie S3). Electron micrographs confirmed the muscular nature of axochordal cells and revealed a tight physical connection to transverse muscles (Fig. 3, E to I). Laser ablation of the axochord impaired crawling (fig. S10 and movie S4) and confirmed anchoring of the transverse musculature. Additionally, we found that the Platynereistransverse muscles share a specific molecular profile (en+foxD; fig. S11) with the vertebrate pioneer myocyctes flanking the notochord (17).


Fig. 3.  The axochord is a ventral medial longitudinal muscle.  (A) Ventral view of Platynereis immunolabeled musculature and nervous system, z-projection of confocal stack. Vnc, ventral nerve cord; mvlm, median ventral longitudinal muscle (axochord). (B) Pseudocolored scanning electron micrograph of Platynereisjuvenile trunk, dorsal view. Axochord (deep pink) and attached oblique muscles (light pink). (C and D) Pseudocolored scanning electron micrographs. (C) Adult cross section. g, gut; vbv/dbv, ventral/dorsal blood vessel. (D) Dissected specimen showing axochord and oblique muscles embedded in the vnc sheath. Closeup illustrates axochord cell morphology. (E) Transmission electron micrograph showing axochordal cells, ventral oblique muscles, neuronal midline (nm), and the neuropil (np). (F) Closeup of area in black square in (E). One axochordal cell is outlined with dashed black line; asterisk indicates extracellular matrix. (G) Schematic drawing of (E). (H) Closeup of are in white square in (E), showing interdigitations between axochordal cells and oblique muscles. (I) Closeup of (F) (orange square) showing cross-sectioned myofibers (red arrow).

Panel A 

A is a high resolution ventral image of Platynereis musculature (purple) and nervous system (green). These structures have been labeled using an antibody against microtubules (green) and a fluorescent small-molecule that binds actin filaments (purple). 

Panel B

B is a scanning electron micrograph that has been artificially colored to show the axochord (pink) and the transverse muscles (light pink) that attach to it in a juvenile Platynereis. 

Panel C

C is a scanning electron micrograph of an adult Platynereis. It's a cross section through the body. The gut, the major blood vessels, and the nerve bundle are clearly visible. Note the position of the axochord between the ventral nerve bundle and the ventral blood vessel. These relative positions are similar to the arrangement in chordates. The only difference is that the chordate nerve bundle is dorsal. 

Panel D

D is a scanning electron micrograph of a dissected adult Platynereis. This image shows the organization of the mature muscles as well as the shape and size of the cells that make up the axochord. 

Panel E

E is a transmission electron micrograph of a cross section through the axochord. 

Panel F

F is a higher magnification view of the black box drawn in E. A black dotted line denotes the outline of a single axochord cell.

Panel G

G is a diagram of E. It shows the different cell types captured in image E. 

Panel H

H is a close up image of the white box in E. There is a ruffled interface between the axochord cell and the oblique muscles. This texture provides more surface area for the cells to adhere to each other and makes a sturdy connection between the muscular midline and the oblique muscles. 

Panel I

I is a higher magnification view of the orange box from F. The red arrow points to a single muscle fiber. 

We next examined whether an axochord is also present in other annelids, lophotrochozoans, or protostomes (Fig. 1C). Phalloidin stainings had revealed ventral midline muscles in almost all annelid families (table S3), yet in some cases only pairs of ventral muscles had been reported (18, 19). One such example is Capitella teleta, which belongs to the second big clade of annelids, the Sedentaria (beside Errantia, to which Platynereis belongs). We investigated development and molecular identity of ventral muscle fibers in Capitella (18) and found that (i), preceding metamorphosis, these converge and fuse into an axochord (Fig. 4, A to D, and fig. S13); (ii) the expression patterns of foxAnogginbrachyurynetrin (Fig. 4, B and D, and fig. S13, B and C); twist2 (20); andhedgehog (21) are consistent with coexpression in the axochord; and (iii) pairs of transverse muscles connect to the Capitella axochord (Fig. 4E) as in Platynereis. We also investigated the annelid Owenia fusiformis that belongs to the most basal annelid family (22) and likewise found an axochord connected to transverse muscle fibers (Fig. 4F). A similar arrangement also occurs in mollusk (23) and brachiopod larvae (24), and ventral midline muscles are observed in most lophotrochozoan phyla (table S3), suggesting that an axochord is ancestral for lophotrochozoa. The lophotrochozoan axochord is a genuinely paired structure, composed of left and right adjacent muscle strands that often bifurcate anteriorly and/or posteriorly, as also observed in chaetognaths (Fig. 4, G and H), a possible protostome outgroup (2526) (Fig. 1C). An ancestral state reconstruction based on our data and on a survey of bilaterian musculature (table S3) favored the presence of an axochord in protostome ancestors (fig. S14).


Fig. 4.  An axochord is widespread in protostomes.  (A to D) Convergence of ventral muscles into an axochord (white arrows) in Capitella, illustrated by phalloidin stainings (A and C) and WMISH (red in B and D) of selected axochord markers. (E) Axochord and transverse musculature in a stage-9 Capitellalarva. Axochord bifurcation at the level of the foregut. (F) Axochord (white arrows) and transverse muscle fibers (green arrows) in juvenile Owenia, below the circumferential layer of longitudinal muscles. (G) Axochord (white arrow) in adult chaetognath (inset) bifurcating at the level of foregut. (H) Closeup of the axochord in the same specimen. Note elongated median nuclei and the two parallel strands of striated myofibers. All panels are z-projections of confocal stacks, ventral view, anterior up, white dashed line outlining foregut.

Panel A 

At this developmental stage, the phalloidin stain (pink) shows one muscle on each side of the midline. DNA is stained blue. 

Panel B

A probe for the axochord marker noggin reveals a similar arrangement. Two regions of expression, one on each side of the midline. 

Panel C

This is Capitella at a later stage of development. The muscles have converged to form a single axochord at the midline. 

Panel D

A probe for the axochord marker brachyury shows expression that is limited to the midline. 

Panel E

Capitella phalloidin stain at a later stage of development. Note how the axochord splits into two near the foregut. Their axochord is a paired structure.  

Panel F

This is a phalloidin stain in another species of annelid, Owenia. Note that there is a similar structure to the other annelids. An axochord (white arrows) at the midline with attached transverse muscles (green arrows). 

Panel G

Chaetognath axochord also bifurcates at the foregut. This is because the axochord is paired. 

Panel H

This is a higher magnification view of the axochord from G. Note how the phalloidin stained muscle fibers do not form a single structure, but are paired. 

Regarding deuterostomes, previous studies on the origin of the notochord focused on the hemichordate stomochord, an unpaired chordoid outpocketing of the pharynx, as a possible notochord homolog (27). Speaking against this hypothesis is its very anterior position and the absence of brachyuryfoxA, and noggin expression (27,28). goosecoid, hedgehog, and colA expression rather suggest homology to the vertebrate prechordal plate (2933). The pygochord, a stiff vacuolated rod in the posterior trunk of ptychoderid hemichordates (34), lies dorsal to the ventral blood vessel in the ventral mesentery. This stands in contrast to both axochord and notochord, which are positioned between blood vessel and nerve cord (Fig. 1). Thus, vacuolization in the hemichordate stomochord and pygochord might have occurred independent to that of the chordate notochord. Unfortunately, no data are available for the specification and developmental fate of ventral mesodermal midline cells in hemichordates or larval echinoderms; except for Protoglossus, no ventromedian musculature has yet been observed (table S3).

Our study of annelid development reveals a population of mesodermal cells that converge and extend along the ventral midline and express a combination of transcription factors, signaling molecules, and guidance factors that closely matches that of the vertebrate chordamesoderm. These comparative data suggest that a similar population of mesodermal midline cells already existed in urbilaterian ancestors but leave open its ancient developmental fate. In annelids, these cells differentiate into an axochord; our investigation of chaetognath musculature and an ancestral state reconstruction based on comparative anatomy (fig. S14) suggest that a similar paired longitudinal muscle existed in protostome ancestors. Yet, in the absence of detailed investigations of expression profile and developmental fate of mesodermal midline cells in basal ecdysozoans and deuterostome ambulacrarians, the nature of ventral midline tissue in urbilaterians remains undecided. It might have constituted sheath-secreting mesenchyme that was independently converted into muscular axochord and notochord in lophotrochozoans and chordates, respectively; alternatively, this tissue might have been contractile already and transformed into axochord in protostomes and notochord in chordates (Fig. 1C). Regardless of its nature, dorsoventral axis inversion (35) would have brought the ventral midline tissue into a dorsal position in the chordate lineage, and the appearance of incompressible vacuoles (14) would have gradually transformed it into a stiff rod of constant length; the amphioxus notochord could then be regarded a vestige of a contractile-cartilaginous transition. This transition could have involved the co-option of not expression (which is absent from the axochord, see above) given that zebrafish not null mutants form muscle tissue instead of notochord (36). Last, in vertebrates, the notochord was complemented by a rigid backbone that crucially contributed to the success of our phylum.

Supplementary Materials


Materials and Methods

Figs. S1 to S16

Tables S1 to S3

References (37–152)

Movies S1 to S4

References and Notes

  1. E. M. De Robertis, Y. Sasai, A common plan for dorsoventral patterning in Bilateria. Nature 380, 37–40 (1996).

  2. D. L. Stemple, Structure and function of the notochord: An essential organ for chordate development. Development 132, 2503–2512 (2005).

  3. N. Satoh, K. Tagawa, H. Takahashi, How was the notochord born? Evol. Dev. 14, 56–75 (2012).

  4. S. Denes, G. Jékely, P. R. Steinmetz, F. Raible, H. Snyman, B. Prud’homme, D. E. Ferrier, G. Balavoine, D. Arendt, Molecular architecture of annelid nerve cord supports common origin of nervous system centralization in bilateria. Cell 129, 277–288 (2007).

  5. D. Arendt, K. Nübler-Jung, Dorsal or ventral: Similarities in fate maps and gastrulation patterns in annelids, arthropods and chordates. Mech. Dev. 61, 7–21 (1997).

  6. R. Tomer, K. Khairy, F. Amat, P. J. Keller, Quantitative high-speed imaging of entire developing embryos with simultaneous multiview light-sheet microscopy. Nat. Methods 9, 755–763 (2012).

  7. J. B. Wallingford, S. E. Fraser, R. M. Harland, Convergent extension: The molecular control of polarized cell movement during embryonic development. Dev. Cell 2, 695–706 (2002).

  8. O. Simakov, F. Marletaz, S. J. Cho, E. Edsinger-Gonzales, P. Havlak, U. Hellsten, D. H. Kuo, T. Larsson, J. Lv, D. Arendt, R. Savage, K. Osoegawa, P. de Jong, J. Grimwood, J. A. Chapman, H. Shapiro, A. Aerts, R. P. Otillar, A. Y. Terry, J. L. Boore, I. V. Grigoriev, D. R. Lindberg, E. C. Seaver, D. A. Weisblat, N. H. Putnam, D. S. Rokhsar, Insights into bilaterian evolution from three spiralian genomes. Nature 493, 526–531 (2013).

  9. D. Arendt, U. Technau, J. Wittbrodt, Evolution of the bilaterian larval foregut. Nature 409, 81–85 (2001).

  10. M. Catala, M. A. Teillet, E. M. De Robertis, M. L. Le Douarin, A spinal cord fate map in the avian embryo: While regressing, Hensen’s node lays down the notochord and floor plate thus joining the spinal cord lateral walls. Development 122, 2599–2610 (1996).

  11. R. Burgess, P. Cserjesi, K. L. Ligon, E. N. Olson, Paraxis: A basic helix-loop-helix protein expressed in paraxial mesoderm and developing somites. Dev. Biol. 168, 296–306 (1995).

  12. W. E. Reintsch, A. Habring-Mueller, R. W. Wang, A. Schohl, F. Fagotto, beta-Catenin controls cell sorting at the notochord-somite boundary independently of cadherin-mediated adhesion. J. Cell Biol. 170, 675–686 (2005).

  13. H. Fischer, T. Henrich, D. Arendt, The normal development of Platynereis dumerilii (Nereididae, Annelida). Front. Zool. 7, 31 (2010).

  14. J. E. Webb, The role of the notochard in forward and reverse swimming and burrowing in the amphioxus Branchiostoma lanceolatum. J. Zool. 170, 325–338 (1973).

  15. M. M. Suzuki, N. Satoh, Genes expressed in the amphioxus notochord revealed by EST analysis. Dev. Biol. 224, 168–177 (2000).

  16. D. G. Baskin, Fine structure, functional organization and supportive role of neuroglia in Nereis. Tissue Cell 3, 579–587 (1971).

  17. P. D. Currie, P. W. Ingham, Induction of a specific muscle cell type by a hedgehog-like protein in zebrafish. Nature 382, 452–455 (1996).

  18. S. D. Hill, B. C. Boyer, Phalloidin labeling of developing muscle in embryos of the polychaete Capitella sp. I. Biol. Bull. 201, 257–258 (2001).

  19. Filippova, G. Purschke, A. B. Tzetlin, M. C. M. Müller, Reconstruction of the musculature of Magelona cf. mirabilis (Magelonidae) and Prionospio cirrifera (Spionidae) (Polychaeta, Annelida) by phalloidin labeling and cLSM. Zoomorphology 124, 1–8 (2005).

  20. K. K. Dill, K. Thamm, E. C. Seaver, Characterization of twist and snail gene expression during mesoderm and nervous system development in the polychaete annelid Capitella sp. I. Dev. Genes Evol. 217, 435–447 (2007).

  21. E. C. Seaver, L. M. Kaneshige, Expression of ‘segmentation’ genes during larval and juvenile development in the polychaetes Capitella sp. I and H. elegans. Dev. Biol. 289, 179–194 (2006).

  22. Weigert, C. Helm, M. Meyer, B. Nickel, D. Arendt, B. Hausdorf, S. R. Santos, K. M. Halanych, G. Purschke, C. Bleidorn, T. H. Struck, Illuminating the base of the annelid tree using transcriptomics. Mol. Biol. Evol. 31, 1391–1401 (2014).

  23. M. Scherholz, E. Redl, T. Wollesen, C. Todt, A. Wanninger, Aplacophoran mollusks evolved from ancestors with polyplacophoran-like features. Curr. Biol. 23, 2130–2134 (2013).

  24. Altenburger, A. Wanninger, Comparative larval myogenesis and adult myoanatomy of the rhynchonelliform (articulate) brachiopods Argyrotheca cordata, A. cistellula, and Terebratalia transversa. Front. Zool. 6, 3 (2009).

  25. F. Marlétaz, E. Martin, Y. Perez, D. Papillon, X. Caubit, C. J. Lowe, B. Freeman, L. Fasano, C. Dossat, P. Wincker, J. Weissenbach, Y. Le Parco, Chaetognath phylogenomics: A protostome with deuterostome-like development. Curr. Biol. 16, R577–R578 (2006).

  26. F. Marlétaz, A. Gilles, X. Caubit, Y. Perez, C. Dossat, S. Samain, G. Gyapay, P. Wincker, Y. Le Parco, Chaetognath transcriptome reveals ancestral and unique features among bilaterians. Genome Biol. 9, R94 (2008).

  27. K. J. Peterson, R. A. Cameron, K. Tagawa, N. Satoh, E. H. Davidson, A comparative molecular approach to mesodermal patterning in basal deuterostomes: The expression pattern of Brachyury in the enteropneust hemichordate Ptychodera flava. Development 126, 85–95 (1999).

  28. J. Gerhart, C. Lowe, M. Kirschner, Hemichordates and the origin of chordates. Curr. Opin. Genet. Dev. 15, 461–467 (2005).

  29. J. Gerhart, The deuterostome ancestor. J. Cell. Physiol. 209, 677–685 (2006).

  30. N. Miyamoto, H. Wada, Hemichordate neurulation and the origin of the neural tube. Nat. Commun. 4, 2713 (2013).

  31. J. A. Belo, L. Leyns, G. Yamada, E. M. De Robertis, The prechordal midline of the chondrocranium is defective in Goosecoid-1 mouse mutants. Mech. Dev. 72, 15–25 (1998).

  32. E. M. Pera, M. Kessel, Patterning of the chick forebrain anlage by the prechordal plate. Development 124, 4153–4162 (1997

  33. W. L. Leung, S. Y. Y. Wong, D. Chan, P. P. L. Tam, K. S. E. Cheah, Loss of procollagen IIA from the anterior mesendoderm disrupts the development of mouse embryonic forebrain. Dev. Dyn. 239, 2319–2329 (2010).

  34. K. Nübler-Jung, D. Arendt, Dorsoventral axis inversion: Enteropneust anatomy links invertebrates to chordates turned upside down. J. Zool. Syst. Evol. Res. 37, 93–100 (1999).

  35. D. Arendt, K. Nübler-Jung, Inversion of dorsoventral axis? Nature 371, 26 (1994).

  36. Acknowledgments: We thank A. Miyawaki for kikGR, R. Renkawitz-Pohl for the antibody against β-tubulin; P. McCrea for the antibody against β-catenin, A. Altenburger for brachiopod data; S. Kaul-Strehlow, G. Mayer, M. V. Sørensen, and K. Worsaae for insightful discussions; I. Haußer-Siller at the Electron Microscopy Core Facility (EMCF) (BioQuant, Heidelberg University); and the EMBL EMCF. The work was supported by the European Union’s Seventh Framework Programme project “Evolution of gene regulatory networks in animal development (EVONET)” [215781-2] (A.L.), the Zoonet EU-Marie Curie early training network [005624] (A.F. and R.T.), the European Molecular Biology Laboratory (A.L., T.B., M.T.H., A.H.L.F., O.S.,P.R.H.S., and R.T.), the EMBL International Ph.D. Programme (A.L., T.B., A.H.L.F., P.R.H.S., R.T.), the Boehringer Ingelheim Foundation (O.S.), the Howard Hughes Medical Institute (P.J.K.), and the Visiting Scientist Program at the Janelia Farm Research Campus (M.H.T., R.T., D.A., and P.J.K.). A.L. and T.B. designed, performed, and documented most of the experiments; developed protocols; and wrote the manuscript. R.T. initiated and, together with M.H.T. and P.J.K., developed the Platynereis SiMView live imaging assay, and M.H.T. acquired the SiMView data. A.H.L.F. investigated musculature development and gene expression in Platynereis. M.H.T. contributed to the expression analysis. O.S. did transmission electron microscopy with T.B. and A.L. P.R.H.S. cloned and characterized twist. R.T. developed the injection protocol with N. Kegel and B. Backfisch. D.A. coordinated the study, wrote the manuscript, and drew the summary figure. The plasmid encoding the kikGR construct is covered by a material transfer agreement provided by the RIKEN Brain Science Institute. Alignments used for tree building are deposited on Dryad (10.5061/dryad.1ct82).