FGF-Dependent Mechanosensory Organ Patterning in Zebrafish

See allHide authors and affiliations

Science  27 Jun 2008:
Vol. 320, Issue 5884, pp. 1774-1777
DOI: 10.1126/science.1156547


During development, organ primordia reorganize to form repeated functional units. In zebrafish (Danio rerio), mechanosensory organs called neuromasts are deposited at regular intervals by the migrating posterior lateral line (pLL) primordium. The pLL primordium is organized into polarized rosettes representing proto-neuromasts, each with a central atoh1a-positive focus of mechanosensory precursors. We show that rosettes form cyclically from a progenitor pool at the leading zone of the primordium as neuromasts are deposited from the trailing region. fgf3/10 signals localized to the leading zone are required for rosette formation, atoh1a expression, and primordium migration. We propose that the fibroblast growth factor (FGF) source controls primordium organization, which, in turn, regulates the periodicity of neuromast deposition. This previously unrecognized mechanism may be applicable to understanding segmentation and morphogenesis in other organ systems.

Primordia segmentation into reiterated structures occurs in the development of such diverse structures as photoreceptor ommatidia in the Drosophila compound eye and segregation of vertebrate paraxial mesoderm into somites. We describe the organization and regulation of segmentation in the zebrafish lateral line, where mechanosensory organs—called neuromasts—form in discrete clusters along the trunk. This pattern is produced by repeated segmentation of the pLL primordium, which migrates along the trunk, depositing 20 to 30 cells at regular (five-to-seven somite) intervals (1). Although the lateral line is an emerging system for the study of collective migration, mechanosensory hair cell development, and hair cell regeneration, the regulation of its initial development and cyclic segmentation is unknown.

The migrating pLL primordium is organized into rosettes, each corresponding to a proto-neuromast (1). Cells within each rosette have distinct polarity, with apical ends centrally constricted and nuclei basally localized. We analyzed the relationships between rosette addition, primordium migration, and neuromast deposition by time-lapse imaging. The pLL primordium is first recognized as a group of cells caudal to the otic vesicle with a single anteriorly positioned rosette (Fig. 1A). Over the next few hours, additional rosettes are sequentially added to the posterior, toward the future primordium leading edge (Fig. 1, A to D, and movie S1). Migration begins once two to three rosettes have formed, and the onset of neuromast deposition correlates with formation of the fourth rosette in the leading zone. During deposition, the trailing proto-neuromast gradually slows down before coming to a complete stop (movie S2). Over the course of migration, the primordium exhibits a cyclical behavior with a new rosette generated near the leading (posterior) zone soon after a proto-neuromast separates from the trailing (anterior) edge, alternating from three to four proto-neuromast rosettes (Fig. 1, E and F, and movie S2). This cyclical behavior produces five to six primary neuromasts [named L1 to L6, following the convention in (2)] deposited along the trunk and a terminal cluster of two to three neuromasts at the end of the tail.

Fig. 1.

Rosettes are renewed from the leading-edge progenitor zone. (A to F) Lateral views of the pLL primordium in a single confocal plane in Tg(Bactin:HRAS-EGFP) fish (20) (right) and corresponding schematic representation (left). One rostral rosette [arrowhead, (A)] is formed at 19 hpf, and three more rosettes [arrowheads, (B) to (D)] are sequentially added between 19 and 25 hpf (see movie S1). Neuromast deposition begins once the primordium is organized into four rosettes [(D) and (E)]; a new rosette is then organized from the leading end (F). (G) Single confocal plane of the primordium leading zone at 24 hpf immediately after photoconversion of green Kaede protein to red (arrow). (H and I) Single confocal planes of the L5 neuromast (H) and terminal cluster (I) from the embryo shown in (G) at 54 hpf. Differential labeling possibly reflects different proliferation rates and/or differential origin. s, somite; L1 to L5, trunk neuromasts. Scale bar in (A), 50 μm.

We reasoned that if leading-edge cells act as a progenitor zone, then they should contribute to new neuromasts once already-formed proto-neuromasts are deposited. Following cell fates by photoconversion of Kaede protein (3, 4) confirmed this hypothesis (Fig. 1G). In a majority of embryos (77%, n = 17), the labeling of leading cells, after two to three rosettes had formed, resulted in labeled cells only in L3 to L6 neuromasts and the terminal cluster (Fig. 1, H and I, and table S1). In every case, almost all cells within neuromasts were labeled (Fig. 1H), indicating that a small number of cells act as a progenitor zone.

FGFs regulate cellular morphogenesis in many systems (5, 6). Using the FGF receptor (Fgfr) inhibitor SU5402 (7) or an inducible dominant-negative Fgfr1 transgenic strain hsp70:dn-fgfr1 (8), we found that FGF signaling is necessary for rosette formation, neuromast deposition, and proper primordium migration (fig. S1, A to H). Application of increasing concentrations of SU5402 caused corresponding reductions in neuromast numbers, confirming specificity of the drug-induced phenotype (fig. S1, I and J). Application of SU5402 beginning at 13 hours postfertilization (hpf), before lateral line placodes were induced (9), did not affect formation of the first, most rostral rosette but blocked formation of more caudal rosettes (fig. S2, A and B, and movie S3). This defect in rosette formation did not alter initial migration onset. These results suggest that FGF signals are not needed for initial primordium formation or polarization, a process whose regulation remains unknown.

Later FGF inhibition during primordium migration altered its organization and movement (fig. S1, D to G, and movie S4). Migration was not affected immediately after inhibitor application, and a single neuromast was deposited. However, subsequent rosette renewal was blocked, and after migrating a distance of 8 to 10 somites, the primordium became disorganized and stopped moving. Alterations in migration were probably caused by changes in expression of the chemokine receptors cxcr4b and cxcr7b (fig. S3), which are required for proper migration (1012). FGF-inhibitor effects were reversible; after inhibitor removal, the primordium reorganized into three to four rosettes and then resumed migration (movie S5).

Expression of FGF signaling components is consistent with a role in primordium organization. fgfr1 was strongly expressed in the primordium trailing region, with lower expression in the leading zone (Fig. 2A). pea3, a downstream target of FGF signaling (13, 14), was expressed in the trailing zone (Fig. 2B). Of 22 FGF ligands screened, only fgf3 (15) and fgf10 (16) were expressed in the pLL primordium, broadly in the leading zone and restricted to a focus of one to two cells in trailing rosettes (Fig. 2, C and E). As the primordium organizes into rosettes, both fgf3 and fgf10 exhibit dynamic expression (fig. S4). fgf transcripts were first detected at the trailing and middle regions (19 to 23 hpf), then confined to the leading zone (24 hpf), and resolved to a single focus in the trailing edge by 25 hpf. When FGF signaling was blocked, fgf3 and fgf10 transcripts were detected throughout the primordium (Fig. 2, D and F), suggesting negative-feedback regulation.

Fig. 2.

Dynamic expression of FGF signaling components in the migrating pLL primordium. (A) Fluorescent in situ hybridization of fgfr1 (red) and deltaA (green) transcripts at 31 hpf. Nuclei were labeled with 4′,6′-diamidino-2-phenylindole (DAPI) (blue). Cell shapes were revealed with antibody against pan-Cadherin (magenta). Arrows indicate rosette centers. fgfr1 expression is strong in the trailing part of the primordium and weaker in the leading zone (brackets). (B) At 31 hpf, pea3 (red) is expressed in trailing rosettes [revealed by basally displaced nuclei (arrows)]. (C and E) fgf3 and fgf10 are expressed in the leading zone and in distinct foci in the trailing zone. fgf3 is expressed at relatively low levels as compared with fgf10. (D and F) Expanded fgf3 and fgf10 expression after treatment with SU5402 between 27 to 31 hpf. Primordia are outlined by dotted lines. Scale bars in (A) to (C), 50 μm.

Functional importance was demonstrated by injecting fgf10 morpholino oligonucleotides (fgf10-MO) (fig. S5) into zygotes derived from fgf3+/– crosses. Whereas reduction of Fgf3 or Fgf10 alone resulted in mild alterations, combined reduction significantly reduced neuromast number and inhibited primordium migration (Fig. 3, A to D). Intermediate phenotypes seen with fgf10-MO injection into zygotes with one functional fgf3 allele (Fig. 3, C and D) were analogous to those found at intermediate concentrations of SU5402 inhibitor, suggesting that neuromast position and number are sensitive to overall levels of FGF signal. In Fgf3/10-deficient embryos, the most rostral rosette formed normally, but subsequent rosette formation was blocked (fig. S2C and movie S6). The combined Fgf3/10-deficient phenotype, although similar to that observed when all FGF signaling was blocked with an SU5402 inhibitor, is not as severe (compare fig. S1G and Fig. 3D), and thus we cannot exclude the involvement of another FGF ligand in these processes.

Fig. 3.

Fgf3 and Fgf10 are necessary for primordium organization and neuromast deposition. (A and B) Zygotes derived from fgf3+/– parents were injected with fgf10-MO, collected at 48 hpf, analyzed for eya1 expression, and genotyped. (A) In wild-type (fgf3+/+ or fgf3+/–) embryos, the pLL primordium migrated through the trunk and deposited five to six neuromasts. (B) fgf3–/– embryos injected with fgf10-MO displayed only one or no neuromasts and showed complete disorganization of the primordium. (C and D) Neuromast (NM) numbers decrease and neuromast positions shifted caudally, as FGF signaling is reduced. Data are shown as means ± SD. (*P < 0.05; **P ≪ 0.001, Mann-Whitney U test). Scale bar in (A), 50 μm. prim, primordium.

Correct patterning of hair cell and support cell precursors, already established within the migrating primordium (2, 17, 18), also requires Fgf3 and Fgf10 function. The Atonal homolog atoh1a and Notch ligand deltaA are specifically expressed in sensory hair cell precursors, whereas notch3 is expressed in undifferentiated and support cells (18). Expression of all three genes was absent in fgf3–/–;fgf10-MO embryos, whereas fgf3+/–;fgf10-MO embryos exhibited intermediate phenotypes (fig. S6, A to I). Brief (3-hour) treatment with SU5402 almost completely abolished precursor gene expression (fig. S6, J to L).

Taken together, our results demonstrate that Fgf3 and Fgf10 are critical for both proto-neuromast organization and for proper specification of precursors within the migrating primordium. These processes might be separately regulated by FGF signaling or may be interdependent. To distinguish among these possibilities, we transplanted cells derived from homozygous hsp70:dn-fgfr1 embryos into the prospective pLL placode domain of wild-type embryos. When intracellular FGF signaling was blocked after heat shock, transgenic cells failed to differentiate into hair cells but were capable of incorporating into all neuromasts (fig. S7, A to C). These results indicate that hair cell precursors may directly require FGF signals to differentiate, whereas support cells do not. Alternatively, FGF signals may be necessary for general differentiation of progenitor cells into hair and support cells, a possibility that could be resolved as specific support cell markers become available. To test whether proto-neuromast organization might depend on differentiation of hair cell precursors, we examined embryos injected with atoh1a-MO (19). Rosette formation, as well as neuromast number and placement, were not affected by loss of atoh1a function (fig. S8, A to I), and expression of fgf10 and pea3 was not significantly altered (fig. S8, J to M). Our results suggest that FGF signaling acts independently to organize rosettes and then to specify differentiation of one or more cell types within each rosette.

Finally, we asked whether reintroduction of wild-type cells could restore neuromast production in Fgf3/10-deficient embryos. Transplantation of wild-type cells rescued migration defects and restored some neuromast deposition, with all rescued neuromasts containing some wild-type cells (Fig. 4, A and B). Hair cell differentiation was partially restored, with hair cells derived from donor or host in any combination (Fig. 4C). Rosette formation and deltaA expression within the primordium were also partially restored in the presence of wild-type cells (n = 7; Fig. 4, D and E). In four cases, we observed a single host-derived deltaA-positive cell that was always immediately adjacent to wild-type transplants (Fig. 4D, arrow), indicating that FGFs function as short-range signals within individual rosettes. Wild-type cells partially restored rosette formation with apical constriction and basally displaced nuclei (Fig. 4E). However, because randomization of wild-type cells (the only FGF source) in mosaic embryos caused irregular rosette formation, neuromast position and spacing were also randomized (Fig. 4F).

Fig. 4.

FGF signaling is necessary for rosette formation and proper segmental deposition. (A) Primordium migration and neuromast deposition (arrowheads) are partially rescued after transplantation of wild-type cells (rhodamine dextran, red) into embryos injected with fgf3- and fgf10-MO. Donor and host carry Et(krt4:EGFP)sqet20 transgene (green) (21) to mark the lateral line. (B) Contralateral control side. (C) Wild-type cells (red) restore hair cell differentiation [acetylated tubulin (AcTub), green] at 48 hpf in fgf3–/– embryos injected with fgf10-MO; nuclei are DAPI-stained (blue). Hair cells are derived from both wild-type donor (arrowheads) and fgf3–/–;fgf10-MO host. (D and E) Wild-type cells (fluorescein dextran, green) are sufficient to rescue deltaA expression [red, (D)] and rosette organization [(E), arrows]. (D) Arrow indicates deltaA-positive cell derived from fgf3–/–;fgf10-MO host, seen in four of seven embryos. (E) Same embryo as in (D). Wild-type cells restore apical constriction (arrows; pan-Cadherin antibody, magenta) and basally displaced nuclei (arrowheads) (Fig. 2A). (F) Distribution of neuromasts in 13 fgf3/10-MO embryos that received wild-type transplants [embryo 4 is shown in (A)] is randomized. Scale bars, 50 μm [in (A) and (D)] and 10 μm [in (C) and (E)]. Error bars in (F) indicate SDs.

We suggest that ultimate placement of neuromasts along the trunk (pLL segmentation) is a direct result of primordium organization into proto-neuromasts (fig. S9). Our data indicate that primordium organization and neuromast position depend on both the overall levels and the location of FGF signaling. Lower levels of FGF signaling, caused either by exposure to intermediate doses of FGF inhibitor or by injection of fgf10-MO into fgf3 heterozygotes, presumably delayed generation of new rosettes from the leading progenitor zone, which translated into altered neuromast position and number. Randomization of the FGF source in mosaic primordia resulted in sporadic rosette formation and subsequent random neuromast placement. We presume that rosette progenitors are still generated in the absence of FGF signaling, because relatively high proliferation levels still persist in the SU5402-treated embryos and fgf3/10 morphants (fig. S10). In support of this idea, primordium cells reorganize into rosettes after SU5402 removal (movie S5). Our study demonstrates that FGF signaling coordinates multiple processes during pLL primordium segmentation, including organization into polarized rosettes, formation of hair cell precursors, and regulation of migration. Control of primordium patterning and migration by the same FGF signaling pathway represents an elegant way to couple these two processes and would ensure that migration continues only as rosettes are renewed. Understanding how FGF signals are translated into polarity and patterning signals during pLL development may shed light on segmentation and morphogenesis in other organ systems.

Supporting Online Material

Materials and Methods

Figs. S1 to S10

Table S1


Movies S1 to S6

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article