These expression patterns suggest that the Hh pathway is involved in segmental patterning in Platynereis. However, striped expressions of Hh pathway genes appears distinctly later during trochophore development (Fig S5A-F) than those of the previously identified Platynereis segment polarity-like genes Pdu-en (8), Pdu-NK4 and Pdu-Lbx (9). Thus the Hh pathway might play a late role in segment formation, in shaping the individual segments after the specification of the segmental pattern by other genes. To test this hypothesis, we used two small molecules, cyclopamine (13) and SANT-1 (14), known for antagonizing Hh signaling in vertebrates by binding to Smo proteins (7). We defined three time windows for treatments (Fig S7): an early trochophore window before any component of the Hh pathway is distinctly expressed in stripes (20 to 30 hpf), a mid-trochophore window when Pdu-hh, Pdu-Gli and Pdu-ptc are all expressed in stripes (32 to 48 hpf) and a late trochophore window (54 hpf onward) when striped expression for Pdu-ptc and Pdu-Gli fade, but still a few hours before the larva starts to elongate and to form externally visible segments (15). Mid-trochophore treatments with increasing doses of cyclopamine or SANT-1 on mixed batches of trochophores coming from several females gave remarkably similar abnormal larvae (Fig S8.I and II). They fail to develop properly shaped segments at 3 days: They do not elongate properly, remaining ovoid in shape (Fig 2 A, B, Fig S8.II.A-C); Segmental grooves do not appear (Fig 2A, B, yellow arrows) and segmental trochae (circular ciliary belts borne by larval segments) are reduced (Fig S8.II.D-F); Appendage outgrowths and chaetae do not appear (Fig S8.II.A-C). In addition, the axonal scaffolds of the ventral nerve cord and segmental peripheral nervous system (Fig 2C, D, Fig S8.II.D-F) both appear underdeveloped and disorganized. In place of the metameric peripheral nerves (red asteriks), an asymmetric proliferation of axons projecting haphazardly to the periphery forms, indicating that the segmental pattern of the nervous system is lost. The effects of two unrelated drugs and of different treatment time windows on embryos suggest that the phenotypes obtained are caused by blocking Hh signaling (Fig S8, Supporting Online Text). In particular, the phenotypes cannot be due to a general toxicity of the drugs since late trochophore treatments at a time of massive proliferation and morphogenesis cause few defects (Fig S8.II.G-K). We also tested the role of Pdu-smo on posterior segment addition by treating individuals with cyclopamine after caudal regeneration. One half (12/24) of treated individuals show regenerated posterior parts with fewer, abnormally shaped segment anlagen and reduced segmental grooves (Fig S12).

To analyze further the effects of cyclopamine on the segmental pattern, we looked at the expression patterns of the previously known segment polarity-like genes in cyclopamine treated larvae (7). To take into account a strong maternal effect polymorphism in the response to cyclopamine (Supporting Online Text and Fig S11A), we carefully selected only those batches of embryos that presented medium and strong morphological phenotypes (scored in 3-day larvae) for the gene expression study. Expression patterns for medium strength morphological phenotypes are presented in Fig 2E-L and Fig S10. Complete results including 1 or 2 repeat experiments for each gene indicate that the effects of cyclopamine on gene expression correlate well with the strength of the phenotype in each same-female batch studied (Fig S11B). Early-stage cyclopamine treatment has no effect on the striped expressions of Pdu-en, Pdu-NK4 and Pdu-Lbx (Fig S10 and Fig S11B) seen in 30 hpf larvae, showing that Pdu-Smo does not act on the segmental pattern independently of the morphogen Pdu-Hh and that the Hh pathway does not intervene in the initial laying of the segmental pattern. By contrast, mid-trochophore cyclopamine treatment at high dose leads to a complete or almost complete loss of striped expression for Pdu-hh, Pdu-en, Pdu-Wnt1 (Fig 2E, F, G, respectively), showing that the segmental pattern is largely affected well before metamorphosis. Pdu-ptc is also downregulated (Fig S9) as it is observed in hh-mutant flies (16) and vertebrates lacking shh expression (17). Pdu-Gli (Fig 2I) stripes are mostly persistent even at high doses of cyclopamine. The specificity of the effects of mid-trochophore cyclopamine treatment on the patterning of the segmental ectoderm is demonstrated by the persistence of gene expression in tissues other than segmental ectoderm. Hence the expression of Pdu-en and Pdu-Lbx in segmental mesoderm (Fig 2F and 2K, red arrows), the stomodeal expression of Pdu-hh and Pdu-Wnt11 (Fig 2E and 2H, blue asterisks) and the pygidial expression of Pdu-Wnt11 and Pdu-Cdx (Fig 2H and 2L, green asterisks) are maintained whereas the proctodeal expression of Pdu-Wnt1 is slightly enlarged (Fig 2G, green asterisks).

The effect of cyclopamine on Pdu-hh stripes must be indirect if we suppose that the direct targets of Hh signaling are only those cells expressing Pdu-Gli, away from segment borders. In Drosophila (Fig 3), wingless (wg) expression is maintained by Hh signaling in the anterior compartment just anterior to the en stripe whereas wg signaling is necessary to maintain en and hh striped expression in the posterior compartment (18). In Platynereis, Pdu-Wnt1, the orthologue of Drosophila wingless (8), is expressed just anterior to Pdu-en/Pdu-hh stripes on the other side of segmental grooves. Its strong downregulation by cyclopamine (Fig 2G) is consistent with a loop of regulation similar to the one known in Drosophila. Ladybird, the fly gene orthologous to Pdu-Lbx is expressed in epidermal stripes overlapping wingless stripes in the anterior compartment of the epidermis and is positively regulated by wingless (19). By contrast, cyclopamine treatment in the annelid does not abolish completely Pdu-Lbx stripes in trochophores (Fig 2K), suggesting that other factors are involved in its maintenance. Another difference between the annelid and the fly is that Pdu-Wnt1 is not expressed in complete circular stripes in the trochophore (it does during posterior growth) but only in the lateral parapodial field (Fig 2G). It was thus interesting to look at Pdu-Smo regulation of other Wnt genes potentially involved in segment formation. Pdu-Wnt11 is expressed in thick stripes in the posterior halves of segments and is strongly downregulated by cyclopamine (Fig 2H). Pdu-Wnt5 is expressed in the anterior halves of segments and is by contrast much more resistant to cyclopamine (Fig 2J).

Our work demonstrates the involvement of the Hh pathway in segment formation outside of arthropods. It comes in contrast with earlier studies of hh orthologues in annelid species (20, 21). The comparison of the effect of Hh signaling inhibition in Platynereis and in insects, i.e. the fruitfly and the coleopteran Tribolium (3), reveals extensive similarities. As in insects, Platynereis Hh presumably diffuses anteriorly to maintain Wnt signaling anterior to the hh expressing cells (Fig 3), which may in turn be crucial to maintain segment boundary gene expressions, including Pdu-hh. As in insects, Hh is not necessary in the initial setting of the segmental pattern but is required to maintain this pattern before the morphological appearance of segments. The comparison of segment polarity gene patterns between Platynereis and arthropods (Fig 3) reveals four independent players with remarkably similar expressions: engrailed, ladybird/Lbx, the Hh pathway and Wnt1/wingless signaling. The most likely explanation of these similarities is that these genes were already playing similar roles in a metameric protostome ancestor. The alternative explanation would be a parallel recruitment of these genes for similar functions in annelids and arthropods. Since they are not known to be part of a conserved core regulatory network or “kernel” (22) that might have been co-opted en bloc, each gene would have been recruited independently, which seems unlikely. Altogether, these four players are expressed in the same spatial relationship across the annelid segment boundary as they are across the parasegmental boundary in arthropods (Fig 3). This suggests that these two boundaries are homologous as was proposed earlier (8). Therefore, the segmented exoskeleton of the arthropods must have evolved out of frame with the ancestral protostome segmentation (Fig S13). This ancestral protostome segmentation is nowadays “recapitulated” as parasegmental patterning in the embryos of extant arthropods.

Figs. S1to S13, Supporting Text, Materials and Methods, Additional references