The mechanosensory lateral line is an excellent model to study phenotypic diversity in the vertebrate nervous system (Dijkgraaf,1963; Coombs et al.,1988; Webb,1989). Although many behavioral and ecological factors correlate with various lateral line patterns, little is known about the developmental mechanisms underlying this variability (Pichon and Ghysen,2004). The lateral line is composed of mechanoreceptors, called neuromasts, which are distributed over the body surfaces of fish and amphibians. During zebrafish embryonic development, the placode-derived posterior lateral line primordium migrates caudally under the epidermis and deposits neuromast precursors or proneuromasts (Gompel et al.,2001). The proneuromasts integrate into the epidermal cell layer (Sapede et al.,2002) and differentiate into neuromasts, which are composed of sensory hair cells, support cells, and mantle cells (reviewed in Ghysen and Dambly-Chaudière,2007). During postembryonic development, the differentiated neuromasts migrate ventrally across the epidermal cells to their eventual destinations in the trunk region (Sapede et al.,2002). Subsequently, each founder neuromast gives rise to accessory neuromasts through a budding process, and forms a row of neuromasts, which is called a ‘stitch’ (Stone,1933; Ledent,2002; Ghysen and Dambly-Chaudière,2005). In contrast to the highly conserved embryonic patterns of neuromast distribution, the adult patterns show extensive interspecies variability (Sapede et al.,2002, Pichon and Ghysen,2004).
In the present study, we analyze the neuromast distributions on the caudal fins of teleosts. Caudal fins exhibit extensive morphological diversity among species (Goodrich,1930), and the lateral line pattern on the caudal fin also displays considerable interspecies variation (Coombs et al.,1988). In adult medaka, a single closely packed neuromast cluster is located in the medial region of the caudal fin (Fig. 1A,B; Ishikawa,1990). In contrast, adult zebrafish have multiple lines of neuromasts distributed over the caudal fin (Fig. 1C,D; Dufourcq et al.,2006). In the present study, we demonstrate that the lateral line patterns in these species are established through different developmental processes in relation to the patterning of the founder neuromasts and the diversity of accessory neuromast formation. In addition, we show that the diverse neuromast patterns observed in a variety of teleost species can be interpreted as variations and combinations of the patterns observed in medaka and zebrafish.
Differences in Lateral Line Patterns Between Medaka and Zebrafish
Spatial distributions of neuromasts on the caudal fin were compared between adult medaka and zebrafish using the hair cell marker DiAsp (Collazo et al.,1994). In medaka, up to 10 neuromasts constitute a single, closely packed cluster in the medial inter-ray region of the caudal fin (Fig. 1A,B; Ishikawa,1990). We refer to the cluster of neuromasts on the caudal fin of medaka as the caudal neuromast cluster (CNC) in the following experiments. In contrast, four lines of neuromasts extend along inter-ray spaces in zebrafish (Fig. 1C,D). The set of neuromasts that form on the caudal fin is referred to as the caudal lateral line (CLL, Dufourcq et al.,2006). To understand the mechanisms that regulate the differences between caudal neuromasts in medaka and zebrafish, we examined the embryonic and postembryonic development of neuromasts in both species.
The teleost caudal fin originates from a ventral region of the embryonic fin fold that will subsequently form the hypural bones and fin rays (Goodrich,1930; Ishikawa,1990; Ledent,2002). In medaka, a single pair of neuromasts was located symmetrically on the left and right sides of the prospective caudal fin region at 8.5 days postfertilization (dpf; Fig. 1E, difference in position of neuromasts between both sides of each embryo was less than 10 μm in length along the axis of the tail, n = 42). The neuromasts on the prospective caudal fin region are referred to as “terminal neuromasts” (Gompel et al.,2001; Ledent,2002; Pichon and Ghysen,2004). The neuromasts on the remaining part of the trunk and tail region are designated as “trunk neuromasts” in the following description. The terminal neuromast was precisely located in medaka (317 ± 12.8 μm from tip of the notochord, n = 84, Fig. 1F). These results contrast with the variable and asymmetrical distribution of the trunk neuromasts (Fig. 1E).
In zebrafish, on the other hand, two to four terminal neuromasts were observed on each side of the body (Fig. 1G, 2. 83 ± 0.57 on average at 5 dpf, n = 236; as previously reported, Gompel et al.,2001). In most cases (65%), there were three neuromasts on one side of the body, but a significant proportion had two or four neuromasts on one side (26% and 9%, respectively, n = 236). There was no correlation between the number of terminal neuromasts on the left and right sides of an individual (Fig. 1H, κ2 = 9.26, P > 0.05). The position of the most rostral terminal neuromast was more variable in zebrafish than that of the terminal neuromast in medaka (Fig. 1I, 249 ± 41.0 μm from tip of the notochord in embryos with three neuromasts, n = 131). Moreover, even when an embryo had the same number of terminal neuromasts on both sides, most of these neuromasts were located asymmetrically (Fig. 1J), resembling the pattern of distribution of trunk neuromasts in terms of asymmetrical localization (Fig. 1G). These results demonstrate that the terminal neuromasts in zebrafish are variable both in number and position, in contrast to the single pair of neuromasts that are located symmetrically in medaka.
Terminal Neuromast Is Established Prior to the Caudal Fin Formation in Medaka
We then examined the timing of determination of the position of the terminal neuromast during medaka development. The eye absent 1 (eya1) gene was used as a marker of proneuromasts before differentiation of hair cells. In zebrafish, eya1 mRNA is expressed in neuromasts and in their precursors (Sahly et al.,1999; Dufourcq et al.,2006). A partial cDNA fragment of the medaka eya1 gene (Koster et al.,2000) was used to isolate the remaining coding region, and we confirmed that this gene is orthologous to zebrafish eya1 (Fig. 2A). As predicted, eya1 mRNA was strongly expressed in medaka neuromast cells (Fig. 2B,C).
eya1 mRNA was symmetrically expressed on the left and right sides of the prospective caudal fin region at 3 dpf, before the morphological caudal fin structures became visible (Fig. 2D,D′). A ventral loop of the caudal vein was observed in the prospective caudal fin region at 4 dpf (Fig. 2E,F). The proneuromast, recognized as a spherical aggregate of eya1-expressing cells, was located near the ventral loop of the caudal vein (Fig. 2E′,F′). At 6 dpf, hair cells had differentiated and had become DiAsp-positive (Fig. 2G). The differentiated neuromast remained near the ventral loop of the caudal vein, which was encompassed by developing hypural plates H1 and H2 (Fig. 2H,H′; Ishikawa,1990). These results demonstrate that the terminal neuromast in medaka is precisely located before caudal fin structures, such as the vein and the hypural plates, become apparent.
Displacement of the Terminal Neuromast During Caudal Fin Formation in Medaka
At 9 dpf, shortly after hatching, the terminal neuromast moved dorsally, maintaining a consistent position relative to the hypural plates and the fin rays during caudal fin formation (Fig. 2I,J). The terminal neuromast retained the same position relative to caudal fin structures in the dorsoventrally symmetrical (homocercal) adult caudal fin (Fig. 1B), suggesting that the dorsal movement of the neuromast may be associated with morphological changes of the caudal fin structures. In the mutant Da (double anal fin), the epural bone adopts a hypural-like morphology (Ishikawa,1990), and the notochord fails to bend up (Fig. 2K, Ishikawa,1990). In the adult Da mutant, the terminal neuromast remains ventral to the medial region of the caudal fin, keeping its relative position to the hypural plates (Fig. 2K′; Ishikawa,1990). These results support our interpretation that the dorsal movement of the terminal neuromast is caused by the displacement of morphological structures during caudal fin formation.
Formation of the Accessory Neuromasts in Medaka
In adult medaka, the terminal neuromast developed into the CNC (Fig. 3A,A′), and the number of neuromasts per cluster was correlated with increasing body size (Fig. 3B). In amphibians and zebrafish, founder neuromasts give rise to rows of accessory neuromasts through a budding process (Stone,1933; Ledent,2002; Ghysen and Dambly-Chaudière,2005). The CNC in medaka may, therefore, also originate from the founder neuromast on the caudal fin. The neuromasts comprising the CNC were innervated by collaterals of the same axonal fascicle (Fig. 3A′), supporting this idea. In normal development, however, addition of new neuromasts is a relatively slow process, and is difficult to study. In the following experiments, we studied regeneration of the CNCs because the developmental process can be accelerated by eliminating the neuromasts (Dufourcq et al.,2006; López-Schier and Hudspeth,2006).
We removed the CNCs mechanically as described previously (Baker and Montgomery,1999; Fig. 3C). At 4 days after ablation, a spherical aggregate of cells appeared near the ablated site of the axonal fascicle (Fig. 3D, n = 9 of 12 samples). The cell aggregate resembled a proneuromast and expressed eya1 mRNA (Fig. 3E). Moreover, some of the cells in the aggregate were DiAsp-positive (Fig. 3F), confirming that the cell aggregate was a regenerated neuromast that contained differentiating hair cells. In some cases, a large neuromast was associated with a smaller neuromast carrying fewer hair cells, and eya1 mRNA was expressed in the area surrounding both neuromasts (Fig. 3G, n = 3 of 12 samples), suggesting that the smaller neuromast may be derived by budding of the founder neuromast. At 12 days after ablation, two to three small neuromasts emerged in close proximity to the presumptive founder neuromast, and these small neuromasts were innervated by collateral branches which bifurcated from the root of the axonal fascicle (Fig. 3H,H′, n = 8 of 8 samples). These results suggest that the CNC is composed of accessory neuromasts that originate from a single founder neuromast through a budding process. Further experiments are required to elucidate whether the same process underlies normal development of the CNC.
In adult medaka, 6–8 scales (7.43 ± 0.63 on average, n = 84) developed in the proximal region of the caudal fin (Fig. 3A) and the CNC was located close to the caudal margin of these scales (Fig. 3A′, see Fig. 3I for a schematic drawing).
Migration of the Terminal Neuromasts During Caudal Fin Formation in Zebrafish
In zebrafish, the terminal neuromasts migrate and repattern in the early stages of caudal fin formation (Ledent,2002). We performed a more detailed analysis of the final destinations of the terminal neuromasts, in experiments with embryos that had three terminal neuromasts on each side (designated T1–3, Fig. 1I), because this type was most frequently observed (Fig. 1H).
The earliest morphological sign of the caudal fin structure was the dispersal of melanocytes (Ledent,2002), which was followed by the formation of a ventral loop of the caudal vein at 9 dpf, similar to that in medaka at 4 dpf (Fig. 4A,A′, compare with Fig. 2F). The terminal neuromasts remained asymmetrically positioned between the left and right sides of the body at this stage (Fig. 4A). At a few weeks, when they reached approximately 4.4 mm in length, the hypural structures became visible as condensations of mesenchymal cells (Bird and Mabee,2003; Fig. 4B,B′). Intriguingly, the positions of the terminal neuromasts became more symmetrical as the caudal fin structures formed (Fig. 4C,D). Zebrafish develop five hypural bones (Bird and Mabee,2003; Fig. 4B′–D′); the anterior three (hy1–3) and the posterior two (hy4 and hy5) correspond to the H1 and H2 hypural plates in medaka, respectively (Ishikawa,1990). The ventral loop of the caudal vein was surrounded by hy3 and hy4, and this position coincided with the medial region of the adult caudal fin (Fig. 4E), as has been described in medaka (Fig. 2G–J; Ishikawa,1990). In approximately 5.0-mm fry, the middle of the three terminal neuromasts (T2) was located near the loop of the vein, at the prospective medial region of the caudal fin (Fig. 4E′). We quantified the position of the terminal neuromasts after caudal fin formation, and showed that T2 was located precisely in the medial region of the caudal fin (86%, n = 130), while the remaining two neuromasts (T1 and T3) were less precisely located in dorsal and ventral regions of the caudal fin (Fig. 4G,J). Thus, the positions of terminal neuromasts become determined during caudal fin formation at least partly by neuromast migration in zebrafish, in contrast to the displacement of the terminal neuromast in medaka (Fig. 2G–J).
Accessory Neuromast Formation of Terminal Neuromasts in Zebrafish
Soon after caudal fin formation, when zebrafish fry were approximately 7.0 mm in length, additional neuromasts appeared in distal regions of the caudal fin to form the caudal lateral line system (CLL, Fig. 4G). CLL neuromasts form lines that extend in inter-rays spaces (Fig. 1C, D, Dufourcq et al.,2006). In the adult caudal fin, there were two to four CLL lines on each side (3.85 ± 0.53 on average, n = 708). In the majority of the cases (83%), four lines were present at stereotypic positions on the caudal fin (designated as C1–4, Fig. 4G,K); thus subsequent experiments were conducted on animals with four lines of the CLLs. The four lines were called C1 to C4, from dorsal to ventral.
To identify the origin of CLL neuromasts, we first observed axonal projections. The Isl1-GFP transgenic line expresses green fluorescent protein (GFP) in the efferent lateral line nerve (Higashijima et al.,2000). In adult Isl1-GFP fish, GFP is also expressed in neuromast cells (Dufourcq et al.,2006). We observed that some of the cells surrounding the DiAsp-positive hair cells also expressed Isl1-GFP, and that the signal was mainly restricted to the apical region of the neuromast (Fig. 4F). Support cells possess fine processes that are associated with the apical surface of the skin, and these processes are attached to the hair cells (Owens et al.,2007). Thus, GFP appears to be expressed in support cells, and possibly in hair cells as well (Dufourcq et al.,2006). Isl1-GFP expression in neuromast cells started when the fish were approximately 7.0 mm in length. However, isl1 mRNA expression in the neuromast was not detected by in situ hybridization (data not shown). Thus, it is likely that Isl1-GFP expression in the neuromast does not recapitulate endogenous isl1 gene expression, but rather reflects an artifactual effect of the transgene.
We demonstrated that the efferent nerve innervating the dorsal neuromast T1 extended two collateral axons to the first neuromasts of the C1 and C2 lines (designated as C1-1 and C2-1, respectively, Fig. 4H,I). Similarly, the efferent nerve innervating the ventral neuromast T3 extended two collateral axons to the C3 and C4 neuromasts, whereas the medial neuromast T2 did not extend any axonal collaterals (data not shown). These results suggested that the C1/2 and C3/4 lines may be derived from T1 and T3, respectively. To test this hypothesis, we analyzed the behaviors of T1 and T2 after removal of the CLL neuromasts. When the caudal fin is amputated, the most distal remaining CLL neuromast gives rise to new neuromasts on the regenerated fin, recapitulating normal development (Dufourcq et al.,2006). We amputated the caudal fin distal to the T1-T3 neuromasts when the CLL neuromasts appeared (Fig. 5A). At 3 days after amputation, the caudal fin regenerated (Fig. 5A). In the region distal to T1 and T3, an aggregate of cells resembling an enlarged proneuromast emerged (Fig. 5B,B′, n = 12 of 12 samples). Some of the cells in the aggregate were DiAsp-positive (Fig. 5C), and the cell aggregate was followed by an axonal process of the lateral line efferent nerve (Fig. 5C′). Confocal microscopy revealed that the DiAsp-positive cells were centrally located within the cell aggregate (Fig. 5D). Moreover, the eya1 mRNA was expressed in the T1 and T3 neuromasts and the presumptive proneuromasts (Fig. 5E). As expected, some of the T1 or T3 neuromasts restored two proneuromasts (Fig. 5F). All of these results suggest that the newly formed neuromasts of the CLLs are accessory neuromasts originated from the T1 and T3 neuromasts by a budding process.
In adult zebrafish, three scales are vertically aligned in the caudalmost part of the body, covering the joint of the body and the caudal fin at the hypural region. Scales in the dorsal and ventral regions are much larger than that in the medial region (Fig. 6A, n = 90). The position of T1-T3 neuromasts coincided with the proximal region of each scale (Fig. 6A). T1-T3 neuromasts on the scale gave rise to rows of neuromasts that elongated dorsoventrally in the proximal region of the scale (Fig. 6B). These vertical rows of neuromasts closely resembled the rows of neuromasts on the body surface (stitches; Ledent,2002), where each stitch corresponds to each scale (Fig. 6C; Metcalfe,1989). The number of neuromasts per row increased on the scale of the proximal region of the caudal fin (Fig. 6D) by addition of neuromasts at the distal end of the row (Fig. 6B), suggesting that these were also accessory neuromasts formed by a budding process as described in the trunk region (Ledent,2002).
As we described, a significant proportion of embryos possessed two (26%) or four (9%) terminal neuromasts (Fig. 1H), and in these cases each neuromast was also in a position that corresponded to one of the three scales in adult fish (Fig. 6E,F). This suggests that the scales or the scale precursors may play an important role in the positioning of terminal neuromasts during postembryonic development, or that both neuromasts and scales depend on a common patterning factor. Moreover, the caudal fin in which the neuromast was absent at the ventral T3 position did not have the C3 or C4-CLL neuromasts (n = 12 of 130 fin sides, Fig. 6E), consistent with our results showing that C3 and C4-CLL neuromasts originated from T3. As described above, the majority (83%) of adult caudal fins had four lines of the CLLs on one side, although 26% of the prospective caudal fins had only two terminal neuromasts. This apparent inconsistency is due to the fact that one of the two terminal neuromasts became always located at the dorsal T1 position after caudal fin formation, and that in a majority of these cases the second terminal neuromast is found at the T3 position. Out of 25 fin sides with two terminal neuromasts, 10 sides had neuromasts at the T1/T2 positions, and 13 sides had neuromasts at the T1/T3 positions.
CLL neuromasts increased in number in a growth-dependent manner (Fig. 6D) by serial addition of neuromasts, with each newly formed neuromast being generated by budding from the distal-most CLL neuromast (Fig. 6G,G′,H,H′). In summary, both the lines of CLL neuromasts and the vertical rows of neuromasts on the scales of the caudal fin originate from the terminal neuromasts in a growth-dependent manner (Fig. 6I for a schematic drawing) as seen with CNC formation on the medaka caudal fin (Fig. 3I).
Diverse Neuromast Patterns on the Caudal Fin in Various Teleost Species
Finally, we asked whether the CLL patterns seen in medaka and zebrafish were also observed in other teleost species. We showed that seven related species of medaka (belongs to the genus Oryzias) had a CNC similar to that in medaka (Oryzias dancena, Fig. 7A, O. luzonensis, O. mekongensis, O. celebensis, O. marmoratus, O. minutillus, and O. javanicus, data not shown), indicating that this pattern is conserved in this taxonomic group. Dermogenys pusillus (wrestling halfbeak), which belongs to the same order Beloniformes as Oryzias, possessed a single line of the CLL neuromasts in the center of the caudal fin (Fig. 7B). Newly born D. pusillus fry had an embryonic pattern identical to that of Oryzias (Fig. 7B), suggesting that the CNC in Oryzias and the medial line of CLL neuromasts in other species may be homologous structures.
In two species from different taxonomic groups, Poecilia reticulata (Cyprinodontiformes, guppy, Fig. 7C) and Takifugu niphobles (Tetraodontiformes, puffer fish, Fig. 7D), there was no CNC or CLL on the caudal fin. In P. retuculata, we showed that each of the two terminal neuromasts remained as a single neuromast through postembryonic development (Fig. 7C). Interestingly, the same is true for the trunk neuromasts: in P. reticulata, no accessory neuromast formation (stitch) is observed anywhere on the body surface (Fig. 7C,D). The same is true in T. niphobles, where neuromasts remain single both on the caudal fin and on the body.
Among the species that present CLL lines on their caudal fin, the number of lines was variable among species: one in D. pusillus (Fig. 7B), two in Tridentiger trigonocephalus (Perciformes) and Misgurnus anguillicaudatus (Cypriniformes, Fig. 8A,C), four in zebrafish, and eight in Leucopsarion petersi (Perciformes) and Carassius gibelio (Cypriniformes, Fig. 8B,D). The number of CLL lines may be determined by the number of terminal neuromasts in embryonic stages and/or by the number of buds originating from each terminal neuromast. Intriguingly, the number of lines of the CLLs does not always reflect phylogeny, but rather appears to reflect ecology. T. trigonocephalus (chameleon goby), which resides in ocean shallows, possessed only two lines of the CLL neuromasts (Fig. 8A). In contrast, the closely related species L. petersi (ice goby, referred to as a neotenic goby; Kon and Yoshino,2002), which belongs to the same family Gobiidae as T. trigonocephalus, is an active swimmer and possessed eight lines of the CLL neuromasts (Fig. 8B). Similarly, M. anguillicaudatus (weather loach, Fig. 8C), a bottom-dwelling fish in rice fields, had fewer lines of the CLLs than the related active swimmer C. gibelio (goldfish, Fig. 8D).
Possible Mechanisms for Neuromast Migration Through the Epidermis
We showed that the terminal neuromasts reach their final position before their differentiation in medaka and after their differentiation in zebrafish. The lateral line primordium migrates under the epidermis (Gompel et al.,2001), whereas the differentiated neuromasts integrate into the epidermis and migrate across the epidermal cells (Sapede et al.,2002). Primordial migration is partly regulated by cxcr4/sdf-1 signaling (David et al.,2002; Li et al.,2004; reviewed in Ghysen and Dambly-Chaudière,2007). However, the mechanisms underlying neuromast migration through the epidermis are not known. In the neuroepithelium of the developing hindbrain, the differentiated facial motor neurons migrate across the neuroepithelial cells, in a process that resembles neuromast migration through the epidermis. Previously, we showed that the Wnt/noncanonical signaling pathway functions in the neuroepithelium to regulate motor neuron migration (Wada et al.,2005,2006). The Wnt/non-canonical signaling pathway also functions in the epidermis, and controls the orientation of hair follicles in the mouse (Guo et al.,2004).
We showed that the final position of each terminal neuromast coincided with the position of a scale in the proximal region of the caudal fin in zebrafish. The prospective scale region may emit chemoattractive factors that regulate neuromast migration. Although the teleost elasmoid scales and hairs in mammals are not homologous structures, both require the ectodysplasin (EDA) pathway for development (Kondo et al.,2001; Sharpe,2001). In mammals, many signaling molecules, including Wnts, Fgfs, Tgfβs, and Shh, are expressed in developing hair precursors (reviewed in Pispa and Thesleff,2003). These diffusible factors are potential chemoattractive cues for neuromast migration (reviewed in Charron and Tessier-Lavigne,2005). To date, only Shh expression in scales has been examined in zebrafish, and shh mRNA is expressed in the posterior margin of the developing scales (Sire and Akimenko,2004).
In contrast to neuromast migration in the trunk region of zebrafish, the trunk neuromasts differentiate only after they have reached their final destination in medaka (Sapede et al.,2002). These observations are similar to our results showing that terminal neuromasts differentiate before their migration in zebrafish, whereas the terminal neuromasts of medaka differentiate after they reach their final position. Similarly, in zebrafish, the CLL neuromasts differentiate and integrate into the epidermis soon after they are derived from the founder neuromasts (Fig. 5B–E), and they continue to migrate distally away from the founder neuromasts (Fig. 4F), which suggests that CLL neuromasts are also capable of migrating across epidermal cells. In contrast, the CNC in medaka does not extend distally on the caudal fin, but remains tightly packed in proximity to the founder neuromast (Fig. 3H). These results raise the possibility that neuromast migration through the epidermis is lost across the entire body surface of medaka.
Developmental Processes Underlying Diverse Neuromast Patterns on the Caudal Fin
We examined the patterns of neuromasts on the caudal fin of medaka and of zebrafish. We observed that the massive differences between the two patterns can be reduced to a small number of developmental differences between the two systems. First, medaka has only one terminal neuromast, whereas the zebrafish has several (two to four). Second, the terminal neuromast of medaka occupies a fixed position relative to prospective fin structure before it differentiates, whereas the neuromasts of zebrafish form in variable numbers and at ill-defined positions, and reach their final position relative to fin structure long after they have differentiated. Third, differentiated neuromast can migrate across the epidermis in zebrafish, but apparently not in medaka. Finally, in both medaka and the zebrafish founder neuromasts can bud off accessory neuromasts, but this capability is not present in other teleost species. We propose that each of these differences points to a variation in a developmental process, and that changes in the strength and combination of these processes can account for large differences in CLL pattern.
Figures 7 and 8 reveal changes in CLL pattern that correspond to changes in a single process. For example, the capability of differentiated neuromasts to migrate across epidermis would account for the difference between the closely related species, O. dancena and D. pusillus (Fig. 7A,B). Likewise, the absence of accessory neuromasts may account for the dramatic differences between the CLL patterns of zebrafish on one hand, and of P. reticulata or T. niphobles (Fig. 7C,D) on the other hand. Of interest, the latter two species also lack stitches on their body surface, suggesting a general loss of the capability to form accessory neuromasts in both species. Loss of stitch formation has also been described for amphibian species (Schlosser,2002), which indicates that “stitch-less” phenotypes have appeared repeatedly during evolution. Simple variations on the zebrafish pattern may account for other types of CLL (Fig. 8). Of interest, patterns that correspond to one species may occasionally appear as rare variants among another species, for example, the pattern in Figure 8A resembles a pattern that is occasionally observed in wild-type zebrafish (Fig. 6E). We do not know, however, whether these two patterns are truly the result of similar developmental program, or display a fortuitous similarity.
After amputation of the caudal fin, mantle cells at the distal margin of the remaining neuromast start to divide and form a bud (Duourcq et al.,2006). Thus, the mantle cells are probably the source of the accessory neuromasts. The neuromast is maintained by the continuous renewal of hair cells, which are regenerated from dividing mantle cells (reviewed in Ghysen and Dambly-Chaudière,2007). Therefore, similar mechanisms may regulate accessory neuromast formation and neuromast renewal. We show that the number of neuromasts in the CNC correlates with increasing body size in medaka (Fig. 3B). In O. dancena, the CNC was composed of up to 24 neuromasts (Fig. 7A), whereas the CNC in medaka animals of similar body size was composed of 6–8 neuromasts (Fig. 3B), suggesting a markedly higher frequency of neuromast budding in O. dancena than in medaka. These results indicate that the frequency of neuromast budding is differentially regulated in different species, bringing a further element of diversity into lateral line development and patterning.
Functional Significance of the Evolution of Neuromast Patterns on the Caudal Fin
The caudal fin contributes to the generation of propulsive forces during locomotion. Given that caudal fin morphology varies significantly between species, fin neuromasts are exposed to different hydrodynamic stimulations depending on fin shape and locomotive activity (Lauder,1989). Our data suggest that the number of CLL lines on the caudal fin reflects ecology rather than phylogeny (Fig. 8), suggesting that diverse neuromast patterns may reflect functional constraints.
Alternatively, different lateral line patterns may not be adaptive morphologies, but responses to different developmental constraints (Webb,1989). In medaka, the caudal structure (ventral loop of the caudal vein) first appears at 4 dpf before hatching (Fig. 2F), whereas in zebrafish, the corresponding structure appears at 9 dpf, long after hatching (Fig. 4A′). Moreover, the medaka caudal skeleton comprises two hypural plates (H1 and H2; Fig. 2H; Ishikawa,1990), while the zebrafish comprises five hypural bones (hy1–5; Fig. 4D′; Bird and Mabee,2003). Changes in developmental timing (heterochrony) of caudal fin formation or morphological changes in caudal skeletons may affect the lateral line patterning of the caudal fin during evolution.
Obviously the two types of explanation, functional adaptation or developmental constraints, need not be exclusive, and are probably complementary. We have showed that wide variations in pattern may reflect changes in a small number of developmental processes, and it seems likely that the inventory of developmental processes involved in lateral line patterning puts limits on the types of changes that can be introduced in CLL patterning. Inversely, it seems equally likely that each of the patterns illustrated Figures 7 and 8 must be functionally appropriate for the survival of the species.
Medaka (Oryzias latipes) T5 strain (Shimada and Shima,2004) was used. Because the T5 strain carries the lf (leucophore free) locus and lacks fluorescent pigments (Wada et al.,1998), neuromasts can be easily visualized. The Da (double anal fin) strain was originally identified by Tomita (Tomita,1969; Ishikawa,1990). Da/Da homozygous fish were used. All of the strains and embryos were kept at 27°C. The zebrafish (Danio rerio) Isl1-GFP transgenic line (Higashijima et al.,2000) was maintained at 28.5°C. Embryos were staged by days postfertilization (dpf). Fry were staged by body length (mm), because size is a better indicator of skeletal development in the postembryonic stages (Bird and Mabee,2003). Fish species used for analysis of variation in neuromast patterns were described by conventional taxonomy (Nelson,2006). Seven medaka-related species, Oryzias dancena, O. luzonensis, O. mekongensis, O. celebensis, O. marmoratus, O. minutillus, and O. javanicus, were used (Takehana et al.,2005). Dermogenys pusillus (wrestling halfbeak) and Poecilia reticulata (guppy) were obtained from local pet shops. All of the fish were kept at 27°C. Newly born fry of D. pusillus and P. reticulata were used. Leucopsarion petersi (ice goby) was obtained from a local fishery and maintained at the Niigata City Aquarium (Chuo-ku, Niigata, Japan). Tridentiger trigonocephalus (chameleon goby) and Takifugu niphobles (puffer fish) were collected from the seashore in Niigata (Nishi-ku, Niigata, Japan). Misgurnus anguillicaudatus (weather loach) and Carassius gibelio (goldfish) were collected from a rice field in Niigata (Shirone, Niigata, Japan). At least three samples of each species were analyzed and the representative one was shown in the figures.
Isolation of the Medaka eya1
The 3′ region of the medaka eyes absent 1 (eya1) cDNA had already been isolated (Koster et al.,2000). We isolated the remaining 5′ region by RT-PCR. Total RNA was extracted from 8.5 dpf-embryos of the Hd-rR strain (Hyodo-Taguchi,1980; Wada et al.,1998) using an RNA extraction kit (Qiagen). A partial cDNA fragment was amplified with a first strand cDNA synthesis kit (Takara) and PCR using specific primers (5′-ATGGAAATGCAGGATCTAGCC-3′ and 5′-GTTGTTACTCGTCATGTAGGG- 3′) designed from the Sanger Centre genome database (http://www.ensembl.org/index.html). The PCR product was cloned into pCRII-TOPO (Invitrogen), and sequenced using a BigDye terminator cycle sequence kit (PE Applied Biosystems) with a DNA sequencer (ABI PRISM/310). The amino acid sequence of Eya1 was deduced from the nucleotide sequences of the isolated cDNA fragment (accession no. AB370342) and published data (CAB65317; Koster et al., 2000). The phylogenetic relationship between Eya proteins was analyzed using the CLUSTAL W program at the DDBJ (http://www.ddbj.nig.ac.jp/search/clustalw-e.html). The amino acid sequences were deduced from the following data: Drosophila eya (AAA28310), mouse eya1 and eya4 (NP_034294 and NP_034297, respectively), and zebrafish eya1 (AAD25366).
Neuromast Staining and In Situ Hybridization
Neuromast hair cells were labeled by incubating live fish in 2 μm/ml 4-(4-diethylaminostyryl)-N-methylpyri- dinium iodide (DiAsp, Sigma; Collazo et al.,1994). Fish were kept in the solution for 30 min and then rinsed briefly. Fish were anaesthetized in 0.17 mg/ml 3-aminobenzoic acid ethylester (tricane, Sigma) before observation (Westerfield,2000). In situ hybridization using RNA probes for medaka eya1 (Koster et al.,2000) and zebrafish eya1 (Sahly et al.,1999) were performed as described (Westerfield,2000). At least five embryos were monitored in each experiment. Images were captured using a fluorescent dissecting microscope (Leica MZ APO) and a differential interference contrast (DIC) microscope (Olympus BX-51WI) with a CCD camera (Nikon DXM1200F), and a confocal microscope (Olympus FV300). Some of the fluorescent color images were converted to negative and monochrome using Adobe Photoshop software.
Ablation of Neuromasts in Medaka
The CNC on the caudal fin was removed as described (Baker and Montgomery,1999) with modification. Each adult medaka (20.5–26.0 mm, n = 10, carrying 4–6 CNCs) was stained with DiAsp, anesthetized and placed on a plastic dish. Under fluorescent dissecting microscope, the DiAsp-positive neuromasts were carefully removed using a sterile needle (0.45 mm diameter, Terumo, see Fig. 3C). Operated fish were kept in egg water (60 mg/ml “Instant Ocean” Sea Salts; Westerfield,2000). One day after ablation, the fish was stained with DiAsp again to confirm that all of the hair cells had been completely removed. At 4 or 12 days after ablation, recovery of the neuromasts was analyzed. Samples were fixed and analyzed by in situ hybridization using medaka eya1 RNA probe.
Amputation of Caudal Fins in Zebrafish
Amputation of the zebrafish caudal fins was carried out as previously described (Dufourcq et al.,2006). Juvenile fish shortly after the caudal fin formation (7.8–8.7 mm, n = 12) were stained with DiAsp. Each fish was mounted in 3% methyl cellulose on a cover slip (Westerfield,2000). The caudal fin was amputated using a sterile needle at the level between the terminal neuromasts and the CLL neuromasts (see Fig. 5A). Operated fish were analyzed 3 days after amputation. Samples were fixed and analyzed by in situ hybridization using zebrafish eya1 RNA probe.
Neuromast and Scale Counting
To count neuromasts, fish stained with DiAsp were observed. The position of the terminal neuromasts was quantified as the distance from tip of the notochord in 8.5 dpf medaka (n = 84, both sides of the 42 embryos were analyzed) and 5 dpf zebrafish (n = 131, in the case of three neuromasts on one side). Numbers of terminal neuromasts were counted in 5 dpf zebrafish (n = 236, 118 embryos). Correlation in neuromast number between left and right sides of the embryos was analyzed using the same samples (118 embryos). Of 118 embryos, 53 possessed three neuromasts on each side. The ratio of the symmetrically located neuromasts (difference in position is less than 10 μm in length between left and right sides of the embryos) was analyzed using these embryos. The number of neuromasts per CNC was counted in medaka (3.8–37.5 mm, n = 154, 77 fish). Number of the C1-CLL neuromasts and T1-associated neuromasts on the scale (see Fig. 3R) was counted in zebrafish (8.8–23.5 mm, n = 100, 50 fish). Positions of the terminal neuromasts (T1-3) and the CLLs (C1-4) were quantified by the relative position to the fin rays (Fig. 3H; Dufourcq et al.,2006). Frequency of appearance in positions for the terminal neuromasts and the CLLs was analyzed using juvenile fish (7.2–13.2 mm, n = 130, 65 fish) and older fish (11.5–25.0 mm, n = 708, 354 fish), respectively. The number of the scales in the proximal region of the caudal fin was counted in medaka (20.5–26.0 mm, n = 84, 42 fish) and in zebrafish (19.0–25.0 mm, n = 90, 45 fish). Data were expressed as mean ± SEM.
We thank J. Wittbrodt and C. Petit for in situ probes; A. Shimada and A. Shima for the medaka T5 strain; S. Higashijima and H. Okamoto for the zebrafish Isl1-GFP line; K. Yamahira for D. pusillus; members of Niigata City Aquarium for L. petersi, and for support throughout this project; M. Iwasaki for collecting wild fish; and T. Mitsui for confocal microscopy. We acknowledge Y. Ishikawa and T. Mukai for helpful discussions. We greatly thank the anonymous reviewer for many valuable comments and suggestions on the manuscript. H.W. was funded by the Grant-in-Aid for Young Scientists (B) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.