Conserved gene regulatory module specifies lateralneural borders across bilateriansYongbin Lia,1, Di Zhaoa,1, Takeo Horieb,c,1, Geng Chena,1, Hongcun Baod, Siyu Chena, Weihong Liua, Ryoko Horiec,Tao Lianga, Biyu Donga, Qianqian Fenge, Qinghua Taoa, and Xiao Liua,2

aMinistry of Education Key Laboratory of Bioinformatics, Center for Synthetic and Systems Biology, School of Life Sciences, Tsinghua University, Beijing100084, China; bShimoda Marine Research Center, University of Tsukuba, Shimoda, Shizuoka 415-0025, Japan; cLewis-Sigler Institute for IntegrativeGenomics, Princeton University, Princeton, NJ 08544; dCollege of Life Sciences, Zhejiang University, Hangzhou 310058, China; and eCenter of BiomedicalAnalysis Platform, Tsinghua University, Beijing 100084, China

Edited by Michael S. Levine, Princeton University, Princeton, NJ, and approved June 21, 2017 (received for review March 13, 2017)

The lateral neural plate border (NPB), the neural part of thevertebrate neural border, is composed of central nervous system(CNS) progenitors and peripheral nervous system (PNS) progenitors.In invertebrates, PNS progenitors are also juxtaposed to the lateralboundary of the CNS. Whether there are conserved molecularmechanisms determining vertebrate and invertebrate lateral neuralborders remains unclear. Using single-cell-resolution gene-expressionprofiling and genetic analysis, we present evidence that orthologsof the NPB specification module specify the invertebrate lateralneural border, which is composed of CNS and PNS progenitors. First,like in vertebrates, the conserved neuroectoderm lateral borderspecifier Msx/vab-15 specifies lateral neuroblasts in Caenorhabditiselegans. Second, orthologs of the vertebrate NPB specificationmodule (Msx/vab-15, Pax3/7/pax-3, and Zic/ref-2) are significantlyenriched in worm lateral neuroblasts. In addition, like in other bilat-erians, the expression domain of Msx/vab-15 is more lateral thanthose of Pax3/7/pax-3 and Zic/ref-2 in C. elegans. Third, we show thatMsx/vab-15 regulates the development of mechanosensory neuronsderived from lateral neural progenitors in multiple invertebrate spe-cies, including C. elegans, Drosophila melanogaster, and Ciona intes-tinalis. We also identify a novel lateral neural border specifier,ZNF703/tlp-1, which functions synergistically with Msx/vab-15 inboth C. elegans and Xenopus laevis. These data suggest a commonorigin of the molecular mechanism specifying lateral neural bordersacross bilaterians.

C. elegans | neural plate border | neural border | Msx/vab-15 | ZNF703/tlp-1

The vertebrate neural border is a transient embryonic domainlocated between the neural plate and nonneurogenic ecto-

derm from late gastrulation to early neurulation. The neuralborder is composed of the lateral neural plate border (NPB) andpreplacode ectoderm (PPE) subdomains (1, 2). The NPB and PPEgive rise to the neural crest and placode, respectively, both ofwhich undergo epithelial-to-mesenchymal transition/delamination,migrate in prototypical paths, and give rise to the peripheralnervous system (PNS) and many other cell types (3, 4). However,the NPB and PPE also have many different features (5). For ex-ample, the PPE is confined to the anterior half of embryos anddoes not contribute to the central nervous system (CNS), whereasthe NPB is the lateral border of the whole neural plate and consistsnot only of progenitors for the PNS but also those for the CNS inthe dorsal neural tube. The juxtaposed localization of the CNSneuroectoderm and PNS progenitors also occurs in the trunk ofinvertebrate embryos such as nematodes, arthropods, annelids, andurochordates (6–9), reminiscent of vertebrate NPB. In Caeno-rhabditis elegans, lateral neuroblasts (P, Q, and V5 cells) are locatedbetween the embryonic CNS and skin from the birth of these cells(10). In addition, worm lateral neuroblasts possess several keycellular and developmental features of vertebrate NPB. For ex-ample, the medially located P neuroblasts display nuclear migra-tion and give rise to multiple cell types, including CNS motorneurons (10). The laterally located V5 and Q neuroblasts give rise

to PNS neurons; that is, V5 neuroblasts give rise to the mecha-nosensory neuron PVD, whereas QL and QR neuroblasts give riseto the migratory mechanosensory neurons AVM and PVM, re-spectively (10). Furthermore, Q neuroblasts undergo delaminationand migrate along prototypical paths while dividing and differen-tiating (10), similar to the neural crest in vertebrates and mecha-nosensory bipolar tail neurons (BTNs) in tunicate (3, 9). At themolecular level, Msx/vab-15 establishes the identity of the lateralboundary of the neuroectoderm in fly, annelids, amphioxus, lam-prey, and higher vertebrates. Also, their Msx/vab-15–expressingdomains include both CNS and PNS progenitors (8, 11–14). WormMsx/vab-15 expression was observed in lateral P neuroblasts, pro-genitors of larval CNS motor neurons. Furthermore, developmentof mechanosensory neurons is defective in Msx/vab-15 mutants(15). These evolutionary echoes suggest that studies in C. elegansmay reveal new clues regarding whether there is a conserved mo-lecular mechanism underlying development of worm lateralneuroblasts and vertebrate NPB.Using genetic analysis, we show here that worm Msx/vab-15

expresses in and specifies all lateral neuroblasts. We also didcomparative analysis in multiple species, including C. elegans,Drosophila melanogaster, and Ciona intestinalis, to show thatMsx/vab-15 regulates the development of mechanosensory neuronsderived from their lateral neural borders, similar to its role in thevertebrate NPB. We profiled the expression of 90% conservedtranscription factors in worm trunk ectoderm at the resolution ofsingle cells and found that orthologs of the vertebrate NPBspecification module are significantly enriched in worm lateral

Significance

The lateral neural plate border (NPB) gives rise to the neuralcrest, one of the precursors of the peripheral nervous system(PNS) and generally considered an evolutionary innovation ofthe vertebrate lineage. Recently, it has been reported that arudimentary neural crest exists in protovertebrate Ciona, butwhether this is true in other invertebrates and there is con-served molecular machinery specifying the NPB lineage areunknown. We present evidence that orthologs of the NPBspecification module specify lateral neural progenitor cells inseveral invertebrates, including worm, fly, and tunicate. Wepropose that an ancient lateral neuroblast gene regulatorymodule was coopted by chordates during the evolution ofPNS progenitors.

Author contributions: Y.L., D.Z., T.H., and G.C. designed research; S.C., W.L., R.H., and B.D.performed research; H.B., T.L., and Q.F. analyzed data; and Q.T. and X.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Y.L., D.Z., T.H., and G.C. contributed equally to this work.2To whom correspondence should be addressed. Email: xiaoliu@mail.tsinghua.edu.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1704194114/-/DCSupplemental.

E6352–E6360 | PNAS | Published online July 17, 2017 www.pnas.org/cgi/doi/10.1073/pnas.1704194114

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

neuroblasts. More interestingly, a novel NPB specifier, ZNF703/tlp-1, was identified and shown to be synergistic with Msx/vab-15in regulating lateral neural progenitors in both C. elegans andXenopus laevis, providing another molecular clue to trace theorigin and evolution of the lateral neural border.

ResultsMsx/vab-15 Specifies Lateral Neuroblasts in C. elegans. A previousstudy in C. elegans did not detectMsx/vab-15 reporter expression inV5 and Q neuroblasts (15). To reveal the expression pattern ofendogenous Msx/vab-15, we generated a knockin worm strain inwhich GFP was integrated into the 3′ end of theMsx/vab-15–codingsequence in the genome. Worms of the knockin strain do not showany detectable phenotype, suggesting the GFP knockin does notdisturb the expression of vab-15. The knockin strain shows signif-icantMsx/vab-15 expression in embryonic AB.p(lr)apapaa, which isthe mother cell of Q and V5 neuroblasts (Fig. 1A). After the di-vision of this mother cell, Msx/vab-15 expression turns off in

V5 neuroblasts but is maintained in Q cells until they divide (Fig. 1B and C). P-neuroblast expression of Msx/vab-15 starts beforehatching, is maintained while P nuclei migrate into the ventralmidline in early L1-stage larvae, and shuts down after P neuroblastsdivide (Fig. 1 B and D).To investigate the function of Msx/vab-15 during worm de-

velopment, we phenotyped Msx/vab-15 mutants. These mutantsappear to have wild-type embryonic cell lineage and producelateral neuroblasts that can even divide in larvae (Figs. 2 and 3Band Fig. S1), but their P neuroblasts are significantly defective innuclear migration, and those that migrate into the midline fre-quently fail to give rise to hypodermal daughter cells or neurons(Fig. 2 and Table 1). Q neuroblasts in mutants show a defect inmigration too (Fig. 3B), and fail to give rise to mechanosensoryneurons with nearly full penetrance (Fig. 3 D and F and Table 1).Similarly, V5 neuroblasts in mutants also fail to give rise toneurons (Fig. 3H).V5 neuroblasts show a strong phenotype in Msx/vab-15 mu-

tants even though Msx/vab-15 expression shuts down right afterthe separation from its sister cell Q in late embryos, indicatingMsx/vab-15 specifies the cell fate of these neuroblasts at theembryo stage, instead of functioning at the larval stage. To dis-sect the temporal role of Msx/vab-15, we generated an Msx/vab-15somatic knockout strain in which CAS9 is driven by a heat-shock promoter. Disrupting Msx/vab-15 at an early stage by heatshocking embryos results in a significant migration and differ-entiation defect of the lateral neuroblasts. On the contrary,disrupting Msx/vab-15 in newly hatched L1 larvae does not leadto a significant differentiation phenotype in V5 and Q neuro-blasts (Table 1). Similarly, knocking out Msx/vab-15 at the earlyL1 stage does not affect the development of P neuroblasts, either(Table 1). In summary, our endogenous gene-expression analysisand temporal knockout assay show that Msx/vab-15 specifies thelateral neuroblasts at their initial formation and regulates thedevelopment of both the CNS and PNS.

Msx/vab-15 Regulates Development of PNS Derived from LateralNeuroblasts in Fly. As in nematodes and vertebrates, Msx/vab-15in D. melanogaster is also expressed in the lateral neural borderencompassing both CNS and PNS progenitors (7, 16). Further-more, fly Msx/vab-15 is a specifier of lateral CNS progenitors (7,16). To test whether Msx/vab-15 also regulates the PNS in fly, wegenerated an Msx/vab-15 mutant, msh4-4, in D. melanogaster andfocused on the lateral clusters of the mechanosensors, the chor-dotonal organs. In wild-type Drosophila embryos, each chordotonalorgan is made of four cells derived from a single sense organprecursor (SOP) cell, including a neuron, glial cell, and two sup-porting cells. During PNS development, five SOPs for the chor-dotonal organs are formed in the Msx/vab-15–expressing domainlateral to the neuroectoderm in each segment (7). The progenycells of these five SOPs migrate dorsally and give rise to a stretch offive characteristically arranged chordotonal organs (17). However,in msh4-4 mutants, the chordotonal organs contained fewer neu-rons and glial cells (Fig. 4A). Furthermore, the migration of theprogeny cells derived from the SOPs in msh4-4 embryos appearedto be significantly desynchronized among different segments, asindicated by the position of the glial cells, whereas in the wild-typeembryos these cells migrated to a similar position in all segments(Fig. 4B). So, like its role in nematodes and vertebrates,Msx/vab-15in fly regulates PNS development in the lateral neural border.

Systematic Prediction of Lateral Neuroblast Regulators in C. elegans.To comprehensively reveal regulators that distinguish lateralneuroblasts from trunk CNS and the skin cells flanking them, wegenerated reporter transgenic lines for 339 out of all 377 wormtranscription factors that have human orthologs (18) (DatasetS1). Seventy-five percent of our cloned promoters are more than2-kb-long upstream regions. Other promoters are less than 2 kb,

Fig. 1. Expression ofMsx/vab-15 in the trunk nervous systems of bilaterians.(A–D) Lateral view of Msx/vab-15::GFP knockin comma-stage embryo (A),newly hatched L1 larvae (B), middle L1 (C), and late L1 (D). The red channelrepresents Histone1::mCherry knockin to label nuclei universally. Arrowsindicate the mother cell of Q and V5 (A), Q neuroblast (B), or Q daughtercells (C). Arrowheads indicate P neuroblasts (B and C) and their daughterand granddaughter cells (D). In middle L1 stage, P nuclei have arrived at theventral midline (C). Anterior end of larvae is to the right, and ventral side isup. (Scale bars, 10 μm.) (E) The schematic drawing of neurulae represents thedorsal view of vertebrates and the ventral view of protostomes. The headregion in each animal lies above the dotted arc. Only in the ectoderm ofC. elegans are neuroblasts drawn according to their precise numbers: a pairof PNS progenitors [AB.p(lr)apapaa, the mother cell of V5 and Q neuro-blasts]; six pairs of P neuroblasts; and nine progenitors of ventral cord motorneurons in the midline. The Msx/vab-15–expressing domain in the annelid isbased on Platynereis dumerilii (8). The Msx/vab-15–expressing domain in flyis based on D. melanogaster (7). The vertebrate Msx/vab-15–expressing do-main represents the merged expression patterns of Msx1, Msx2, and Msx3 inmouse (34, 35).

Li et al. PNAS | Published online July 17, 2017 | E6353

DEV

ELOPM

ENTA

LBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

mostly because their upstream intergenic regions are short. It hasbeen demonstrated that 2 kb upstream of a worm gene accu-rately drives gene expression in most cases (19, 20). Therefore,the majority of our reporters should largely recapitulate endog-enous gene expression. We profiled their expression in trunkectoderm of newly hatched L1 larvae at the resolution of singlecells using an imaging analysis pipeline (19) (Fig. 5). The an-notation of L1 trunk ectodermal cells is highly reliable becauseof their well-separated and invariable position and characteristicnuclear morphology (10). All cell annotations were manuallyverified to be precise.

To predict neuroblast regulators, we defined a neuroblast-enrichment score as the logarithmic ratio of expression in neu-roblasts over expression in skin or ventral cord motor neurons(see Materials and Methods for details). To test the predictionpower of our neuroblast-enrichment score, we functionally vali-dated seven top-scored genes: ZNF703/tlp-1, Msx/vab-15, Atonal/lin-32, Zic/ref-2, Znf/ztf-6, Ash/hlh-3, and Pax3/7/pax-3 (Fig. 5). Inaddition to Msx/vab-15 described in this study, Atonal/lin-32 is awell-known master regulator of Q and V5 neuroblast develop-ment (21, 22). Our genetic analysis showed that Pax3/7/pax-3 andZic/ref-2 are required for both migration and differentiation of P

Fig. 2. Msx/vab-15, Pax3/7/pax-3, and Zic/ref-2 are required for migration and differentiation of P neuroblasts. Somatic knockout was induced by heatshocking embryos carrying the Phsp::CAS9 transgene. The Pref-2::H1::mCherry reporter was used as a marker to label P-neuroblast progeny. The Punc-25::GFPreporter was used to label D-type ventral cord motor neurons. VDs are P-derived D-type neurons. (A, E, I, and M) Wild-type L2 larvae, when all P progeniesarrive at the ventral midline (A) and give rise to hypodermal Pn.p (E) and neurons (I and M) distributed with stereotypical patterns. Arrowheads indicateprogenies of P neuroblasts that are defective in migration, as shown retained in the lateral body. In mutant worms, some P neuroblasts cannot migrate to theventral cord and retain in the middle body of worm (B–D). Among P neuroblasts that successfully migrated into the ventral cord, some failed to give rise tohypodermal Pn.p (F–H), neurons (J–L), such as VD neurons (N–P). Dotted circles indicate missing nuclei. Each worm was arranged such that its anterior end wasto the right and its ventral midline was at the top. Green fluorescence in pharynx represents the Pmyo-2::GFP coinjection marker (K, L, O, and P). All of theworms detected were at L2 stage. (Scale bar, 10 μm.)

E6354 | www.pnas.org/cgi/doi/10.1073/pnas.1704194114 Li et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

nuclei (Fig. 2 and Table 1). However, mutants of the other threecandidates (ZNF703/tlp-1, Ash/hlh-3, and Znf/ztf-6) show a weakto marginal phenotype (Table 1). ZNF703/tlp-1 is expressed bothin P neuroblasts and the mother cells of V5 and Q, similar toMsx/vab-15 (Figs. 5 and 6A), so we further investigated the ge-netic interaction between ZNF703/tlp-1 and Msx/vab-15. Al-though P neuroblasts appear, there are no phenotypes inZNF703/tlp-1 single mutants. But P neuroblasts in ZNF703/tlp-1;Msx/vab-15 double mutants show a stronger defect in migrationand differentiation than those in Msx/vab-15 single mutants (Fig.6 B and C). Q neuroblasts display little migration defect inZNF703/tlp-1 single mutants, whereas ZNF703/tlp-1;Msx/vab-15double mutants showed a stronger phenotype than Msx/vab-15single mutants (Fig. 6D). Significant synergistic effect was alsoobserved upon V5 differentiation (Fig. 6E). Therefore, ZNF703/tlp-1 is a novel neuroblast regulator that functions as a modulatorof Msx/vab-15 activity in worms. These data demonstrated thatthe neuroblast-enrichment score is strongly related to the mo-lecular mechanisms specifying worm lateral neural precursorcells. Then, examining the correlation between the wormneuroblast-enrichment score and orthologs of the NPB specifi-cation module can suggest the conservation between the wormlateral neuroblasts and the vertebrate NPB.

The NPB Specification Module Is Conserved Between Worm andChordates. A conserved four-gene NPB specification module

(Msx/vab-15, Pax3/7/pax-3, Zic/ref-2, and AP2/aptf-1) estab-lishes the identity of the NPB from jawless fish lamprey tomouse (11, 23). This gene set has neuroblast-enrichment scoressignificantly higher than those of other genes (Mann–Whitney testbased on neuroblast-enrichment score, P value < 1.8 × 10−4).Actually, three out of these four genes are among the top-sevenneuroblast-enriched genes (Fig. 5). Although the expression pat-terns of these three NPB specifiers (Msx/vab-15, Pax3/7/pax-3, andZic/ref-2) are overlapping, the expression domain ofMsx/vab-15 inmany bilaterians expands more laterally than those of Pax3/7/pax-3and Zic/ref-2 (8, 24). Consistently, onlyMsx/vab-15, but not Pax3/7/pax-3 and Zic/ref-2, is expressed and plays a critical role in themost lateral V5 and Q neuroblasts, which are PNS progenitors inworm (Fig. 1 and Table 1). Another example of more lateral Msx/vab-15 expression exists in the urochordate C. intestinalis. TwoNPB cells (b8.18 and b8.21) give rise to BTNs, which are likely thecounterparts to vertebrate dorsal root ganglion (DRG) mecha-nosensory neurons (9). These two NPB cells express Msx/vab-15and Pax3/7/pax-3 but not Zic/ref-2 (9, 25). Nevertheless, our gene-knockdown experiments showed that Msx/vab-15 is required forBTN differentiation (Fig. 4C). Therefore, Msx/vab-15 regulatesthe development of the PNS from the lateral neural border in bothProtostomia and Deuterostomia, regardless of the coexistence ofother NPB specifiers such as Zic/ref-2.

ZNF703/tlp-1 Is a Novel NPB Specifier in X. laevis. We have identifiedZNF703/tlp-1 as a novel regulator of lateral neuroblasts cooper-ating with Msx/vab-15 in C. elegans. To investigate whetherZNF703/tlp-1 has conserved function in vertebrates, we examinedits expression and function in X. laevis. In situ RNA hybridizationshowed that ZNF703/tlp-1 is expressed at the posterior ectodermexcept for the dorsal midline in frog gastrulae. During neurulation,the expression of ZNF703/tlp-1 in nonneurogenic lateral ectodermdiminishes so as to become enriched in the NPB and neural crest(Fig. 6F). As in C. elegans, disturbing ZNF703/tlp-1 alone has nodetectable phenotype. But moderate knockdown of both ZNF703/tlp-1 and Msx/vab-15 significantly down-regulates the trunk ex-pression of the neural crest specifier FoxD3, migratory DRG progen-itor marker islet1, and nonmigratory Rohon–Beard mechanosensoryneuron progenitor marker Runx1 (Fig. 6G), with no detectableeffect on cell proliferation or apoptosis (Fig. S2). High dosageMsx/vab-15 morpholino blocks the differentiation of NPB inXenopus, while low dosage does not (Fig. 6G and Fig. S3).Hence, ZNF703/tlp-1 acts as a modulator of Msx/vab-15 in bothworm lateral neuroblasts and the vertebrate NPB, supporting ourhypothesis on the conservation of the gene regulatory modulespecifying the lateral neural border across bilaterians.

DiscussionAn important feature of Bilateria is the centralized nervoussystem, in which the CNS in the head and trunk midline inte-grates and processes sensory information from the PNS. De-velopmental processes to generate centralized nervous systemsdiffer dramatically among different bilaterian phyla. In nema-todes such as C. elegans, cell fate is coupled with cell lineage andneurogenesis is multiclonal; that is, many independent pro-genitor cells give rise to neurons (6, 10). However, in vertebrategastrulae, accompanying the determination of the dorsoventralaxis by BMP signaling, ectoderm in the dorsal midline thickens toform a neural plate, from which their CNS is derived. Betweenthe neural plate and nonneurogenic ectoderm are the NPB andPPE, from which the majority of the vertebrate PNS is derived(24). How nervous system centralization patterns come into be-ing is a long-standing question in neural development and evo-lution (3, 26).Regarding the molecular mechanisms determining CNS de-

velopment, comparative studies have provided evidence that theCNS of vertebrates, annelids, and arthropods originated from a

Fig. 3. Msx/vab-15 is required for the development of worm lateral PNSprogenitors. The boundary of the worm body is delineated by dashed lines.(A and B) Migration of Q neuroblast progeny cells. Dotted circles indicate Qdaughter or granddaughter cells. Examined worms were at late L1 stage. Qdaughters were separating due to their different migration rate in wild type(A), whereas they failed in migration in the vab-15 mutant so that even theirgranddaughters are clustered together (B). (C–F) Q-derived mechanosensoryneurons AVM (arrow in C) and PVM (arrow in E) were missed in the vab-15mutant (D and F). Examined were L2 larvae. Arrowheads indicate ALM(C and D) or PLM (E and F) neurons as a landmark. (G and H) V5-derivedmechanosensory neuron PVD (arrow). egl-17: markers of Q neuroblasts andtheir progeny (47). mec-4: markers of AVM/PVM (58). mec-10: markers ofAVM/PVM and PVD (59). Arrowheads indicate the PLM neuron as a land-mark. Examined were L3 larvae. (Scale bars, 10 μm.)

Li et al. PNAS | Published online July 17, 2017 | E6355

DEV

ELOPM

ENTA

LBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

common bilaterian ancestor (8, 27). For example, both theneuroectoderm of annelids and neural plate of vertebrates aresubdivided into a sim/hlh-34–expressing midline and longitudinalNk2.2/ceh-22–, Nkx6/cog-1–, Pax6/vab-3–, Pax3/7/pax-3–, and Msx/vab-15–expressing domains from medial to lateral, likely repre-senting the molecular architecture of trunk CNS in their lastcommon ancestor, Urbilateria (8). However, there has been de-bate on the common origin of CNS centralization because somesister or outgroup species of annelids, arthropods, and chordateslack a centralized ventral cord (26, 28). For example, hemichor-dates differ from major chordates by possessing a diffuse nervoussystem and do not have mediolateral neural patterning (29).Characterization of the molecular mechanisms of neural pattern-ing in distantly related bilaterians should provide a broader view ofthe evolution of nervous system centralization (28).Nematodes diverged from arthropods about 550 Mya, close to

the origin of Ecdysozoa (30). Unlike annelids and arthropods,nematodes have an unsegmented body plan and no appendages.C. elegans burrows in rotten matter and has a feeding organisolated from the rest of the animal (31, 32). Correspondingly, itsnervous system is more modestly organized than those in anne-lids and arthropods. The complex mediolateral patterning oftrunk CNS is largely absent in worm, where there are only motorneurons in its ventral cord. Nevertheless, Pax3/7/pax-3 and Msx/vab-15 still express in the lateral border of the CNS neuro-ectoderm in worm (15, 33), consistent with the conserved ex-pression patterns of these two lateral genes in arthropods,annelids, and chordates (8). Our genetic analysis shows that bothMsx/vab-15 and Pax3/7/pax-3 specify P neuroblasts, which areworm CNS progenitors, similar to their roles in the lateral CNSprogenitors in fly and vertebrates (11, 16, 34–36). Combining ourdata with the existing evidence of the common origin of trunkCNS from fly to vertebrates (8, 27), it is more likely that nem-atodes inherited the Msx/vab-15 specification of P neuroblastsfrom Urbilateria than the nematode branch lost the conservedlateral domain after diverging from fly and then regeneratedit independently.In addition to the conserved CNS molecular module shown in

C. elegans, our data suggest that the molecular mechanism de-fining the lateral neural border, from which the majority of thePNS is derived, is also conserved across bilaterians. First, the

lateral neural borders in nematodes, fly, annelids, and urochor-dates show conserved neural patterning where the medial por-tion of the Msx/vab-15–expressing domain gives rise to the CNSwhereas the lateral portion generates the PNS. Second, theirlateral PNS progenitors exhibit several key features of vertebratePNS progenitors, such as migration and differentiation intomechanosensory neurons (7, 8, 10, 13, 34). Our genetic analysisalso shows that Msx/vab-15 specifies these PNS progenitors innematodes, fly (Fig. S4), and urochordates, similar to its role inthe vertebrate NPB and PPE. The PPE is the neurogenic ecto-derm anterolateral to the NPB, and is considered the epidermalsubdomain of the vertebrate neural border that mostly contrib-utes to cranial sensory systems (1, 37). The PPE also gives rise tomechanosensing hair cells of lateral lines in the trunk of anam-niote vertebrates and those in vertebrate inner ears (1). Thedevelopment of the mechanosensing hair cells in vertebrates andthat of the mechanosensory neurons in worm and fly both de-pend on Atoh1/lin-32 (22, 38). On the contrary, Ciona BTNs andvertebrate DRG neurons require Neurogenin/ngn-1 (9). Hence,more molecular and comparative studies are needed to un-derstand the evolutionary relationship of these mechanosensingcell types. Nevertheless, our study on the neural borders suggeststhe conservation of the molecular mechanism underlying thespecification of PNS progenitors across bilaterians (39).In theory, the similarity between the invertebrate and verte-

brate lateral neural border can be explained by either commonorigin or convergence of independent evolution. To distinguishbetween these two hypotheses, it is necessary to compare mo-lecular mechanisms specifying lateral neural borders in variousbilaterians. Unlike the vertebrate NPB and neural crest, themolecular mechanism underlying the specification of wormlateral neuroblasts still remains largely unknown after years ofgenetic study (33, 40, 41). Our systematic expression profilingof 90% conserved transcription factors in C. elegans revealedthat three NPB specifiers (Msx/vab-15, Pax3/7/pax-3, and Zic/ref-2) are coexpressed in worm lateral neuroblasts and thatMsx/vab-15 expression is more lateral than that of Pax3/7/pax-3 andZic/ref-2, like their expression patterns in neural borders ofvertebrates, urochordates, amphioxus, and likely annelids (7, 8,11, 13, 24). Furthermore, we showed that Msx/vab-15 regulatesthe differentiation of PNS progenitors in worm and fly, similar

Table 1. Genetic analysis of migration and differentiation of lateral neuroblasts

Gene Allele

Frequency of wormswith neuroblast

migration defect, % (n) Frequency of worms with neuroblast differentiation defect, % (n)

Q*,† P‡Missing PDE§,

(V5.paaa#)Missing AVM†,

(QR.paa#)Missing PVM†,

(QL.paa#)Not 13 VDs¶,

(Pn.app#)

Wild type 0 (100) 0 (30) 1 (105) 0 (100) 0 (100) 0.5 (200)Msh/vab-15 tm260 34 (44) 81 (101) 58 (112) 96 (104) 96 (101) 100 (44)

Knockout in embryojj 19 (32) 46.46 (99) 30 (50) 28 (50) 16 (50) 46.46 (99)Knockout in L1jj 13 (23) 0 (27) 6 (50) 4 (25) 8 (50) 0 (27)

NLZ/tlp-1 bx85 0 (50) 0 (55) 2 (100) 13 (101) 4 (104) 0 (101)ZNF/ztf-6 tm1803 0 (100) 0 (100) 2 (42) 0 (100) 0 (100) 0 (100)Ash/hlh-3 tm1688 0 (100) 0 (100) 0 (100) 0 (100) 0 (100) 3.6 (280)Zic/ref-2 Knockout in embryojj 0 (42) 54 (63) 0 (50) 0 (42) 0 (42) 54 (63)pax3/7/pax-3 Knockout in embryojj 0 (22) 41.7 (60) 4 (50) 0 (22) 0 (22) 45 (60)

The number of scored animals in each experiment is in parentheses.*A late-stage larva is considered abnormal if either the AVM or PVM is mislocated.†mec-4 reporter as an AVM/PVM neuron marker.‡ref-2 reporter as a marker of P neuroblasts and their progeny.§dat-1 reporter as a PDE marker.¶unc-25 reporter as a VD neuron marker.#Cell lineage of the specified neuron.jjSomatic knockout is induced by heat shock activating Phsp::CAS9 in a specified stage.

E6356 | www.pnas.org/cgi/doi/10.1073/pnas.1704194114 Li et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

to its critical role in specifying PNS lineage in urochordates andvertebrates. We noticed that in PNS specification, only the roleof Msx/vab-15 is conserved in worm, but not the other two NPBspecifiers, Pax3/7/pax-3 and Zic/ref-2. This is not totally sur-prising because there have been indications that the expressionand function of Pax3/7/pax-3 and Zic/ref-2 are not strictly con-served in the evolution of PNS development. For example,whereas Pax3/7/pax-3 is expressed in the lateral domain inboth annelids and nematodes (8, 33), it is active in transversestripes in fly (27). In addition, Pax3/7/pax-3 is only marginallyexpressed in the putative PNS progenitor region in annelids (8).Their expression even varies within vertebrates. Pax3/7/pax-3and Zic/ref-2 expression domains are significantly more medialin lamprey (11) than those in zebrafish (42). Also, in Ciona, theNPB cells that give rise to mechanosensory BTNs do not ex-press Zic/ref-2 (25). Therefore, our data, consistent with otherstudies, showed that only Msx/vab-15, but not other NPBspecifiers, exhibits conserved expression and a functional rolein PNS progenitors across bilaterians (3, 43). Finally, we haveidentified a novel lateral neural border specifier, ZNF703/tlp-1,

and demonstrated that it functions synergistically with Msx/vab-15in regulating the migration and differentiation of lateral neuralprogenitor cells in both C. elegans and Xenopus. Altogether, ourdata revealed deep conservation of vertebrate NPB specifiers’module composition, expression pattern, and regulatory role inneural border development among distantly related species withdrastically different neural anatomy, favoring the common ori-gin hypothesis in the molecular mechanism underlying neuralborder specification.

Materials and MethodsC. elegans-Related Procedures.Orthologous gene pairs.Orthologous gene pairs were obtained from a previouspublication (18) or annotation in WormBase (www.wormbase.org) or TreeFam(www.treefam.org). A 1:N ortholog relationship is allowed.C. elegans strains and reporter transgenes. Mutant strains were obtained fromthe Caenorhabditis Genetics Center or National BioResource Project ofJapan. vab-15mutants: TU2362(u781), FX0260(tm260); tlp-1mutant: EM347(bx85);lin-32 mutant: FXO1446(tm1446); dat-1 reporter: BZ555(Pdat-1::GFP). Promoterreporter constructs and their transgenic strains were generated as described(19, 44), and their description can be found in Dataset S1. vab-15::GFP knockinstrain was generated using the CRISPR/Cas9 system as described (45). mec-10reporter (Pmec-10::H1::mCherry) was generated through mosSCI insertion (46).Primer sequences of mec-10 promoter: 5′-accgaaagatgcacgtccta-3′ and5′-ATATTGTGGTCCAGCTCTCACACTG-3′.Somatic knockout. Somatic KO was conducted as described (47). Guide RNA se-quences: vab-15, 5′-GTGAATGTTGCATAGACAG-3′; pax-3, 5′-GTCGTGTTAAC-CAACTCGG-3′; ref-2, 5′-CTTGGGCTGGAAGAAAGTG-3′. These sequences werecloned into the vector pOG2306 (47). Each vector was injected into wormsmixed with the Pmyo-2::GFP coinjection marker. Synchronized embryos oftransgenic strain were transferred to OP50-seeded nematode growthmedium (NGM) plates and heat shocked as described, at 33 °C for 1 h (47).The embryos then grew at 20 °C until L2 stage.Imaging protocol. Larvae that grew for no more than 4 h after hatching wereconsidered to be at early L1 stage. Larvae whose neuronal lineage P neu-roblasts finished cell division and whose gonad was composed of about50 cells were considered to be at L2 stage. For live imaging, worms wereplaced on a 2% agarose pad and immobilized using levamisole (2.5 mg/mL).Images were captured using a Zeiss Imager A2 epifluorescence microscope.For 3D imaging, collected worms were frozen in liquid nitrogen for a night.Frozen worms were thawed in 200 μL acetone. Then, the acetone was dis-carded and the worms were fixed with a solution composed of 1× modifiedRuvkun’s witches brew and 20% formalin for 1.5 h, and then proceeded toDAPI staining and 60% glycerol mounting as described (19). Three-dimensional images were obtained with a Zeiss LSM 780 confocal micro-scope. The pixel size was 0.132 μm in the x-y plane and 0.300 μm in thez direction.Cell annotation of 3D image stacks of L1 larvae. The 3D image stacks of wormswere straightened computationally along the anterior–posterior axis. Theircell nuclei were segmented and annotated, and then manually edited withthe VANO interactive interface as described (19). The identities of hyp7hypodermal nuclei, seam cells, ventral cord motor neurons, neuroblasts, andtheir progeny’s nuclei were manually confirmed according to WormAtlas(www.wormatlas.org) (10).Gene-expression measurement. The gene-expression level for every cell wasmeasured as described (19). Briefly, background fluorescence was estimatedusing 10 pseudonuclei. After subtracting background fluorescence, mCherryfluorescence was normalized by DAPI fluorescence to account for sphericalaberration. A nucleus with a normalized mCherry fluorescence of 500 wasoften barely distinguishable from background red fluorescence according tovisual inspection. To become robust to background, gene expression wascalculated as (normalized mCherry fluorescence + 500)/500.Neuroblast-enrichment score. For each gene, we calculated its median log2

expression level for hypodermal cell group (hyp7, V1 to V4, plus V6), lateralneuroblast group (Q, P, V5), and ventral cord motor neuron (VMN) (DatasetS2). Mean (MedianP, MedianQ) − Max (Medianhyp7, MedianV, MedianVMN)represents the neuroblast-enrichment score of a gene.Embryo-lineage tracing. vab-15 mutant strain TU2362(u781) was crossed withRW10029 (zuIs178; stIs10024) (48), which carries the his-72::GFP transgeneso that all somatic nuclei become detectable during embryogenesis.Then, one- to four-cell-stage embryos were transferred onto an agarose padfor imaging. We performed confocal imaging with an inverted Leica SP5confocal microscope. Embryo-lineage tracing and imaging were processed asdescribed (49). The u781mutant is edited up to 200 time points with 398 cells.

Fig. 4. Msx/vab-15 is required for the development of worm lateral PNSprogenitors in other invertebrates. (A and B) Disrupted chordotonal organs inMsx/vab-15/msh mutant Drosophila embryos. Shown are lateral views of wild-type and msh4-4 embryos at early stage 12 with the anterior side at the leftand ventral side at the bottom. (A and B) Neurons in the chordotonal organsare revealed by 22C10 immunostaining (A) and glial cells by Repo staining (B).Pictures of whole embryos are in Fig. S4. Arrowheads indicate a five-neuronarray of chordotonal organs in wild type but disorganized with only fourneurons in msh4-4. Arrows indicate glial cells in the chordotonal organs. Themigration of five glial cells of chordotonal organs is significantlydesynchronized, and the glial cell numbers vary in different segments com-pared with the WT embryos. (Scale bar, 100 μm.) (C) Defective BTN differ-entiation in Msx/vab-15 knockdown Ciona. Examined were larvae whosedorsal sides were up. Arrows point to BTNs. (C, Insets) Enlarged areas in themorpholino-injected larva represent BTNs with the down-regulated Asicbreporter, a marker of BTNmechanosensory neurons. Fractions of scored larvaewith down-regulated reporter activity are shown. (Scale bars, 100 μm.)

Li et al. PNAS | Published online July 17, 2017 | E6357

DEV

ELOPM

ENTA

LBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

D. melanogaster-Related Procedures.Generation of Drosophila mutant. Mshmutant was generated using the CRISPR/Cas9method (50) and guide sequence 5′-GGCTAGGACTCAGGTTGGTC-3′ target-ing the N-terminal msh coding region. Msh4-4 contains a single-nucleotideinsertion at position 185 of the ORF so that it is a frameshift allele and pro-duces a truncated MSH protein of only 67 amino acids.Immunocytochemistry and microscopy. Embryos were fixed and stained as pre-viously described (51). Rat anti-Repowas fromA. Tomlinson (Columbia University,New York). Embryos were observed using differential interference contrast op-tics with a Nikon Eclipse 80i microscope.

X. laevis-Related Procedures. Frogs were used with ethical approval from theAnimal Care and Use Committee of Tsinghua University.Xenopus whole-mount in situ hybridization. Whole-mount in situ hybridizationwas performed based on standard protocols (52, 53). cDNA sequences forZNF703, Runx1, Islet1, and Sox2 were cloned into the vector pCS107 (www.xenbase.org). Primer sequences were as follows: ZNF703: 5′-TCCatcga-tATGAACTGTTCTCCCCCTGGATCT-3′ and 5′-cacGTCGACctggtatcctagcgc-tgaagctg-3′; Runx1: 5′-tccATCGATATGGCCTCACATAGTGCCTTCCA-3′ and5′-gagcCTCGAGtcagtatggccgccatacggctt-3′; Islet1: 5′-cgcGAATTCatggga-gatatgggagaccca-3′ and 5′-gagcCTCGAGtgcctctatagggctggctaccatg-3′; andSox2: 5′-gagaGAATTCatgtacagcatgatggagaccgag-3′ and 5′-gcgcctcgagt-cacatgtgcgacagaggcagcgtgcc-3′. In situ probes were labeled with Dig-UTPusing a Promega T7 polymerase kit.Xenopus embryo manipulation and microinjection. Xenopus eggs and embryoswere obtained through standard procedures (54). Briefly, eggs harvestedfrom HCG-injected female frogs were fertilized through artificial in-semination techniques. Fertilized eggs were dejellied using 2% cysteineprepared in 0.1× Marc’s modified Ringer’s (MMR) (pH 7.9), followed bya thorough wash with 0.1× MMR. The dejellied embryos were then transferredinto 2% Ficoll in 0.5× MMR for morpholino (MO) and/or mRNA injection.

The MO sequences were as follows: Msx1: 5′-GCCATACAGAGAGATCCG-AGCTGAG-3′; ZNF703a: 5′-ACCAGGGCGCAAAGATTCCCCTTGT-3′; and ZNF703b:5′-ACCAGGGCGCGGAGATTCCCCTTAT-3′.

All MOs are purchased from Gene Tools. ZNF703 MO was a mixture of 5 ngZNF703a MO and 5 ng ZNF703b MO per embryo. Msx1 MO and/or ZNF703MOmixed with 2 ng RLDx as lineage tracer were injected into the animal-pole twoleft-side cells at the eight-cell stage. For rescue experiments, after MO injection,ZNF703 mRNA was immediately injected into the same two cells before theneedle hole cured. Beta-gal mRNA was coinjected as a lineage tracer in theZNF703 mRNA-alone group. The right-injected embryos were sorted out undera fluorescence dissecting microscope (Olympus; SZX16) and cultured to stages16 to 17. Embryos were fixed in Mops EGTA MgSO4 formaldehyde (MEMFA)solutionin for 4 h at room temperature followed by whole-mount in situ hy-bridization. ZNF703mRNA (mixedwith beta-gal)-injected embryos were stainedby red-gal before in situ hybridization.Western blotting. Western blotting was performed as described previously (55).Briefly, Xenopus embryos were lysed in TNE buffer (10 mM NaCl, 10 mMTris·HCl, 1 mM EDTA, pH 7.4) plus protease inhibitors, and pipetted severaltimes on ice. Lysates were then cleared by centrifugation at 1,000 × g for5 min. Supernatant was mixed with loading buffer and boiled for 15 minfollowed by SDS/PAGE separation. Primary antibody of HA (Roche; 1:2,000 di-lution) and primary antibody of β-actin (1:5,000 dilution) were used.

C. intestinalis-Related Procedures. C. intestinalis adults were obtained fromM-Rep. Sperm and eggs were collected by dissecting the sperm and gonad ducts.Construct. The Asicb reporter construct was designed based on the previouslypublished enhancer sequence (56).Microinjection of antisense morpholino oligonucleotide. MOs were obtained fromGene Tools. The antisense oligonucleotide sequence of the Msx MO is 5′-ATTCGTTTACTGTCATTTTTAATTT-3′. The MO was dissolved at a concentration of0.5 mM in 1 mM Tris·HCl (pH 8.0), 0.1 mM EDTA, 1 mg/mL tetramethylrhodamine

Fig. 5. Expression profile of conserved transcription factors in the trunk ectoderm of L1 larvae at single-cell resolution. The data are in Dataset S2. Cells weremanually arranged according to their cell types. Genes are ranked according to their neuroblast-enrichment score (see Materials and Methods) in descendingorder. (Upper) Expression profile of the top-seven neuroblast-enriched genes, which are barely expressed in skin cells and ventral cord motor neurons.Orthologs of vertebrate NPB specifiers are underlined.

E6358 | www.pnas.org/cgi/doi/10.1073/pnas.1704194114 Li et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

dextran (D-1817; Invitrogen). Microinjection of MO was performed as describedpreviously (57).Image acquisition. Imagesof larvaewere capturedusingaZeissAX10epifluorescencemicroscope.

Statistics. Wilcoxon and Mann–Whitney tests, robust statistics tests with noassumption on the distribution of samples, were used to examine whethera set of worm genes orthologous to a vertebrate gene module had sig-nificantly higher neuroblast-enrichment or mechano-enrichment scores. Ingenetics and gene-knockdown analyses, Fisher’s exact test was used under

the assumption that the fraction of animals with a phenotype followedbinomial distribution, and a t test was used when a quantitative pheno-type was scored under the assumption that the quantitative phenotypefollowed normal distribution. Variation was calculated based on binomialor normal distribution, and the similarity of variations was not assumed todecrease the false positive rate. The minimum number of scored animalswas 22 in one batch analyzed, whereas the maximum was 300.

ACKNOWLEDGMENTS. We thank Haidong Li from the Center of Biomedi-cal Analysis, Tsinghua University for confocal imaging, and Lei Qu and

Fig. 6. ZNF703/tlp-1 is a novel regulator functioning as an Msx/vab-15 modulator both in worm lateral neuroblasts and vertebrate NPB. (A) Embryonicexpression of worm ZNF703/tlp-1. A comma-stage embryo carrying the Ptlp-1::H1::mCherry reporter trangene is shown. Arrowheads indicate P neuroblasts.Arrow indicates the mother cell of V5 and Q neuroblasts. (Scale bar, 10 μm.) (B–E) Phenotypes of ZNF703/tlp-1 and Msx/vab-15 mutant worms of migrationand differentiation of lateral neuroblasts. The y axis represents the number of VD neurons recognized by the Punc-25::GFP marker (B) or the fraction ofworms showing the specified defect (C–E). The number of scored worms is shown. (F) The expression of ZNF703 in Xenopus embryos from stage (st)12 to st19.(Scale bar, 0.5 mm.) The red dot lines represent the location of cross section. (G) Whole-mount in situ hybridization assay in Xenopus embryos (st16 to st17) toexamine the effect of MO knockdown of Msx/vab-15 and/or ZNF703/tlp-1 on the expression of the neural plate marker Sox2, neural crest specifier FoxD3, andRohon–Beard sensory neuron progenitor marker Runx1 and DRG progenitor marker Islet1. Asterisks indicate the injected side of an embryo. The red lines andgreen lines indicate the length of left and right parts of neural tubes, respectively. # indicates mRNA resistant to the ZNF703 MO. Fractions of scored embryoswith phenotype are shown. (Scale bar, 0.5 mm.) Error bars represent standard errors.

Li et al. PNAS | Published online July 17, 2017 | E6359

DEV

ELOPM

ENTA

LBIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020

Hanchuan Peng for processing imaging data. Drs. Cecilia Mello and RobertWaterston edited the manuscript. Dr. Dionne Vafeados provided transgenicworms. We are very grateful to Drs. Martin Chalfie, Andy Fire, David Kingsley,Chris Amemiya, Xiaohang Yang, Ding Xue, Zhirong Bao, Margaret Fuller,and Joanna Wysocka for helpful discussions and comments on the manu-

script. This work was supported by Ministry of Science and Technologyof China Grant 20131970194; National Natural Science Foundation ofChina Grants 20141300429 and 91519326; Tsinghua 985 funding; Grant-in-Aid for Scientific Research from JSPS Grant 24687008; and Pre-Strategic Ini-tiatives, University of Tsukuba.

1. Schlosser G, Patthey C, Shimeld SM (2014) The evolutionary history of vertebratecranial placodes II. Evolution of ectodermal patterning. Dev Biol 389:98–119.

2. Medeiros DM (2013) The evolution of the neural crest: New perspectives from lam-prey and invertebrate neural crest-like cells. Wiley Interdiscip Rev Dev Biol 2:1–15.

3. Green SA, Simoes-Costa M, Bronner ME (2015) Evolution of vertebrates as viewedfrom the crest. Nature 520:474–482.

4. Milet C, Monsoro-Burq AH (2012) Neural crest induction at the neural plate border invertebrates. Dev Biol 366:22–33.

5. Schlosser G (2008) Do vertebrate neural crest and cranial placodes have a commonevolutionary origin? BioEssays 30:659–672.

6. Sulston JE, Schierenberg E, White JG, Thomson JN (1983) The embryonic cell lineageof the nematode Caenorhabditis elegans. Dev Biol 100:64–119.

7. D’Alessio M, Frasch M (1996) msh may play a conserved role in dorsoventral pat-terning of the neuroectoderm and mesoderm. Mech Dev 58:217–231.

8. Denes AS, et al. (2007) Molecular architecture of annelid nerve cord supports commonorigin of nervous system centralization in bilateria. Cell 129:277–288.

9. Stolfi A, Ryan K, Meinertzhagen IA, Christiaen L (2015) Migratory neuronal progen-itors arise from the neural plate borders in tunicates. Nature 527:371–374.

10. Sulston JE, Horvitz HR (1977) Post-embryonic cell lineages of the nematode, Caenorhabditiselegans. Dev Biol 56:110–156.

11. Sauka-Spengler T, Meulemans D, Jones M, Bronner-Fraser M (2007) Ancient evolu-tionary origin of the neural crest gene regulatory network. Dev Cell 13:405–420.

12. Ramos C, Robert B (2005) msh/Msx gene family in neural development. Trends Genet21:624–632.

13. Abitua PB, Wagner E, Navarrete IA, Levine M (2012) Identification of a rudimentaryneural crest in a non-vertebrate chordate. Nature 492:104–107.

14. Sharman AC, Shimeld SM, Holland PW (1999) An amphioxus Msx gene expressedpredominantly in the dorsal neural tube. Dev Genes Evol 209:260–263.

15. Du H, Chalfie M (2001) Genes regulating touch cell development in Caenorhabditiselegans. Genetics 158:197–207.

16. Isshiki T, Takeichi M, Nose A (1997) The role of the msh homeobox gene duringDrosophila neurogenesis: Implication for the dorsoventral specification of the neu-roectoderm. Development 124:3099–3109.

17. Orgogozo V, Grueber WB (2005) FlyPNS, a database of the Drosophila embryonic andlarval peripheral nervous system. BMC Dev Biol 5:4.

18. Shaye DD, Greenwald I (2011) OrthoList: A compendium of C. elegans genes withhuman orthologs. PLoS One 6:e20085.

19. Liu X, et al. (2009) Analysis of cell fate from single-cell gene expression profiles inC. elegans. Cell 139:623–633.

20. Reece-Hoyes JS, et al. (2007) Insight into transcription factor gene duplication fromCaenorhabditis elegans Promoterome-driven expression patterns. BMC Genomics8:27.

21. Zhu Z, et al. (2014) A proneural gene controls C. elegans neuroblast asymmetric di-vision and migration. FEBS Lett 588:1136–1143.

22. Zhao C, Emmons SW (1995) A transcription factor controlling development of pe-ripheral sense organs in C. elegans. Nature 373:74–78.

23. Nikitina N, Sauka-Spengler T, Bronner-Fraser M (2008) Dissecting early regulatoryrelationships in the lamprey neural crest gene network. Proc Natl Acad Sci USA 105:20083–20088.

24. Holland LZ (2009) Chordate roots of the vertebrate nervous system: Expanding themolecular toolkit. Nat Rev Neurosci 10:736–746.

25. Imai KS, Levine M, Satoh N, Satou Y (2006) Regulatory blueprint for a chordate em-bryo. Science 312:1183–1187.

26. Strausfeld NJ, Hirth F (2016) Introduction to ‘Homology and convergence in nervoussystem evolution.’ Philos Trans R Soc Lond B Biol Sci 371:20150034.

27. Arendt D, Nübler-Jung K (1999) Comparison of early nerve cord development in in-sects and vertebrates. Development 126:2309–2325.

28. Hejnol A, Lowe CJ (2015) Embracing the comparative approach: How robust phy-logenies and broader developmental sampling impacts the understanding of nervoussystem evolution. Philos Trans R Soc Lond B Biol Sci 370:20150045.

29. Lowe CJ, et al. (2006) Dorsoventral patterning in hemichordates: Insights into earlychordate evolution. PLoS Biol 4:e291.

30. Rota-Stabelli O, Daley AC, Pisani D (2013) Molecular timetrees reveal a Cambriancolonization of land and a new scenario for ecdysozoan evolution. Curr Biol 23:392–398.

31. Petersen C, Dirksen P, Prahl S, Strathmann EA, Schulenburg H (2014) The prevalenceof Caenorhabditis elegans across 1.5 years in selected north German locations: The

importance of substrate type, abiotic parameters, and Caenorhabditis competitors.BMC Ecol 14:4.

32. Albertson DG, Thomson JN (1976) The pharynx of Caenorhabditis elegans. PhilosTrans R Soc Lond B Biol Sci 275:299–325.

33. Thompson KW, et al. (2016) The paired-box protein PAX-3 regulates the choice be-tween lateral and ventral epidermal cell fates in C. elegans. Dev Biol 412:191–207.

34. Wang W, Chen X, Xu H, Lufkin T (1996) Msx3: A novel murine homologue of theDrosophila msh homeobox gene restricted to the dorsal embryonic central nervoussystem. Mech Dev 58:203–215.

35. Duval N, et al. (2014) Msx1 and Msx2 act as essential activators of Atoh1 expression inthe murine spinal cord. Development 141:1726–1736.

36. Lin J, Wang C, Yang C, Fu S, Redies C (2016) Pax3 and Pax7 interact reciprocally andregulate the expression of cadherin-7 through inducing neuron differentiation in thedeveloping chicken spinal cord. J Comp Neurol 524:940–962.

37. Abitua PB, et al. (2015) The pre-vertebrate origins of neurogenic placodes. Nature524:462–465.

38. Arendt D, et al. (2016) The origin and evolution of cell types. Nat Rev Genet 17:744–757.39. Fritzsch B, Beisel KW, Pauley S, Soukup G (2007) Molecular evolution of the vertebrate

mechanosensory cell and ear. Int J Dev Biol 51:663–678.40. Feng G, et al. (2013) Developmental stage-dependent transcriptional regulatory

pathways control neuroblast lineage progression. Development 140:3838–3847.41. Wrischnik LA, Kenyon CJ (1997) The role of lin-22, a hairy/enhancer of split homolog, in

patterning the peripheral nervous system of C. elegans. Development 124:2875–2888.42. Garnett AT, Square TA, Medeiros DM (2012) BMP, Wnt and FGF signals are integrated

through evolutionarily conserved enhancers to achieve robust expression of Pax3 andZic genes at the zebrafish neural plate border. Development 139:4220–4231.

43. Waki K, Imai KS, Satou Y (2015) Genetic pathways for differentiation of the periph-eral nervous system in ascidians. Nat Commun 6:8719.

44. Murray JI, et al. (2012) Multidimensional regulation of gene expression in theC. elegans embryo. Genome Res 22:1282–1294.

45. Dickinson DJ, Ward JD, Reiner DJ, Goldstein B (2013) Engineering the Caenorhabditiselegans genome using Cas9-triggered homologous recombination. Nat Methods 10:1028–1034.

46. Frøkjær-Jensen C, et al. (2014) Random and targeted transgene insertion inCaenorhabditis elegans using a modified Mos1 transposon. Nat Methods 11:529–534.

47. Shen Z, et al. (2014) Conditional knockouts generated by engineered CRISPR-Cas9 endonuclease reveal the roles of coronin in C. elegans neural development. DevCell 30:625–636.

48. Murray JI, et al. (2008) Automated analysis of embryonic gene expression with cellularresolution in C. elegans. Nat Methods 5:703–709.

49. Ho VWS, et al. (2015) Systems-level quantification of division timing reveals a com-mon genetic architecture controlling asynchrony and fate asymmetry. Mol Syst Biol11:814.

50. Gratz SJ, et al. (2013) Genome engineering of Drosophila with the CRISPR RNA-guided Cas9 nuclease. Genetics 194:1029–1035.

51. Yang X, Yeo S, Dick T, Chia W (1993) The role of a Drosophila POU homeo domaingene in the specification of neural precursor cell identity in the developing embryoniccentral nervous system. Genes Dev 7:504–516.

52. Lander R, et al. (2013) Interactions between Twist and other core epithelial-mesenchymal transition factors are controlled by GSK3-mediated phosphorylation.Nat Commun 4:1542.

53. Zhang Y, Ding Y, Chen YG, Tao Q (2014) NEDD4L regulates convergent extensionmovements in Xenopus embryos via Disheveled-mediated non-canonical Wnt sig-naling. Dev Biol 392:15–25.

54. Sive HL, Grainger RM, Harland RM (2000) Early Development of Xenopus laevis: ALaboratory Manual (Cold Spring Harbor Lab Press, Cold Spring Harbor, NY).

55. Gao L, et al. (2016) A novel role for Ascl1 in the regulation of mesendoderm for-mation via HDAC-dependent antagonism of VegT. Development 143:492–503.

56. Coric T, Passamaneck YJ, Zhang P, Di Gregorio A, Canessa CM (2008) Simple chordatesexhibit a proton-independent function of acid-sensing ion channels. FASEB J 22:1914–1923.

57. Satou Y, Imai KS, Satoh N (2001) Action of morpholinos in Ciona embryos. Genesis 30:103–106.

58. Duggan A, Ma C, Chalfie M (1998) Regulation of touch receptor differentiation by theCaenorhabditis elegans mec-3 and unc-86 genes. Development 125:4107–4119.

59. Huang M, Chalfie M (1994) Gene interactions affecting mechanosensory transductionin Caenorhabditis elegans. Nature 367:467–470.

E6360 | www.pnas.org/cgi/doi/10.1073/pnas.1704194114 Li et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

1, 2

020