Innervation is required for sense organ developmentin the lateral line system of adult zebrafishHironori Wadaa,b,1, Christine Dambly-Chaudièrec,d, Koichi Kawakamib, and Alain Ghysenc,e,f,1,2

aPrecursory Research for Embryonic Science and Technology (PRESTO), Japan Science and Technology Agency, Kawaguchi, Saitama 322-0012, Japan; bDivisionof Molecular and Developmental Biology, National Institute of Genetics, Mishima, Shizuoka 411-8540, Japan; cUniversité Montpellier 2, Montpellier F-34095,France; dUnité Mixte de Recherche (UMR) 5235, Centre National de la Recherche Scientifique (CNRS), Montpellier F-34095, France; eInstitut National de laSanté et de la Recherche Médicale (INSERM) U710, Montpellier F-34095, France; and fÉcole Pratique des Hautes Études (EPHE), Paris F-75007, France

Edited* by A. J. Hudspeth, Howard Hughes Medical Institute, New York, NY, and approved February 12, 2013 (received for review August 15, 2012)

Superficial mechanosensory organs (neuromasts) distributed overthe head and body of fishes and amphibians form the “lateral line”system. During zebrafish adulthood, each neuromast of the body(posterior lateral line system, or PLL) produces “accessory” neuro-masts that remain tightly clustered, thereby increasing the total num-ber of PLL neuromasts by a factor of more than 10. This expansion isachievedbyabuddingprocess and is accompaniedbybranches of theafferentnerve that innervates the founder neuromast. Herewe showthat innervation is essential for the budding process, in completecontrast with the development of the embryonic PLL, where inner-vation is entirely dispensable. To obtain insight into the molecularmechanisms that underlie the budding process, we focused on theterminal system that develops at the posterior tip of the body and onthe caudalfin. In this subset of PLL neuromasts, bud neuromasts formin a reproducible sequence over a few days, much faster than forother PLL neuromasts. We show that wingless/int (Wnt) signalingtakes place during, and is required for, the budding process. We alsoshow that theWnt activator R-spondin is expressed by the axons thatinnervate budding neuromasts. We propose that the axon triggersWnt signaling, which itself is involved in the proliferative phase thatleads to bud formation. Finally, we show that innervation is requirednot only for budding, but also for long-term maintenance of allPLL neuromasts.

Lgr | post-embryonic development | Thunnus thynnus | Danio rerio

Various types of sense organs are distributed over the body ofvertebrates, such as pressure-sensitive Pacini corpuscles or

light touch-sensitive Meissner corpuscles in mammals. The cor-responding afferent neurons are located within dorsal root gangliaat a distance from the organs that they innervate, yet each cor-puscle ends up being innervated, and afferent neurons manage tofind a target corpuscle. The analysis of mutant mice revealed thatinnervation plays a role in the development of Pacini andMeissnercorpuscles. The inactivation of various neurotrophins, or of theirreceptors, leads to a reduction in the number of sense organs(reviewed in ref. 1). Whether innervation is responsible for theearly decision to form a sense organ, for its progressive differen-tiation, or for its maintenance remains unclear. This is, at least inpart, because it is not easy to identify the very first steps of senseorgan development and to determine whether they follow orprecede the arrival of afferent axons. The lateral line system of fishcomprises a number of discrete sensory organs (neuromasts) madeof a core of mechanosensory hair cells surrounded by support cellsand acts as a flow-sensitive device (reviewed in ref. 2). The pos-terior lateral line system (PLL), which extends over the trunk andtail, is innervated by sensory neurons clustered in a ganglion justposterior to the otic vesicle. Sensory neurons are entirely dis-pensable for the early development of the zebrafish PLL, however,as shown both by genetic and by surgical ablation of the ganglion(3, 4). Embryonic and larval development of the PLL (reviewed inref. 5 for zebrafish) lead to the juvenile stage, around 1 month postfertilization (mpf). At this stage, about 50 PLL neuromasts arealigned as four antero-posterior lines (6) (Fig. S1A). Neuromastsincrease in numbers throughout adult life due to the formation, by

each juvenile neuromast, of a cluster of neuromasts roughly orientedalong the dorsoventral axis and known as a “stitch” (Fig. S1B).Stitches can comprise more than 10 neuromasts in 2.5-mpf adultsand up to 40 in 18-mpf fish. At the posterior end of the body,a vertical line of terminal neuromasts is present just anterior to thecaudal fin (arrowheads, Fig. S1B), and four lines of neuromastsextend on the caudal fin (caudal lateral lines, or CLL; Fig. S1B) (7).CLL neuromasts have been proposed to be formed by migratingprimordia (8).Stitching has been studied in amphibians and shown to depend

on a budding process (9). Newly formed neuromasts within a stitchhave therefore been called “bud-neuromasts,” or “accessory”neuromasts (9). Here we show that stitch formation, including theformation of the CLL, entirely depends on innervation. Based onpatterns of gene expression and signaling activity, we propose thatcell proliferation, which is required for bud formation, depends onwingless/int (Wnt) signaling and that Wnt activation is triggeredwhen the leucine-rich, repeat-containing, G-protein–coupledreceptor (LGR), present on neuromast cells, binds the ligandR-spondin (Rspo), produced by afferent axons. We further showthat long-term maintenance of neuromasts also depends on in-nervation. We conclude that the development of the adult systemdiffers entirely from the embryonic one and resemblesmore closelythe interactions observed in the development of mammaliantouch receptors.

ResultsPostembryonic Development of the PLL in Zebrafish. The juvenilepattern of neuromasts is established at the end of larval life (1 mpf,9- to 10-mm fishes) (10). During adult life, further development ofthe PLL is achieved by the formation of additional neuromastsnext to each juvenile neuromast, leading to the formation of dor-soventrally oriented stitches (Fig. S1B). Here we concentrated onthe ventralmost of the four juvenile lines (line V, Fig. S1A), whichcomprises about 30 neuromasts, one on every intersomitic border.Stitch formation in zebrafish has been studied in only one case

so far, the opercular stitch, which belongs to the anterior lateralline system (11) (Fig. S1B). This stitch is unusual in that it beginsto form very early during larval life, around 3 d post fertilization(dpf). In the opercular case, stitch formation involves a buddingprocess that begins when a few cells elongate away from the

Author contributions: A.G. designed research; H.W. and A.G. performed research; H.W.and K.K. contributed new reagents/analytic tools; H.W., C.D.-C., and A.G. analyzed data;and A.G. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Data deposition: The rspo family genes reported in this paper have been deposited in theDNA Data Bank of Japan (DDBJ) [accession nos. AB769405 (rspo1), AB769406, AB769407(rspo2), AB769408 (rspo3), and AB769409 (rspo4)].1H.W. and A.G. contributed equally to this work.2To whom correspondence should be addressed. E-mail: alain.ghysen@univ-montp2.fr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1214004110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1214004110 PNAS | April 2, 2013 | vol. 110 | no. 14 | 5659–5664

NEU

ROSC

IENCE

founder neuromast. Cell movement along this extension, as wellas cell proliferation in its distal region, leads to a distal swellingthat progressively rounds up and differentiates as a new neuro-mast. We observed a similar sequence of elongation, swelling,and rounding up during stitching of PLL neuromasts (Fig. 1 Aand B).

Stitch Innervation. The various neuromasts of a stitch are in-nervated by branchlets of the nerve that innervates the founderneuromast. The order of budding events can usually be deducedfrom the pattern of branching (Fig. 1 C and D). Accessory neu-romasts can be formed on either side of the founder neuromast orsequentially on the same side, and they may in turn form addi-tional neuromasts or may remain quiescent (Fig. S2). We couldnot detect any obvious regularity in the pattern of budding, sug-gesting that bud neuromasts appear in a random sequence. Thereis, however, a general tendency of stitches of the ventral line to addnew neuromasts ventrally (Fig. 1D), possibly related to preferen-tial growth of the ventral scales in the ventral direction, similar tothe tendency of the opercular stitch to extend posteriorly andaccompany the growth of the opercular dermal bone (11).Branching of the nerve that extends to the different neuromasts

of a stitch could correspond to defasciculation of a bundle of axonsif new neurons innervate new neuromasts during stitch develop-ment, or it could correspond to axonal branching if the sameneurons that innervate the founder neuromast extend collateralsto innervate new neuromasts of the stitch. We answered thisquestion by labeling single embryonic neurons through plasmidinjection (12, 13) and found that all neuromasts of a stitch areinnervated by the same labeled neuron (Fig. 2A, 6 mpf).In the course of this experiment we noted that singly labeled

afferents (Fig. S3A) invade budding structures very early on (Fig.S3B), much before the buds become mature accessory neuro-masts. We confirmed this finding in the doubly transgenic strainET20; nbt-dsred, where neuromast mantle cells and interneuromastcells express gfp, whereas all neurons express dsred. We observedthat a neurite invades the earliest swelling of neuromasts (Fig.2 B and C) and accompanies the thin processes that mark theonset of budding (Fig. 2D).As neuromast extensions become thicker, and before hair cells

differentiate, the accompanying neurite has already developed adiscrete arborization (Fig. 2E). As described previously (11),the process expands progressively to form a radially organized

protoneuromast (Fig. 2F). At this early stage in the formation ofthe first bud neuromast, a new extension prefigures the formationof a second bud neuromast (Fig. 2G, arrowhead), and this exten-sion is already accompanied by a neurite (arrow). When the firstbud neuromast differentiates (Fig. 2H), the new extension (arrow-head) and accompanying neurite (arrow) are clearly defined.

Role of Innervation in Stitching. The prevalence of innervation atthe earliest stages of budding led us to examine a possible role ofthis innervation. With that aim, we examined the effect of earlyPLL ganglion ablation on stitching in 1.5-mpf, 13-mm youngadults. Complete ablation of the ganglion on the experimentalside had no effect on stitch formation in the control side (Fig.2I), but led to the complete absence of stitches in the ablatedside in all cases (n = 8). In three cases where ablation was notcomplete, all uninnervated neuromasts remained single (Fig. 2J),whereas the innervated neuromast (arrow) underwent stitching(arrowheads), thus demonstrating a tight link between innerva-tion and the formation of stitches. Although founder neuromaststhat were not innervated never formed accessory neuromasts, we

Fig. 1. Stitches in zebrafish. (A and B) Initiation of stitch formation incxcr4.139:rfp fish. Long cell processes extend from the founder neuromastalong the dorsoventral axis (A, arrows) and progressively swell up (arrow-head). Cell proliferation (B, arrow: cell rounding up beforemitosis) eventuallyleads to the formation of a new, complete accessory neuromasts (B, arrow-head). (C and D) Innervation of stitches in Hgn39d; ET20 fish. The branchingpattern reveals the history of stitch formation. Numbers refer to the succes-sive branching/budding events. F: founder neuromast. In all figures, anterioris left and dorsal is up. (Scale bars: A and B—20 μm; C and D—50 μm.)

Fig. 2. Innervation and budding. (A) Single neuron labeled through plasmidinjection in a cldnb:gfp background. (B–H) Successive stages in the formationof the first bud neuromast as seen in nbt-dsred, ET20 fish. Arrowheads:cellular extensions; arrows: accompanying neurite. (I and J) Effect of de-nervation on stitching. (I) Control side. (J) denervated side. Arrowheads:neuromasts; arrows: neurites. The remaining neurites in J belong to thecutaneous sensory system and are unrelated to the lateral line. (Scale bars:A, 50 μm; B–G, 20 μm; H and I, 100 μm.)

5660 | www.pnas.org/cgi/doi/10.1073/pnas.1214004110 Wada et al.

observed the presence of long, dorsoventrally oriented cellularextensions in all cases (Fig. 2J). This suggests that the first step inthe budding process—the extension of ventral or dorsal pro-cesses by one or a few cells of the founder neuromast—does notdepend on innervation. Innervation is required for later steps ofbudding, including cell proliferation, differentiation, and for-mation of hair cells.

Terminal Stitching in Bluefin Tuna. The analysis of innervation-dependent budding is complicated by the very slow nature of theprocess in which several days or weeks may elapse before a newbud neuromast is formed (Fig. S2). We found, however, that theformation of the terminal pattern in another teleost species, thebluefin tuna (Thunnus thynnus), involves a process of buddingthat is faster than that of body stitches.The embryonic PLL is nearly identical in zebrafish and in

bluefin tuna, although the two species are very distantly related(14). At 6 dpf, however, only two terminal neuromasts are pres-ent in bluefin tuna larvae [terminal 1 (ter1) and ter2, numberedfrom the posteriormost one], instead of three in zebrafish (ter1–3, numbered from the posteriormost one) (7). To better un-derstand whether and how this difference affects the develop-ment of the caudal system, we followed the development of theterminal and caudal system in bluefin tuna.We observed that, at 6–7 dpf (3.6-mm larvae), a neurite

extends away from ter2 (Fig. 3A, arrow) and accompanies longcellular extensions (Fig. 3B, arrow) that will eventually developas a new neuromast, ter2′ (Fig. 3C). The formation of caudallines begins soon after the differentiation of ter2′ and involvesthe successive addition of new neuromasts at the distal tip ofeach line. This process involves again a close correlation betweenthe extension of cellular processes from the distalmost neuro-mast, and the presence of a growing neurite (arrowhead andarrow, Fig. 3D). The extension of the four CLL proceeds oneneuromast at a time, and each line ends up at a neuromast be-tween two rounds of budding [Fig. 3 E and F, illustrating the firsttwo neuromasts of CLL1 and CLL2 (arrowheads)].

Terminal/Caudal Stitching in Zebrafish. In zebrafish, development ofthe terminal-caudal system begins around 10 dpf (4.5 mm). Threeter neuromasts are usually present at the end of embryogenesis(65%, n = 236, two in 26% of the cases, and four in another 9%;average 2.83 ± 0.57). In the juvenile pattern, a vertical row ofthree terminal neuromasts is present at the tip of the body (Fig.S1B). This vertical row was assumed to correspond to the hori-zontal row of terminal neuromasts present at the end of em-bryogenesis due to the dorsal bending of the anteroposterior axisin this region (7, 10). We reexamined this question in light of ourfindings with bluefin tuna. Of the three embryonic ter neuro-masts, ter1 moves dorsally due to dorsal bending of the noto-chord, and ter2 migrates ventrally away from its initial position(Fig. 3 G and H), whereas the most anterior one, ter3, retains itsoriginal position (Fig. 3L). As ter2 moves ventrally, it extendsa process that is innervated as soon as it forms (Fig. 3 G and H,arrows), much before the process swells up to prefigure bud-neuromast ter2′ (Fig. 3I). At 12 dpf, 2 d after the onset of bud-ding, the neurite extending to ter2′ has already sent a branch thatprefigures the tail line CLL3 (Fig. 3 J and K). It is only then thatthe first hair cell precursors, as labeled in the atoh-tomato line,differentiate at the core of ter2′ (Fig. 3K). The row of threeterminal neuromasts found in juveniles therefore corresponds toembryonic ter1 and ter2 and to the bud-neuromast ter2′ (Fig. 3L).A cellular process from ter1 that extends into the growing

caudal fin (Fig. 3M, arrowhead) is also accompanied by axoncollaterals (arrows). This process will subsequently round up toform the first neuromast of CLL1. The caudal lines later extendtogether with the growth of the caudal fin by the progressiveaddition of more distal neuromasts (7).

Because of the similarity of this process with the steps describedabove in ventral line stitching, we examined a possible role forinnervation in the development of ter2′ and of the caudal lines byganglion ablation. Ablation was performed on HuC:kaede larvaetwice, at 5 and 7 dpf, as it turned out to be impossible to ablate allneurons at once. Of 10 successfully ablated larvae that survived upto 1.5 mpf, 8 lacked all caudal lines and 2 had a single caudal lineinstead of the usual four (average of 0.2 line per larva). On thecontralateral side, eight larvae had four caudal lines, one hadthree, and one had two (average of 3.7 lines per larva).The experiment was repeated on Hgn39d fishes, with ablation

performed at 3–4 and 5–6 dpf. Larvae were checked at 10 dpf,and those with one to three neurons still present were ablateda third time. Larvae at 40 dpf that had no neurons left at 10 dpf

Fig. 3. Development of the terminal and caudal systems. (A–F) terminal andcaudal budding in Thunnus simultaneously labeled for actin by phalloidinlabeling and for acetylated tubulin by immunolabeling. (A and B) Neuriteextension from ter2 (arrow) prefigures the formation of bud-neuromast ter2′in Thunnus. (C) ter2′ has formed and remains neurally connected to ter2(arrow). (D) First neuromast of a caudal line with a neurite branch (arrow)extending into a budding extension (arrowhead). (E and F) First two neuro-masts (arrowheads) of the two dorsalmost CLL lines of the same larva,showing that the caudal lines form by sequential budding. (G–M) Terminaland caudal budding inDanio. (G–I) Ventral migration of ter2 and extension ofa neurite in the direction where ter2′ will form (arrows), as visualized in12-dpf nbt-dsred; ET20 fish. (J and K) Formation of hair cell precursors (red)takes place only in the budding structure at 12–13 dpf, after the neurite(green) has sent a branch to presumptive tail line CLL3, as visualized inHgn39d; atoh-tomato fish. (L) Overall pattern of ter neuromasts when bud-neuromast ter2′ appears, as visualized in nbt-dsred; ET20 fish. (M) Onset ofthe budding of tail line CLL1 from ter1. (Inset) Higher magnification of thedash-outlined area. A neurite (arrows) extends along the cell process (arrow-head) that emanates from ter1 to initiate CLL1. (Scale bars in A–M: 20 μm.)

Wada et al. PNAS | April 2, 2013 | vol. 110 | no. 14 | 5661

NEU

ROSC

IENCE

had only ter1 and ter2 on the ablated side, and no dorsal (ter1-derived) CLL were present (Fig. S4A′, n = 3). In larvae that werereablated at 10 dpf, ter2′ formed (presumably due to inductionof budding by the surviving neurons around 9–10 dpf), but nei-ther dorsal nor ventral CLL were present (Fig. S4B′, n = 5),demonstrating that innervation is directly required for the de-velopment of the entire CLL system. ter2′ and at least two CLLwere present on the control side in all cases. Noninnervatedterminal neuromasts were reduced in size relative to controlneuromasts in all cases. In the most extreme case, seven hair cellswere labeled in the control ter1, but only one was labeled in ter1on the experimental side (Fig. S4 C and C′).

Wnt Reporter Activity in Budding Cells. During embryonic de-velopment of the PLL,Wnt/β-catenin signaling is essential for cellproliferation in the leading region of the migrating primordium(15–18). This led us to examine whether Wnt activity can bedetected during stitching. With this aim in mind, we used the Wntreporter line Tcf/Lef-miniP:dgfp (19), where we detected GFP inthe budding cells that form ter2′ (Fig. 4A, arrowhead). TheseWnt-positive cells are accompanied by a neurite revealed byphotoconverted Huc-Kaede labeling (Fig. 4A, arrow). We con-firmed that the presence of GFP reveals bona fide Wnt signalingby in situ hybridization with an anti-lymphoid enhancer-bindingfactor 1 (LEF1) probe (Fig. 4B, arrowheads). After ter2′ hasdifferentiated, we observedWnt activity in cells emanating from it(Fig. 4C, arrowheads). This bud prefigures the first ventral caudalline (Fig. 4D), where the budding process (arrowheads) is also

accompanied by a neurite (arrows). Note that the GFP signal hasdisappeared in the mature bud-neuromast ter2′.If Wnt activity is responsible for cell proliferation and budding,

one would expect that interfering with Wnt signaling would affectbudding. We tested this prediction by overexpressing the Wnt an-tagonist, Dickkopf (Dkk), in Tg(hsp70l:dkk1b-GFP) larvae (20, 21).After heat shock at 12 and 13 dpf, we observed at 18 dpf that ter2′was absent on one side in 6 of 10 hsp:dkk carriers [Fig. 4E, affectedside, vs. Fig. 4F, normal side; CLL3 has begun to form on thenormal side (arrowheads) but not on the affected side]. In non-heat-shocked larvae, ter2′ was absent in only 2/48 sides (4.2%, instead of30% after heat shock). In one of the two cases, ter2 was missing aswell, suggesting that the absence of ter2′ is secondary to that of ter2,and in the other case two neuromasts were present at the position ofter2, suggesting that ter2′ formed but failed to migrate.

Expression of rspo2:gfp in the Sensory Neurons. During both ter2′and caudal budding, we observed Wnt activity in the vicinity ofafferent axons (arrowheads in Fig. 4 A and D). The close cor-relation between innervation and Wnt activity suggests thatsensory neurons may provide a factor that activates Wnt sig-naling in the extending processes. Rspo/Lgr signaling is known toactivate Wnt signaling by stabilizing the Wnt receptor complex(22). We confirmed that the Lgr6 receptor is strongly expressedin most or all neuromast cells up to 3 dpf (23) (Fig. S3C) andshowed that it remains expressed in neuromast cells, includingbudding ter2′, at 11 dpf (Fig. S3D). The Rspo ligand wouldtherefore be a good candidate to act as axonal activator of Wntsignaling during neuromast budding. Of four rspo family genesfound in the zebrafish genome (Fig. S5A), we determined thatrspo2 is expressed in a subset of lateral line neurons (Fig. S5B).Using a Gal4 enhancer trap line for rspo2 (hspGFFDMC131A)(24), we observed intense labeling in the dorsalmost neurons ofthe PLL ganglion (Fig. 4G), corresponding to the neurons thatinnervate the terminal neuromasts (25, 26). Using the sametransgenic combination to detect the axons of rspo2-expressingneurons, we found that at 7 dpf, shortly before the formation ofthe bud-neuromast ter2′, the axons innervating ter1 and ter2express GFP (Fig. S3E), consistent with a possible role for Rspo/Lgr signaling in the budding process. We also observed thepresence of GFP at 11 dpf in a neurite that leaves ter2 and formsan incipient arborization in bud-ter2′ (Fig. S3F), thus confirmingthat neurites that accompany budding processes express rspo2.If Rspo produced by the neurites plays a role in the activation

of Wnt signaling in neuromast cells, one would expect that re-moving the innervation would affect Wnt activity in the buddingcells. We tested this notion by cutting the PLL nerve at 9 dpf,shortly before bud induction takes place, and a second time at 12dpf, just before the regenerating nerve reaches the terminal re-gion in HuC:gal4; UAS:rfp; Tcf/Lef:gfp larvae where neurites andhair cells both express red fluorescent protein (RFP). We ob-served at 17 dpf that, of six larvae, four showed substantial re-duction of Tcf/Lef:gfp activity (Fig. 4H, treated side, vs. Fig. 4I,control side of the same larva) and one showed complete dis-appearance. We repeated the experiment and quantified thelevel of GFP fluorescence. We observed that the ratio of ex-perimental vs. control side was 0.55 ± 0.60 after nerve cut (n =11) vs. 1.14 ± 0.76 after sham illumination of the experimentalside (n = 8), a statistically significant difference (P = 0.042). Weconclude that Wnt activation depends on neuromast innervation.

Innervation and Neuromast Maintenance in Zebrafish. To seewhether stitches could ever form in the absence of in-nervation, we raised adult fish where the PLL ganglion hadbeen ablated early during larval life. We observed thatstitches never form, as described above in 1.5-mpf adults, butin addition we found that the size of neuromasts begins todecrease when the fish reach ∼1.5 cm (Fig. 5A), although

Fig. 4. Reporting Wnt and Rspo2 activity during budding. (A) Wnt activity(green) prefiguring ter2′, next to ter2, in a 9-dpf larva. Wnt activity (arrow-head) is adjacent to extending neurites revealed by photoconverted HuC-Kaede (arrow). (B) Presence of lef1 mRNA in the ter2′ budding process. (C)Wnt activity (green) in cells emanating from ter2′ and prefiguring the firstventral line of the caudal system in 10-dpf larvae (arrowheads). (D) Buddingof the first ventral line of the caudal system. Wnt activity is detected inbudding cells (arrowheads) but not in ter2′ anymore and is tightly correlatedto neurite extension (arrows). (E) Absence of ter2′ after overexpression of theWnt antagonist Dkk. On the contralateral side (F), ter2′ formed normally andthe CLL3 line also developed (arrowheads). Hair cells and sensory axons arerevealed by DiAsp labeling. (G) Rspo2 enhancer trap activity in the PLL gan-glion (outlined in dotted lines). Arrow: root of the PLL nerve extending to theneuromasts. (H) After nerve cut, Wnt activity is severely reduced next to ter2(asterisk), compared with the control side (I). Arrow: branch of the PLL nerveinnervating ter2; double arrow: branchlet innervating ter2′. “m” indicatesGFP-expressing mesenchymal cells that contribute to fin development. Ex-ceptionally, posterior (“P”) is left in E and H.

5662 | www.pnas.org/cgi/doi/10.1073/pnas.1214004110 Wada et al.

the span of dorsal and ventral extensions closely matches thespan observed on the control side. This is consistent with theobservation that the size of ter neuromasts is reduced in 40-dpf fishes where the ganglion had been ablated (Fig. S4).Upon further growth, neuromasts disappear altogether (Fig.5B), and the dorsal and ventral extensions eventually disap-pear as well (Fig. 5C).

DiscussionContrary to most other vertebrates, fish keep growing throughoutadult life. Some sensory organs, e.g., the eyes, increase accordinglyin size during adulthood. In the case of the posterior lateral linesystem, however, individual sense organs retain the same sizethroughout life, and growth of the system is achieved through theformation of “accessory” neuromasts that remain closely associ-ated with the “founder” neuromast of the juvenile pattern andform dorsoventrally oriented stitches.The very first step in stitch formation—the extension of cellular

processes from the founder neuromast—takes place normally in theabsence of innervation. Subsequent steps (swelling, radial organi-zation, and hair cell differentiation) are all blocked, however. Thisresult suggests that innervation is required for cell proliferation andthat, in its absence, budding is not possible. Stitch formationwould therefore depend both on the neuromast’s ability to extendprocesses in preferred directions and on the axonal ability tostimulate cell proliferation and thereby convert extensions intonew neuromasts. A similar two-tier mechanism may also operatein the case of the mammalian touch-sensitive corpuscles: com-petence to form organs may be distributed, but innervation wouldbe essential to the realization of this competence (1).The importance of innervation for budding and stitch forma-

tion contrasts with the demonstrated lack of role for innervationin the formation of the embryonic and larval patterns. A possibleinterpretation of this difference is that the ancestral processleading to the development of superficial sense organs dependedon an interaction between competent tissue and inducing neu-rites, as still observed in adult fish and in developing mammals.The embryonic development of the PLL, where sense organ de-velopment anticipates innervation, may have allowed for fasterdevelopment of a functional escape reaction.Our results show that the formation of the terminal and

caudal fin patterns in both Thunnus and Danio also involves abudding process where innervated processes elongate in preferred

directions and then give rise to additional ter and CLL neuromasts.The extreme fragility of Thunnus larvae made it impossible to as-sess the effect of denervation on the development of the terminal/caudal pattern. InDanio, however, we could show that denervationresults in the absence of ter2′ and of caudal lines, similar to otherPLL stitches.We have documented the reactivation of Wnt sig-naling at early stages of the terminal budding process, consistentwith the well-documented role of this signaling pathway in con-trolling cell proliferation. We confirmed the requirement for Wntsignaling by showing that heat-shock–induced expression of theWnt-signaling antagonist Dkk reduces the number of ter2′ neuro-masts and delays the formation of CLL lines. It has been reportedthat budding takes place normally in lef1 mutant fishes (18). Thissuggests that other T cell-specific factor (tcf) genes, possibly tcf7(27), can substitute for lef1/tcf1 at later stages of development, asshown in many other sytems (reviewed in ref. 28).One of the pathways to activate Wnt signaling involves binding

of the ligand Rspo2 to its receptor Lgr (22). Lgr is present inneuromast cells (23), and we found that Rspo2 is present in asubset of PLL afferent neurons and in the neurites that innervatethe terminal neuromasts at the time they form bud-neuromastter2′. Furthermore, we showed that nerve cuts intended to pre-vent or delay terminal innervation reduce substantially Wnt ac-tivity associated with the budding of ter2′. The observation thatsome Wnt activity remains may be due either to the fact that thefirst nerve cut was done at 9 dpf, a time when some Rspo2 mayhave already been produced by afferent neurites, or to the pos-sibility that other ligands, possibly related to the presence ofnearby proliferating fin tissue, may partially substitute for Lgr.Altogether, our observations are entirely consistent with a

simple scheme where Rspo2 secreted by afferent axons wouldbind to its Lgr receptor present on neuromast cells and inducelocal proliferation through Wnt activation, thus resulting in theformation of a bud neuromast. In addition, we demonstrate thatinnervation is also required for long-term survival of neuromasts,emphasizing the contrast between early and adult developmentof the same sensory system.

Materials and MethodsFish Care and Fish Strains. Fish were kept under standard conditions as de-scribed by Westerfield (29). The experimental animals’ care was in accordancewith the guidelines of the National Institute of Genetics and of UniversityMontpellier 2. For imaging, embryos were anesthetized with Tricain (3-amino

Fig. 5. Degeneration of neuromasts in the absenceof innervation. (A) Control and experimental sides ofa 2-mpf nbt-dsred, ET20 fish; the PLL ganglion hadbeen ablated at 5–7 dpf in the experimental side. (Band C) Control and experimental sides of Hgn39d,ET20, DiAsp-labeled fish where the PLL ganglion hadbeen ablated at 5–7 dpf. (B) Three-month post fer-tilization, 2.4-cm fish. (C) Three-month post fertil-ization, 2.6-cm fish. (Scale bars: 500 μm.)

Wada et al. PNAS | April 2, 2013 | vol. 110 | no. 14 | 5663

NEU

ROSC

IENCE

benzoic acid ethylester; Sigma A-5040): 4 mg/mL in 20 mM Tris (pH 7) andmounted in 0.7% agar in fish water. Sensory neurons, including their neurites,were visualized in nbt-dsred, where all neurons show red fluorescence [ZFINTg(Xla.Tubb:DsRed)] (30) in HuC:kaede fish [ZFIN Tg(elavl3:Kaede)] (31)where they fluoresce in red after photoconversion in the Hgn39d line[ZFIN Et(T2KHG)39d] (27), where only lateral line neurons express GFP, orin larvae combining Huc:gal4 [ZFIN Tg(elavl3:Gal4-VP16)] (32) and UAS:rfp[ZFIN Tg(5xUAS:RFP)] (33). Neuromasts were visualized in green in Et20:gfp[ZFIN Et(krt4:EGFP)sqet20] (34) or in cldnb:gfp [ZFIN Tg(-8.0cldnb:lynEGFP)](35) or in red in cxcr4.139:rfp [ZFIN Tg(cxcr4b:mRFP)] (36) fish. Differentiationof hair cells was monitored using atoh-tomato fish [ZFIN Tg(atoh1a:dTomato)](11). Wnt activity was assessed using the Tcf/Lef-miniP:dGFP [ZFIN Tg(OTM:d2EGFP)] (19), and rspo2 expression was reported in the hspGFFDMC131A[ZFIN Et(hsp70l:Gal4FFDMC)131a] (24) and in the UAS:GFP lines (33).

Labeling and Imaging. Single-neuron labeling was achieved through plasmidinjection in fertilized eggs, as previously described (25), using a Huc:mem-TdTomato construct (26). Tuna embryos were simultaneously labeled foractin by phalloidin labeling and for acetylated tubulin by immunolabeling,as described previously (37). Neuromast hair cells were labeled by 4-(4-dieth-ylaminostyryl)-N-methylpyridinium iodide (DiAsp, Sigma) as described pre-viously (38). Imaging was done on a Leica SPE, a Zeiss LSM700 confocalmicroscope, or a Zeiss Axioimager equipped with a Coolsnap camera.

Laser Ablation and Heat-Shock Treatment. AMicropoint laser system (PhotonicInstruments) was used on a Zeiss Axioplan 2 microscope with either 40× (forganglion ablation) or 63× (for nerve cuts) water immersion objectives, usingcoumarin-440 nm, 5 mM in methanol, as a laser medium. The PLL nerve wasablated by laser illumination at the level of somites 1–3. As a control (sham)operation, the same amount of fluorescent illumination was applied tomyotomes along the lateral line. The levels of GFP fluorescence were quan-tified under ImageJ software (http://rsb.info.nih.gov/ij/) using the “Measure”function. Heat shocks were performed by transferring thefish tank into an airincubator at 39 °C for 70 min, allowing 10 min for the water to reach 39 °C.

Isolation of rspo Family Genes. cDNA fragments of zebrafish rspo genes wereisolated by RT-PCR using specific primers designed from the Sanger Centre ge-nome database (www.ensembl.org/index.html) (rspo1f 5′-CCAGGGACTATG-CATTTGGGA-3′; rspo1r 5′-CCACCTGATGCTGTGATGAAA-3′, rspo2f 5′-ATGCAG-TTTCGCCTGTTCTCA-3′; rspo2r 5′-TATTTTTTACTGGCCTGCCCG-3′, rspo3f 5′-ATG-CAATTGCAACTGATCTCC-3′; rspo3r 5′-TGTCTATACTGTTCCAGCGTC-3′, rspo4f 5′-CAGATGCATTGGCAACTTTTG-3′; rspo4r 5′-TTATTGAGGAACAAAACTGCG-3′). AnRNA extraction kit (Qiagen) and a first-strand cDNA synthesis kit (Takara) wereused. The PCR product was cloned into pCRII-TOPO (Invitrogen) and se-quenced using a BigDye terminator cycle sequence kit with a DNA sequencer(PE Applied Biosystems). The amino acid sequences were deduced from thenucleotide sequences and analyzed using the CLUSTAL W program at theDNA Data Base in Japan (www.ddbj.nig.ac.jp/search/clustalw-e.html).

In Situ Hybridization. RNA probes were synthesized using an RNA transcriptionkit (Invitrogen), and in situ hybridization was performed as described (29).Image was captured using a dissecting microscope (Leica M165FC) witha CCD camera (Nikon DXM1200F).

ACKNOWLEDGMENTS.We thank T. Ishitani for the Wnt-reporter line, whichwas made available to us prior to publication. We also thank H. Okamotofor the HuC:kaede line and for unpublished information on the rspo genes;Y. Kikuchi for the lgr6 gene; S. Higashijima for HuC:gal4; R. Moon for hsp:dkk;D. Gilmour for cldnb:gfp fish; D. Gilmour and M. van Drenth for the nbt-dsredline; H. Lopez-Schier for the Hgn39d line; and V. Korzh and M. Allende for theEt20:gfp line. We acknowledge the help of H. Kakinuma and H. Okamoto withlines and lenses. We thank Françoise Carbonell for help with plasmid injectionsand Fernando de la Gandara, Aurelio Gonzalez, and Denis Coves for providingtuna larvae. We thank Nicolas Cubedo for excellent fish husbandry, and theMontpellier RIO Imaging facility for making its confocal station available to us.H.W. was funded by the Precursory Research for Embryonic Science and Tech-nology Program of the Japan Science and Technology Agency. Part of thiswork was supported by the National BioResource Project.

1. Vega JA, García-Suárez O, Montaño JA, Pardo B, Cobo JM (2009) The Meissner andPacinian sensory corpuscles revisited new data from the last decade. Microsc Res Tech72(4):299–309.

2. Coombs S, Montgomery JC (1999) Comparative Hearing: Fish and Amphibians, edsFay RR, Popper AN (Springer Verlag, New York), pp 319–362.

3. Andermann P, Ungos J, Raible DW (2002) Neurogenin1 defines zebrafish cranialsensory ganglia precursors. Dev Biol 251(1):45–58.

4. Grant KA, Raible DW, Piotrowski T (2005) Regulation of latent sensory hair cell pre-cursors by glia in the zebrafish lateral line. Neuron 45(1):69–80.

5. Ghysen A, Dambly-Chaudière C (2007) The lateral line microcosmos. Genes Dev 21(17):2118–2130.

6. Nuñez VA, et al. (2009) Postembryonic development of the posterior lateral line in thezebrafish. Evol Dev 11(4):391–404.

7. Wada H, Hamaguchi S, Sakaizumi M (2008) Development of diverse lateral line pat-terns on the teleost caudal fin. Dev Dyn 237(10):2889–2902.

8. Dufourcq P, et al. (2006) Mechano-sensory organ regeneration in adults: The zebra-fish lateral line as a model. Mol Cell Neurosci 33(2):180–187.

9. Stone LS (1937) Further experimental studies of the development of the lateral-linesense organs in amphibians observed in living preparations. J Comp Neurol 68(1):83–115.

10. Ledent V (2002) Postembryonic development of the posterior lateral line in zebrafish.Development 129(3):597–604.

11. Wada H, et al. (2010) Dermal morphogenesis controls lateral line patterning duringpostembryonic development of teleost fish. Dev Biol 340(2):583–594.

12. Nagiel A, Andor-Ardó D, Hudspeth AJ (2008) Specificity of afferent synapses ontoplane-polarized hair cells in the posterior lateral line of the zebrafish. J Neurosci28(34):8442–8453.

13. Faucherre A, Pujol-Martí J, Kawakami K, López-Schier H (2009) Afferent neurons of thezebrafish lateral line are strict selectors of hair-cell orientation. PLoS ONE 4(2):e4477.

14. Ghysen A, et al. (2010) Development of the posterior lateral line system in Thunnusthynnus, the Atlantic blue-fin tuna, and in its close relative Sarda sarda. Int J Dev Biol54(8–9):1317–1322.

15. Gamba L, Cubedo N, Lutfalla G, Ghysen A, Dambly-Chaudiere C (2010) Lef1 controlspatterning and proliferation in the posterior lateral line system of zebrafish. Dev Dyn239(12):3163–3171.

16. Aman A, Nguyen M, Piotrowski T (2011) Wnt/β-catenin dependent cell proliferationunderlies segmented lateral line morphogenesis. Dev Biol 349(2):470–482.

17. Valdivia LE, et al. (2011) Lef1-dependent Wnt/β-catenin signalling drives the pro-liferative engine that maintains tissue homeostasis during lateral line development.Development 138(18):3931–3941.

18. McGraw HF, et al. (2011) Lef1 is required for progenitor cell identity in the zebrafishlateral line primordium. Development 138(18):3921–3930.

19. Shimizu N, Kawakami K, Ishitani T (2012) Visualization and exploration of Tcf/Lef functionusing a highly responsive Wnt/β-catenin signaling-reporter transgenic zebrafish. Dev Biol370(1):71–85.

20. Aman A, Piotrowski T (2008) Wnt/beta-catenin and Fgf signaling control collective cellmigration by restricting chemokine receptor expression. Dev Cell 15(5):749–761.

21. Stoick-Cooper CL, et al. (2007) Distinct Wnt signaling pathways have opposing roles inappendage regeneration. Development 134(3):479–489.

22. de Lau W, et al. (2011) Lgr5 homologues associate with Wnt receptors and mediateR-spondin signalling. Nature 476(7360):293–297.

23. Hirose K, Shimoda N, Kikuchi Y (2011) Expression patterns of lgr4 and lgr6 duringzebrafish development. Gene Expr Patterns 11(7):378–383.

24. Yano T, Abe G, Yokoyama H, Kawakami K, Tamura K (2012) Mechanism of pectoralfin outgrowth in zebrafish development. Development 139(16):2916–2925.

25. Sato A, Koshida S, Takeda H (2010) Single-cell analysis of somatotopic map formationin the zebrafish lateral line system. Dev Dyn 239(7):2058–2065.

26. Pujol-Martí J, Baudoin JP, Faucherre A, Kawakami K, López-Schier H (2010) Pro-gressive neurogenesis defines lateralis somatotopy. Dev Dyn 239(7):1919–1930.

27. Nagayoshi S, et al. (2008) Insertional mutagenesis by the Tol2 transposon-mediatedenhancer trap approach generated mutations in two developmental genes: tcf7 andsynembryn-like. Development 135(1):159–169.

28. Hoppler S, Kavanagh CL (2007) Wnt signalling: Variety at the core. J Cell Sci 120(Pt 3):385–393.

29. Westerfield M (2000) The Zebrafish Book (Univ. of Oregon Press, Eugene, OR), 4th Ed.30. Peri F, Nüsslein-Volhard C (2008) Live imaging of neuronal degradation by microglia

reveals a role for v0-ATPase a1 in phagosomal fusion in vivo. Cell 133(5):916–927.31. Sato T, Takahoko M, Okamoto H (2006) HuC:Kaede, a useful tool to label neural

morphologies in networks in vivo. Genesis 44(3):136–142.32. Kimura Y, Satou C, Higashijima SI (2008) V2a and V2b neurons are generated by the

final divisions of pair-producing progenitors in the zebrafish spinal cord. Development135(18):3001–3005.

33. Asakawa K, et al. (2008) Genetic dissection of neural circuits by Tol2 transposon-mediated Gal4 gene and enhancer trapping in zebrafish. Proc Natl Acad Sci USA105(4):1255–1260.

34. Parinov S, Kondrichin I, Korzh V, Emelyanov A (2004) Tol2 transposon-mediated en-hancer trap to identify developmentally regulated zebrafish genes in vivo. Dev Dyn231(2):449–459.

35. Haas P, Gilmour D (2006) Chemokine signaling mediates self-organizing tissue mi-gration in the zebrafish lateral line. Dev Cell 10(5):673–680.

36. Gamba L, Cubedo N, Ghysen A, Lutfalla G, Dambly-Chaudière C (2010) Estrogen re-ceptor ESR1 controls cell migration by repressing chemokine receptor CXCR4 in thezebrafish posterior lateral line system. Proc Natl Acad Sci USA 107(14):6358–6363.

37. Ghysen A, Dambly-Chaudière C, Coves D, de la Gandara F, Ortega A (2012) Deve-lopmental origin of a major difference in sensory patterning between zebrafish andbluefin tuna. Evol Dev 14(2):204–211.

38. Collazo A, Fraser SE, Mabee PM (1994) A dual embryonic origin for vertebrate mecha-noreceptors. Science 264(5157):426–430.

5664 | www.pnas.org/cgi/doi/10.1073/pnas.1214004110 Wada et al.