BMP and Delta/Notch signaling control the development of amphioxus epidermal sensory neurons: insights into the evolution of the peripheral sensory system

The evolution of the nervous system in metazoan animals has been a topic of great interest. It is thought that the first nerve cell (neuron) evolved in a common ancestor of cnidarians and bilaterians, because neuronal cell types are absent in their sister group, the sponges (Galliot et al., 2009). Although putative sensory cells do exist in the sponge, these cells do not form synapses and do not possess dendrites and axons (Richards et al., 2008). Although cnidarians typically have a diffuse nerve net, some regionalization patterns of neurons and their neurites have been identified in several types of cnidarians (Galliot et al., 2009; Watanabe et al., 2009). It is possible that these cnidarian nerve rings might represent a primitive organized nervous system. After the divergence of cnidarians and bilaterians, a centralized nervous system evolved in the bilaterian lineage (Arendt et al., 2008; Holland, 2003). The evolutionary origin of the central nervous system (CNS) in bilaterian animals has been highly controversial. Traditionally, it was widely accepted that the CNSs of protostomes and deterostomes (or specifically, chordates) evolved independently because they are structurally different and are located in opposite positions along the dorsal-ventral axis (reviewed by Arendt and Nubler-Jung, 1999). However, during the last two decades, many discoveries have been made that challenge this view. Between the protostome and chordates, there are remarkable similarities in gene expression and conserved signaling pathways in the anterior/posterior and medial/lateral patterning of the CNS (reviewed in De Robertis, 2008). This finding has led to the hypothesis that a single CNS evolved in the basal bilaterians, and a dorsal-ventral axis inversion occurred in the chordate lineage.

Conversely, the presence of an ectodermal nerve net and the lack of an apparent CNS in the hemichordates (a group of deuterostome animals) reinforce the argument for independent origins of the CNS in protostomes and chordates (Arendt et al., 2008; Lowe et al., 2003). This controversy is complicated further by the fact that the relationship between the CNS and the peripheral nervous system (PNS) is not always well defined. In addition to the CNS, both protostomes and chordates have a PNS with ectodermal sensory neurons. In the fruit fly Drosophila, inhibition of Dpp/BMP signaling by its antagonist Sog (also known as Chordin) during early embryogenesis induces the formation of a ventral neurogenic region that gives rise to the ventral nerve cord and some ventral PNS neurons; subsequently, high levels of BMP signaling in the dorsal ectoderm promote the formation of the dorsal and lateral PNS neurons (Rusten et al., 2002). Similarly, in vertebrate embryos, the CNS is derived from the dorsal neural ectoderm, where antagonists are present to inhibit the BMP signaling. By contrast, most of the neurons of the vertebrate PNS are derived from two ectodermal tissues, the neural crest and the neurogenic placodes, which originate from the boundary between the CNS and the epidermal ectoderm (Baker and Bronner-Fraser, 2001; Graham and Begbie, 2000; Le Douarin and Kalcheim, 1999; Le Douarin and Dupin, 2003; Schlosser, 2006). However, the neural crest and the neurogenic placodes are usually considered to be vertebrate innovations (Shimeld and Holland, 2000), and their evolutionary origins remain largely unresolved. In hemichordates, it is highly debated whether the dorsal or ventral tracts of the axons along with the collard cord correspond to the CNS, and whether the ectodermal nerve net corresponds to the CNS or PNS (Holland, 2009; Nomaksteinsky et al., 2009). These uncertainties have hindered our understanding of the evolution of the nervous system. Therefore, it is important to determine the correct homologous relationships of both the CNS and PNS among the key metazoan animal groups by comparative studies.

To gain more insight into the evolution of the nervous system, we focused on the mechanisms of neurogenesis in the cephalochordate amphioxus. Amphioxus is an invertebrate chordate that shares key chordate characteristics (a dorsal CNS, notochord, segmented somites, and pharyngeal gill slits) with vertebrate animals. Among the three major chordate groups (amphioxus, tunicates and vertebrates), the adult body plan and embryology of the amphioxus are highly similar to those of the vertebrates (Whittaker, 1997), whereas most tunicate larvae undergo metamorphosis and lose many chordate characteristics as they become sessile adults. Therefore, it was traditionally thought that amphioxus is the sister group of the vertebrates and that tunicates compose the basal chordate group; however, recent phylogenomic analyses have reverted their positions and placed amphioxus as the basal chordate group and tunicates as the sister group of the vertebrates (Blair and Hedges, 2005; Bourlat et al., 2006; Delsuc et al., 2006; Philippe et al., 2005; Putnam et al., 2008). In addition to the dorsal CNS, amphioxus also have a PNS comprising several types of sensory neurons (Wicht and Lacalli, 2005). Among these amphioxus sensory neurons, the solitary type I receptor cells are the most abundant population of the epidermal sensory neurons (ESNs) and are scattered along the bodies of developing amphioxus larvae and adults (Holland and Yu, 2002; Lacalli and Hou, 1999) (Fig. 1A,B). The cell body of the receptor is located within the epidermis, and a long axon from the base of the cell body extends into the CNS (Fig. 1C,D). Several axons originated from local ESNs merge into parallel bundles, and these axon bundles run along the myotome boundaries on the flank and connect to the CNS (Fig. 1E-I). During the early neurula stage, these ESNs differentiate from the ventral ectoderm and are first characterized by the expression of a pan-neural marker Hu/Elav (Fig. 1J-M) (Benito-Gutierrez et al., 2005a; Satoh et al., 2001); subsequently, at the mid- to late-neurula stage, many of these differentiating ESNs appear to delaminate from the ectoderm and invaginate into the sub-epidermal space to migrate towards the dorsal side of the neurula (Fig. 1N-Q) (Benito-Gutierrez et al., 2005b; Kaltenbach et al., 2009). At the late neurula to early larval stage, the ESNs re-insert into the epidermal ectoderm and send out axons to innervate the CNS (Kaltenbach et al., 2009), and many of these ESNs are immunoreactive to antibodies against the neurotransmitter γ-aminobutyric acid (GABA) (Fig. 1R-W) (Anadon et al., 1998). Using embryonic ontogeny and expression data for several homologs of placode marker genes, it has been speculated that these amphioxus ESNs are homologs of vertebrate placode-derived neurons (Benito-Gutierrez et al., 2005a; Benito-Gutierrez et al., 2005b; Holland, 2005; Holland and Yu, 2002; Kaltenbach et al., 2009; Kozmik et al., 2007; Mazet et al., 2004; Meulemans and Bronner-Fraser, 2007; Rasmussen et al., 2007). However, the molecular mechanism controlling the specification of amphioxus ESNs remains unclear, and this situation has impeded our understanding of the relationship of these ESNs to the peripheral sensory neurons in other animals. To address this limitation, we set out to determine the developmental mechanisms underlying the formation of the ESNs in amphioxus.

Here, we show that the ventral ectoderm of the amphioxus embryo is a neurogenic domain for ESNs. We demonstrate that Delta/Notch signaling is involved in the specification of individual ESNs from neighboring epidermal cells in this domain and that BMP signaling levels regulate the formation of this putative neurogenic domain within the ventral ectoderm. Additionally, we show that the inductions of the dorsal CNS and the ventral ESN progenitor domain are separate events that require opposing BMP signaling levels during amphioxus embryogenesis. Our results demonstrate that conserved Delta/Notch and BMP signaling pathways govern the specification and distribution of peripheral sensory neurons in amphioxus, respectively. These results suggest that a deep homology exists in the PNS sensory neuron formation in metazoan animals.

Amphioxus (Branchiostoma floridae) adults were collected in Tampa Bay, Florida, USA, during the summer breeding season. The gametes were obtained by electric stimulation. Fertilization and subsequent culturing of the embryos were carried out as previously described (Yu and Holland, 2009). Amphioxus embryos were staged according to Hirakow and Kajita (Hirakow and Kajita, 1991; Hirakow and Kajita, 1994). For embryos in the neurula stage, we further defined N0 stage as the onset of neurulation, which is characterized by the dissociation of the dorsal epidermal ectoderm from the neural plate before formation of somites. We made a clear distinction between the N1 (early neurula, 1-3 somites), N2 (hatching neurula, 4-8 somites) and N3 (late neurula, 9-12 somites) stages based on the number of somites.

Notch signaling was inhibited by DAPT (Geling et al., 2002) treatment. DAPT (Calbiochem) was dissolved in DMSO to prepare a 10 mM stock solution. The DAPT stock solution was diluted in filtered seawater to a final concentration of 50 μM and applied to amphioxus embryos from the G5 stage (late gastrula). The control embryos were treated with filtered seawater containing an equal amount of DMSO. Manipulation of BMP signaling was achieved by treatment with 250 ng/ml of recombinant zebrafish Bmp4 protein (zBMP4, R&D Systems), as previously described (Yu et al., 2007), or by the small molecular inhibitor dorsomorphin (Sigma). Dorsomorphin was dissolved in DMSO to prepare a 10 mM stock solution, and the experimental embryos were treated with dorsomorphin at 8 or 16 μM from the G5 stage. The control embryos were treated with an equal amount of DMSO in filtered seawater.

According to the draft genome sequence from the Joint Genome Institute (JGI) and the annotated protein ID 233300 for amphioxus Achaete-scute homolog gene product (Satou et al., 2008), a pair of Achaete-scute homolog specific primers were designed, and BamHI and EcoRI restriction enzyme sites were added to their 5′ ends, respectively (AS1F-BamHI, 5′-CGGGATCCCGCAAGCGGAGGATAAACTTC-3′ and AS1R-EcoRI, 5′-CGGAATTCCGAACCACGATGTGAAGTCG-3′). The 5′ and 3′ parts of the Achaete-scute homolog were amplified from a cDNA library (Langeland et al., 1998) constructed using the pBluescript vector. Vector-specific primer pBluescript-NotI (5′-CTTGCGGCCGCTTCACTATAGGGCGAATTGGGTACC-3′) and AS1R-EcoRI were used to amplify the 5′ part of the cDNA, and primers pBluescript-NotI and AS1F-BamHI were used to amplify the 3′ part. The amplicon of the 5′ and 3′ parts of the Achaete-scute homolog were cloned into the pBluescript vector and sequenced to confirm identity.

The plasmid containing the 3′ domain of the Achaete-scute homolog was linearized by BamHI digestion and used to synthesize a digoxigenin (DIG)-labeled antisense riboprobe with T3 RNA polymerase. Full-length cDNAs of AmphiElav/Hu, AmphiDelta and AmphiSoxB1a were isolated from an EST library, and the PCR products of the insert cDNA were used as templates for synthesizing DIG-labeled antisense riboprobes with T7 RNA polymerase (Yu et al., 2008a; Yu et al., 2007). For the synthesis of the AmphiTlx antisense riboprobe, the primer pDONR222F-SP6 (5′-ATTTAGGTGACACTATAGAAGACGGCCAGTCTTAAGCTC-3′) and an AmphiTlx-specific primer with a 5′ added T7 polymerase binding site (Tlx-R-T7, 5′-TAATACGACTCACTATAGGGAGACCAACGACTTGGTATGAAGC-3′) were used to amplify a 1276 bp AmphiTlx cDNA fragment from the cDNA clone bfne092g20 (Yu et al., 2008b). The procedure for in situ hybridization was performed as previously described (Yu et al., 2008a). Images were taken using a Zeiss Axio Imager A1 microscope with a Zeiss AxioCam MRc CCD camera.

Double fluorescent in situ hybridization and immunostaining of the amphioxus embryos were performed as previously described (Wu et al., 2011). For acetylated α-tubulin and actin filament (F-actin) staining, the embryos were fixed in 4% PFA-MOPS-EGTA and then stored in PBST (PBS with 0.1% Tween 20) at 4°C until needed. For pSmad1/5/8 staining, the embryos were fixed in PFA-MOPS-EGTA and then stored in 70% ethanol at –20°C. The embryos were blocked for 1 hour in blocking solution (3% BSA and 0.1% Triton X-100 in PBS) and incubated in primary antibodies against pSmad1/5/8 (Cell Signaling, 1:150) or acetylated α-tubulin (Sigma, 1:1000) and GABA (Sigma, 1:1000) in blocking solution at 4°C overnight. The immunofluorescent signals were detected after incubation with secondary antibodies (Alexa Fluor 594-conjugated goat anti-mouse antibody for acetylated α-tubulin, Alexa Fluor 488-conjugated goat anti-rabbit antibody for GABA, and Alexa Fluor 594-conjugated goat anti-rabbit antibody for pSmad1/5/8, Invitrogen, 1:400). DAPI (Invitrogen, 1 μg/ml in PBST) was used for nuclear staining, and BODIPY FL Phallacidin (Invitrogen, 1:40 in 1% BSA/PBST) was used for F-actin staining. Fluorescent images were acquired using a Leica TCS-SP5 confocal microscope.

The embryos subjected to in situ hybridization were mounted with a 100 μm spacer and orientated as ventral side up under a Zeiss Axio Imager A1 microscope. The numbers of solitary or clustered Hu/elav- or Delta-positive cells were counted in at least ten embryos for both the control and DAPT-treated embryos. Lateral images of the embryos were captured to measure the expression boundary of gene expression. The height of the embryo (H, the vertical distance from the ventral edge to the dorsal edge of the embryo) and the height of the expression boundary (h, the vertical distance from ventral edge to the most dorsal edge of the signal) were measured using the Zeiss AxioVisionLE software. The h/H ratio indicated the relative distance of signal from the base to the dorsal expression boundary. t-tests were performed using SigmaStat, and bar histograms were drawn using SigmaPlot.

We confirmed that the earliest differentiating type I receptor cells are individual cells within the ventral ectoderm during the early neurula stage that express the pan-neural marker Hu/elav (Fig. 1K,M,O). As development proceeds, these Hu/elav-positive cells are observed on the flanks of the embryos (Fig. 1N,P). Interestingly, we noticed that the Notch ligand Delta is expressed in a similar pattern in the ventral ectoderm (Rasmussen et al., 2007) and that the expression of Delta always precedes Hu/elav expression (Fig. 1X-AA,FF). This suggests that Delta/Notch-mediated lateral inhibition (Bray, 2006) might be involved in the specification of these solitary sensory neurons from the neighboring epidermal cells. To test this hypothesis, we used the γ-secretase inhibitor DAPT to block the Notch signaling pathway. The embryos treated with DAPT showed a significant increase of total Hu/elav-positive cells, and many of the Hu/elav-expressing units were composed of multiple cells (Fig. 2A-C). We also observed a significant increase of Hu/elav-expressing units (composed of either solitary or clustered Hu/elav-positive cells) in the DAPT-treated embryos (Fig. 2D), which suggests that lateral inhibition might be responsible for limiting the number of these individual sensory neurons. Additionally, when we used other classes of inhibitors that block different functional sites within γ-secretase (Morohashi et al., 2006), we observed similar phenotypes as DAPT treatment (supplementary material Fig. S1). These results indicate that the observed phenotypes are probably caused by blocking Notch signaling. We also observed that in the DAPT-treated embryos, the Delta-expressing cells were increased and present in clusters (Fig. 2E-H), suggesting that an auto-regulatory loop might function to restrict the expression of the Delta ligand to each solitary neuronal progenitor cell.

Furthermore, we followed the morphogenesis of the ESNs in the control and DAPT-treated embryos by Hu/elav expression and GABA immunostaining (Fig. 2I-K). During the late neurula stage, individual ESN progenitor cells were observed within the sub-epidermal space during their migration in the control embryos; whereas in the DAPT-treated embryos, we observed many aggregates of Hu/elav-positive cells within the sub-epidermal space (Fig. 2I, white arrowheads), suggesting that these ectopic ESN progenitor cells might migrate together as a unit. At the early larval stage when many ESNs become GABA-positive and send out axons to innervate the CNS, we observed a significant increase of clustered GABA-positive cells in the DAPT-treated larvae (Fig. 2J,K). Moreover, the axon bundles connected to each GABA-positive unit appeared to be thicker compared with those in the control (Fig. 2J, white arrows), suggesting that there are more axon tracts in each of the bundles coming from the ectopic ESNs. These data indicate that DAPT treatment not only changes the expression of the neuronal markers during morphogenesis but also affects the actual number of neurons forming at the epidermis.

In the Drosophila and vertebrate models, proneural basic helix-loop-helix (bHLH) genes, such as Achaete-scute homologs, are required for the correct differentiation of sensory neurons (Quan and Hassan, 2005). Expression of proneural bHLH genes in the neural progenitors activates Delta expression, whereas their own expression is inhibited in the adjacent Notch signal-receiving cells. We identified one amphioxus Achaete-scute homolog (Ash; sequence deposited under GenBank accession number JF779676), the expression of which was colocalized with Delta in the individual cells within the ventral ectoderm during the early neurula stage (Fig. 3A,B). DAPT treatment in the amphioxus embryos also increased the number of Ash-expressing cells (most of which also expressed Delta, as shown in Fig. 3B) and caused clustering of many Ash-expressing cells within the ventral ectoderm (Fig. 3B-E). This result is consistent with the known regulatory loop between proneural bHLH genes and Delta/Notch signaling in both Drosophila and vertebrates.

In amphioxus, BMP signaling is involved in patterning the overall dorsoventral (D/V) axis (Onai et al., 2010; Yu et al., 2007), and manipulation of BMP levels severely affects neural induction and medial-lateral patterning across the dorsal ectoderm (Yu et al., 2008a). To test whether BMP signaling levels control the formation of ESNs from the ventral ectoderm, we used recombinant zebrafish Bmp4 protein (zBMP4) to elevate BMP signal levels globally (Yu et al., 2007). We found that the embryos treated with zBMP4 at the blastula stage were ventralized without a dorsal CNS, whereas the punctate epidermal expression of both Delta and Hu/elav was expanded throughout the entire embryo (Fig. 4A). We used nuclear phospho-Smad1/5/8 (pSmad1/5/8) staining to visualize BMP signaling levels. In control embryos, we observed strong pSmad1/5/8 staining on the ventral side (Fig. 4B, upper panels, red), which is consistent with the notion that a high BMP level determines ventral structure (Onai et al., 2010; Yu et al., 2007). In the zBMP4-treated embryos, pSmad1/5/8 staining was present globally (Fig. 4B, lower panels, red), and the expression patterns of Delta and Hu/elav changed corresponding to altered BMP signaling levels across the entire dorsal-ventral axis (Fig. 4B, lower panels, punctate green; supplementary material Fig. S2A,B). To quantify further the effect of manipulation of BMP signaling on the formation of ESNs, we selected one batch of zBMP4-reated embryos that displayed a mild phenotype and a partially formed neural tube, by which we can recognize the D/V polarity of the embryos (Fig. 4C). In the control embryos, the dorsal limits of the Delta- and Hu/elav-positive cells in the epidermal ectoderm were <40% of the body height at the N1 and N2 stages, respectively, whereas in the zBMP4 treated embryos, the Delta- and Hu/elav-positive cells were expanded to >80% of the body height at the N1 and N2 stages, respectively (Fig. 4C,D). These results indicate that the level of BMP signaling controls the dorsal-ventral distribution of ESNs.

We also noticed that in our severe zBMP4-treated phenotype, the dorsal neural tube (the CNS, indicated by arrows in Fig. 4A with Hu/elav staining) was completely abolished (Fig. 4A); however, the progenitors of the ESNs remained present throughout the ectoderm. This suggests that the patterning of the amphioxus CNS and the ESNs are separate events and require different BMP signaling levels. To test this hypothesis, we applied zBMP4 treatment at a later time point (stage G5) when the CNS was already determined. After this late treatment, the formation of the CNS was mostly unaffected, as shown by the expression of the CNS marker SoxB1-a (Fig. 4E). Interestingly, the distribution of punctate Delta and Hu/elav expression in the ectoderm was still expanded dorsally after this late zBMP4 treatment (Fig. 4E,F). Additionally, combination of DAPT and zBMP4 treatments caused a dramatic increase of clustering Hu/elav-positive cells across nearly the entire epidermal ectoderm (supplementary material Fig. S3). These observations suggest that ESN patterning is an independent event that occurs later than CNS induction.

To analyze the role of BMP signaling on ESN patterning further, we used dorsomorphin, a small molecule inhibitor that blocks BMP signaling through the Smad pathway (Yui et al., 2008). Dorsomorphin treatment at the G5 stage caused a dose-dependent reduction of Hu/elav-positive ESNs in the neurula stage embryos (Fig. 4G,H). We also observed a decrease of BMP signaling levels as visualized by pSmad1/5/8 staining (Fig. 4I), indicating that BMP signaling is required for the formation of ESNs from the ventral ectoderm. Taken together, these results suggest that high BMP signaling levels might generate a putative ESN progenitor domain in the ventral ectoderm of the amphioxus embryos.

Next, we examined this putative ESN progenitor domain further and investigated the gene that might respond to the high level of BMP signaling. Through a literature search, we identified a homeobox transcription factor gene, AmphiTlx (Tlx), the expression domain of which is closely similar to the putative ESN progenitor domain (Kaltenbach et al., 2009) (Fig. 5A). At the N0-N1 stage, Tlx is expressed in the ventral ectoderm that corresponds to the area from which the dotted Delta- and Hu/elav-positive cells originate during the early neurula stage (Fig. 1L,M,X-AA; Fig. 5A); subsequently, at the N2-N3 stage, Tlx expression is gradually restricted to colocalize with the Hu/elav-positive cells (Fig. 5A,B), suggesting a role for Tlx in differentiating ESNs. We found that zBMP4 treatment shifted this ectodermal Tlx expression domain dorsally (Fig. 5C,D) and that this dorsal expansion correlated with the change in BMP signaling level as visualized by double fluorescence staining of pSmad1/5/8 and Tlx (Fig. 5E; supplementary material Fig. S2C,D). Furthermore, DAPT treatment also increased the number of Tlx-expressing cells and caused a clustering of Tlx-positive cells in the N2 stage embryos when Tlx expression was restricted to the individual developing ESNs (Fig. 5F-I). Taken together, our results suggest that Tlx is probably a downstream BMP gene that mediates ESN development in conjunction with Ash and Delta/Notch signaling. Further experiments are needed to verify this hypothesis and to link Tlx expression to the functions of Ash and Delta in selected ESN progenitor cells.

Our results demonstrate that BMP signaling is required for the formation of a putative ESN progenitor domain in the amphioxus ventral ectoderm (Fig. 6A). Within this domain, the proneural bHLH transcription factor Ash and Delta are co-expressed in the ESN progenitor cells. Treatment with DAPT and other γ-secretase inhibitors caused a significant increase in ESN formation and in the clustering distribution of these ectopic ESNs. Although we cannot exclude the possibility that these reagents might also affect other pathways that contribute to ESN specification, and further experiments will be necessary to confirm this, our data suggest that Delta/Notch-mediated lateral inhibition is probably involved in the specification of individual sensory neurons in amphioxus (Fig. 6B). Further experiments using gene-specific manipulations, such as Delta knockdown and dominant-negative or constitutively active forms of Notch pathway molecules (Pasini et al., 2006), will help to confirm this finding. Despite the described limitations, our current data are consistent with the theory that the Achaete-scute subfamily proneural bHLH genes and Delta/Notch signaling constitute a conserved genetic module to generate external sensory neurons in both protostome and deuterostome animals (Bertrand et al., 2002; Quan and Hassan, 2005). Recently, a functional study in the cnidarian sea anemone Nematostella vectensis demonstrated that one of its Achaete-scute homologs, NvashA, is also expressed in the neuronal cells within the ectoderm, and that manipulating NvashA function by morpholino knockdown or mRNA over-expression experiments affects the expression of neuronal genes (Layden et al., 2012). Together with data showing that Delta/Notch signaling is involved in the neurogenesis of certain neural cell types in Nematostella embryos (Marlow et al., 2012), these observations suggest that proneural bHLH genes and Delta/Notch signaling might represent an ancient genetic module for generating sensory neurons that evolved in the common ancestor of cnidarians and bilaterians. It is worth noting that in the sponge Amphimedon queenslandica, homologs of the proneural bHLH gene AmqbHLH1 and the Delta ligand gene are both expressed in a putative sensory cell type (globular cells) (Richards et al., 2008), although this cell type is not considered to be a bona fide neuron. Interestingly, overexpression of AmqbHLH1 in Xenopus and Drosophila embryos can induce ectopic formation of sensory neurons in these organisms (Richards et al., 2008), suggesting that proneural bHLH genes and Delta/Notch signaling might constitute an ancient cell-determination genetic module that originated in the metazoan stem group. In the eumetazoan lineage (cnidarians and bilaterians), this genetic module might have been co-opted into the genetic program for sensory neuron differentiation. In cnidarians and bilaterians, the expansion of proneural bHLH genes and the recruitment of other context-specific co-factors into this conserved genetic module might have further facilitated the spatial and temporal diversification of distinct lineages of neuronal cells within each organism during ontogenetic development. Therefore, they might have allowed for the evolutionary modifications of neuronal cell types and sensory systems in the different organisms (Layden et al., 2012; Quan and Hassan, 2005). Thus, comparing how this conserved genetic module is activated and utilized for neurogenesis in different organisms should provide important information for understanding the evolution of the nervous system.

Inhibition of BMP signaling by BMP antagonists in the dorsal ectoderm is required for amphioxus CNS formation (Onai et al., 2010; Yu et al., 2008a; Yu et al., 2007). Our work demonstrates that high BMP signaling positively regulates the developmental program of the peripheral ESNs in the ventral ectoderm and that manipulating BMP signaling levels affects the distribution of ESNs during amphioxus embryogenesis. Our data are consistent with findings from Drosophila and Platynereis which conclude that the motor neurons and interneurons in the CNS develop from the non-BMP side of the body, whereas the formation of the peripheral sensory neurons are positively regulated by BMP signaling (Denes et al., 2007; Rusten et al., 2002). This finding suggests that the function of biphasic BMP signaling in CNS and PNS development exists in the bilaterian common ancestor.

Within the chordate lineage, the patterning process of amphioxus ESNs is highly similar to the formation of the ascidian larval epidermal sensory neurons derived from the ventral epidermal midline (Pasini et al., 2006), suggesting that they might be homologous. However, whether these amphioxus ESNs are homologous to the vertebrate sensory neurons derived from the neural crest and neurogenic placodes remains a contentious issue. In vertebrates, neural crest and placode development is intimately associated with neural induction, and the level of BMP signaling plays a crucial role in establishing the medial-lateral boundaries along the dorsal ectoderm (Meulemans and Bronner-Fraser, 2004; Schlosser, 2008). It has been determined that BMP signaling is active at the neural plate border in the chick embryo (Faure et al., 2002). Moreover, BMP signaling is involved in sensory neurogenesis in epibranchial placodes (Holzschuh et al., 2005; Kriebitz et al., 2009). The amphioxus sensory cells show some similarity to the vertebrate placode-derived neurons in the following ways: (1) they both originate from the epidermis and are induced by relatively high BMP signaling levels; (2) they express the proneural bHLH homolog of Achaete-scute and differentiate into neuronal cells by Delta/Notch-mediated lateral inhibition; and (3) they can delaminate and migrate (Benito-Gutierrez et al., 2005b; Kaltenbach et al., 2009). However, there are differences between the amphioxus ESNs and the vertebrate placode-derived neurons. In vertebrates, the pre-placodal field is closely adjacent to the dorsal neural plate (Schlosser, 2008), whereas the presumptive ESN domain in amphioxus is not. Thus, it can be hypothesized that a conserved developmental module for PNS neurogenesis is activated at different locations of the body in the different chordate lineages. It is tempting to speculate that BMP signaling might function as a labile plug-in (Davidson and Erwin, 2006) and can be deployed in different spatial contexts to activate conserved neurogenesis modules for generating epidermal sensory neurons. It is possible that during early vertebrate evolution, certain embryonic ectodermal domains for generating the peripheral sensory system might have shifted more dorsally to become the pre-placodal field in vertebrates. Taken together, our results support the hypotheses that the amphioxus ESNs are possible homologs of the vertebrate placode-derived neurons and that the ventral ectoderm of the amphioxus embryo is a neurogenic region that might be homologous to the pre-placodal field of vertebrates (Holland, 2009). Further study of the regulation of BMP signaling in the ectoderm to initiate sensory neuron development in the different chordate lineages is important for determining the evolution of the chordate sensory systems.

There is evidence to suggest that the amphioxus Tlx gene is probably downstream of BMP signaling and involved in ESN development. We have observed that the initial Tlx expression in the ventral ectoderm roughly corresponds to the area from which the dotted Hu/elav-positive cells originate during the early neurula stage (Fig. 5); subsequently, Tlx expression is gradually restricted to colocalize with Hu/elav-positive cells. In addition, elevated BMP signaling level (visualized by pSmad1/5/8 staining) leads to an expansion of ectodermal Tlx expression in accordance with the ectopic ESN formation in the ectoderm (Fig. 5E). It has been shown that vertebrate Tlx homologs (Tlx1 and Tlx-3) are expressed in certain developing neurons of the CNS as well as in the cranial sensory neurons and that Tlx-3 can control the expression of some proneural bHLH genes to promote neuronal differentiation (Kondo et al., 2008; Langenau et al., 2002). We are currently carrying out gene-specific knockout experiments to examine the necessity of amphioxus Tlx function for Ash and Delta activation and the differentiation of ESNs. Additionally, it will be interesting to investigate the epistatic relationships of Tlx with other transcription factors that are also expressed in the developing ESNs, such as Coe and SoxB1c (Mazet et al., 2004; Meulemans and Bronner-Fraser, 2007). We anticipate that comprehensive comparisons such as these will facilitate our understanding of the modification of the PNS during chordate evolution and might reveal the developmental evolution that led to the emergence of complex peripheral sensory organs in vertebrates.

Finally, our data demonstrate that manipulating BMP signaling in different time windows can cause different phenotypes in the amphioxus embryos. Elevating BMP signaling level at the blastula stage abolishes the formation of the CNS and converts the entire ectoderm into the ESN progenitor domain (Fig. 4A). By contrast, elevating the BMP signaling level at the late gastrula stage (when neural induction has presumably completed) has little effect on the formation of CNS, but it still expands the distribution of peripheral ESNs to almost the entire ectoderm (Fig. 4E). These phenotypes are similar to the hemichordate ectodermal nerve net and raise the intriguing question of whether this ectodermal nerve net is homologous to the CNS or PNS. Hemichordates have a mostly diffused nervous system with dense aggregates of neurons associated with their dorsal and ventral cord (Lowe et al., 2006; Lowe et al., 2003; Nomaksteinsky et al., 2009). Recent anatomical and molecular analyses have shown that the dorsal cord of hemichordates is internalized in the collar region, which is reminiscent of the neural tube in chordates, suggesting that it might represent the CNS in hemichordates (Kaul and Stach, 2010; Nomaksteinsky et al., 2009). These new findings also suggest that the neurons of the hemichordate ectodermal nerve net might actually represent the PNS. It has been shown that in the direct developing hemichordate Saccoglossus kowalevskii embryos, the pan-neural marker Hu/elav is expressed as spots in the entire ectoderm but is stronger in the dorsal and ventral midlines. After exogenous zBMP4 protein treatment, Hu/elav is expressed at a uniformly strong level throughout the ectoderm (Lowe et al., 2006). This observation suggests that the formation of certain neurons in the S. kowalevskii ectoderm is also positively regulated by BMP signaling. We speculate that the Hu/elav-expressing cells in the dorsal ectodermal midline of S. kowalevskii are probably the ones induced by the positive BMP signaling, because the Bmp2/4 ligand is expressed in this position, and that they might be equivalent to the ESNs in amphioxus. Further molecular genetic studies on the formation of neuronal cells in hemichordates will help to resolve this issue and shed light on the evolution of the nervous system.

We thank Dr Li-Tzu Li and the personnel in the ICOB core facility for excellent technical assistance. We thank Dr Linda Holland and Dr Nicholas Holland for organizing the amphioxus collection in Tampa; and Dr Susan Bell at the University of South Florida for providing laboratory facilities.