Antagonistic interactions in the zebrafish midline prior to the emergence of asymmetric gene expression are important for left–right patterning

(b) RNA and morpholino injections

Capped mRNA transcripts were synthesized using Ambion mMessage mMachine™ kits and quantitated by UV spectrophotometry. oep mRNA was transcribed from plasmid pJZoepFlag1–2 linearized with SmaI [31]. Morpholino antisense oligos were obtained from GeneTools, LLC and resuspended to a stock concentration of 50 µg µl⁻¹ in dH₂O. The ntla morpholino sequence we used in these studies is described in [94]. The lefty1 morpholino used is described in [95]. RNA and morpholinos were diluted in 5 mg ml⁻¹ phenol red dissolved in 0.2 M KCl to concentrations that would deliver the desired amount in 100–500 pl injection volumes. Injection volumes were calibrated by micrometer as described [9]. RNA and morpholinos were microinjected into dechorionated embryos at the 1–2 cell stage. Concentrations injected per embryo were as follows: oep RNA 50 pg, ntla morpholino 200 or 400 pg, lefty1 morpholino 250 pg and control morpholino 250 pg. Note, these experiments were conducted in early 2000s and the preparation of morpholinos at that time allowed for clean phenocopies at very low concentrations. Today, we typically use 2 ng or more of these same morpholinos to produce the same effect. To generate LZoep embryos, oep RNA was injected into 1–2 cell stage embryos derived from crossing adults homozygous for the oep^tz257 mutation.

(c) Shield transplantations and ablations

To label donor cells, donor embryos were injected with lysine fixable fluorescein dextran 10 000 MW dye (Molecular Probes) resuspended at 5% w/v in 5 mg ml⁻¹ phenol red dissolved in 0.2 M KCl. For transplantation, donor embryos were injected with 500 pl fluorescein dextran alone or containing the appropriate amount of RNA or morpholino as described above. Shield transplantations and ablations were done as described [96]. Transplantations and ablations were performed in either 1× Danieau as described [96], or in Hank's balanced salt solution. Any transplanted embryos that did not develop normally were omitted from analysis. For example, embryos with abnormal bends or kinks in the axis, abnormal extension of the axis, or embryos with any degree of cyclopia were omitted. In embryos containing shield transplants, labelling of some dorsal spinal cord neurons was occasionally observed [97], but any embryos with substantial labelling outside of axial mesoderm derivatives were omitted from the analysis. Shield-ablated embryos were analysed by in situ hybridization for both pitx2 [33] and for sonic hedgehog (shh) [98] transcripts. Only shield-ablated embryos that seemed to develop normally (see above), but yet had gaps of shh expression in the notochord and floorplate, were included in the analysis.

(b) oep in the midline of LZoep embryos can restore pitx2 expression

The loss of pitx2 LPM expression in LZoep embryos is owing to insufficient oep activity. To determine the spatial requirement for oep in pitx2 expression in the LPM, we generated genetic chimeras by transplantation. Asymmetric pitx2 expression in the LPM is first observed at 10 somites in wild-type embryos (figure 2f) and is present through 26 h post fertilization (hpf). During somitogenesis, oep is expressed in the left and right LPM, in the developing dorsal diencephalon and in derivatives of axial mesoderm including the midline notochord, overlapping with lefty1 expression ([31]; figure 2d,g). As Oep is required for Nodal signalling, and pitx2 is a known target of Nodal signalling, it is conceivable that oep is predominantly required in the LPM to allow Nodal ligand to induce pitx2 to be expressed at wild-type levels. To test this possibility, we first transplanted wild-type cells into 1000-cell stage LZoep embryos. Transplantation at this early stage places wild-type cells in a scattered fashion throughout many tissues in the embryo. LZoep embryos with wild-type cells in the LPM were obtained, but rescue of pitx2 levels was not observed (n = 10). Large numbers of cells can be placed more accurately into the LPM by performing margin transplants at the 50% epiboly stage. Despite having large numbers of wild-type cells in the LPM, but not in other tissues, 10/14 showed no pitx2 expression in the LPM and 4/14 showed weak expression of pitx2 similar to LZoep controls. This suggests that providing Oep in the LPM does not allow for Nodal signalling to occur in an embryo otherwise lacking Oep.

Open in a separate window

Figure 2.

Expression of one-eyed pinhead, lefty1 and pitx2. (a–c) Lateral views of wild-type embryos at the 10 somite stage. Anterior is up; dorsal to the right. Lines and lower case letters mark the level of the embryo shown in (d–g). (d–f) posterior views and (g) dorsal view of the embryos in (a–c). Anterior is up; left (L) and right (R) are indicated. oep is expressed in the dorsal diencephalon (asterisk in a), in the midline (black arrowhead in d,g) and in the LPM (white arrowheads d,g). At this stage, oep is expressed in thin stripes in the posterior LPM (d) and more broadly in more anterior parts of the LPM (g). lft1 is expressed in a small portion of the anterior prechordal plate (dot of expression in the head region of b) and in the posterior midline (black arrowhead in e). The posterior notochord expression of lft1 is observed even earlier at the 3 somite stage, prior to when the first asymmetric expression of pitx2 is observed. pitx2 is expressed in head mesoderm (asterisk in c) and in presumptive Rohon-Beard neurons in the spinal cord. The first asymmetric expression of pitx2 is evident at this stage (white arrowhead in f) to the left of the midline (black arrowhead). Note the view in (f) is slightly displaced to those in (d,e) to better view the asymmetric expression of pitx2. The dark shading in (f) is from the expression in head mesoderm (asterisk in a) directly opposite the plane of view. (h) Expression of lft1 in an LZoep embryo at the 10 somite stage. lft1 expression can vary from weak to undetectable (10/13; as shown in (h), indicated by black arrow), to almost wild-type in intensity (3/13). As lft is induced by Nodal, the loss of late Nodal signalling in LZoep embryos may prevent the full induction of lft1 observed in wild-type embryos. Accordingly, more embryos express lft1 at the 3 somite stage (6/12 wild-type levels; 6/12 weaker levels). (i) Lateral view of lft1 expression in an 18–20 somite stage wild-type embryo. Arrows mark expression domains in the diencephalon, the heart field and the posterior notochord [82,84]. (j) Lateral view of lft1 expression in an LZoep embryo at the 18–20 somite stage. Note only the expression domain in the posterior notochord (black arrow) is observed in LZoep embryos (5/8 wild-type to reduced levels; 3/8 no obvious expression).

As oep is also expressed in the midline during the stages when pitx2 is first expressed in the LPM (figure 2d,g), oep may be required in the midline to affect pitx2 expression. We tested this hypothesis by placing wild-type cells into the midline of LZoep embryos using shield transplantation (figure 3a) [96,97]. The zebrafish shield is equivalent to the organizer in Xenopus and gives rise to axial mesoderm and floor plate. In these experiments, the shield of the host embryos was removed and replaced by the shield from the donor. This type of transplantation places donor cells throughout the floor plate and axial mesoderm derivatives, including the hatching gland, head mesoderm and notochord (figure 3b). We did not observe transplanted cells in the dorsal forerunner cells or in KV with this procedure (data not shown). Transplantation of a wild-type shield into a wild-type host does not strongly affect the levels or the asymmetric expression of pitx2 in the LPM (figure 3c,e,f), indicating that this form of transplantation does not robustly impact normal L–R patterning per se. While strong levels of pitx2 were never observed in LZoep controls, wild-type shields transplanted into LZoep embryos were capable of restoring pitx2 expression to strong levels, similar to wild-type, in the LPM in a proportion of embryos (figure 3d,e,f). To further ensure that this result is not simply a consequence of the transplantation technique, shields from LZoep embryos were transplanted into LZoep hosts. These transplanted embryos expressed weak levels of pitx2 similar to LZoep controls, but strong levels of expression were not observed (figure 3e). These results suggest that restoring oep expression in the midline can promote pitx2 expression in the LPM in LZoep embryos. Given that pitx2 is a bona fide Nodal transcriptional target, and Nodal requires EGF-CFC proteins to function, we assume any pitx2 expression in the LPM requires that residual Oep protein supplied via injection is still present.

Open in a separate window

Figure 3.

Shield transplantation experiments in wild-type and LZoep embryos. (a) Illustration of the experiment. A needle is used to remove the shield from a donor embryo labelled with fluorescein dextran tracer dye. The shield from the host embryo is removed and replaced with the shield from the donor. (b) Lateral view of a wild-type embryo after a wild-type shield transplant. pitx2 expression is in blue, labelled cells from the donor shield are in brown. Note that the floorplate (black arrow) and the notochord (black arrowhead) can be labelled with this technique. Labelled cells were not observed in the dorsal forerunner cells or in KV (data not shown). Dorsal views of pitx2 expression in a wild-type embryo with a wild-type shield transplant (c) and an LZoep embryo with a wild-type shield transplant (d) at the 18 somite stage. Anterior is up; left (L) and right (R) are indicated. (e) Bar graph representing pitx2 expression in shield transplanted embryos. The genotype of the host is listed first, followed by the genotype of the donor. Expression categories and colour codes are described in figure 1. (f) Statistical significance of the experiments was determined using χ² and Fisher's exact tests. Comparisons of strength of pitx2 expression and the sidedness of pitx2 expression were conducted. For sidedness, only embryos that displayed expression were included in the analysis, reflected by the change in n values. Blue denotes p-values of less than 0.01, orange denotes p-values between 0.01 and 0.05, and light orange denotes p-values greater than 0.05.

To determine if a particular region of the midline must express oep to activate pitx2 in the LPM, we monitored where transplanted cells were located. We did not find any correlation between the position of wild-type cells in the midline at the 18 somite stage and rescue of pitx2 expression in the LPM of LZoep embryos. For example, some LZoep embryos with strong expression of pitx2 had wild-type cells only in the most anterior floorplate cells, while others had wild-type cells throughout all axial mesoderm derivatives (data not shown). Intriguingly, while placing wild-type cells into the midline of LZoep hosts could restore pitx2 expression to strong levels in some cases, it did not correct the randomization of pitx2 expression seen in LZoep embryos (see §4).

(c) Blocking ntla activity in LZoep embryos can restore pitx2 expression

Our transplantation experiments suggest that the midline-localized oep has an active role in promoting pitx2 expression in the LPM. This is in apparent contrast with models that support a more passive role for the midline (see §1). To further test the function of the midline in L–R patterning, we removed midline structures from LZoep embryos using morpholino antisense oligonucleotides (MO) against no tail (ntla; the zebrafish Brachyury/T orthologue) [99]. ntla mutants and morphants lack the notochord, have expanded floorplate, have a substantially reduced KV and display bilateral expression of pitx2 and other left side-specific genes [29,34,35,77,78,100–103]. Injection of the ntla MO in wild-type or LZoep embryos results in 100% of the injected embryos displaying the morphological ntla phenotype. In wild-type embryos injected with ntla MO, pitx2 is expressed bilaterally as is seen in ntla mutants (figure 4c). Surprisingly, injection of ntla MO into LZoep embryos can restore expression of pitx2 in the LPM to strong levels in many cases; while LZoep mutants never exhibited strong LPM pitx2, 30% of LZoep mutants injected with ntla MO showed strong levels (figure 4d–g). Unlike injection of ntla MO, injection of a control MO does not dramatically alter pitx2 gene expression levels in wild-type or LZoep embryos (figure 4f,g). The most parsimonious explanation for these results is that the notochord and/or KV in LZoep embryos represses pitx2 transcription in the LPM, and that this repression can be released by blocking ntla function and removing midline structures.

Open in a separate window

Figure 4.

no tail (ntla) morpholino (MO) injections in wild-type and LZoep embryos. Lateral views of 20 somite stage wild-type embryos injected with a control MO (a) or the ntla MO (b). Note the shortening of posterior structures in the ntla MO-injected embryo. Dorsal views of pitx2 expression in ntla MO-injected wild-type (c) and LZoep (d,e) embryos at the 18 somite stage. Anterior is up; left (L) and right (R) are indicated. (f) Bar graph representing pitx2 expression in ntla MO-injected embryos. Genotype of injected embryos is listed first, followed by the injected MO. Expression categories and colour codes are described in figure 1. (g) Statistical significance of experiments conducted as described in figure 3.

(d) Midline ablations in LZoep embryos can restore pitx2 expression

Loss of ntla affects the notochord and the structure of KV [100,103]. To test if the phenotype of ntla MO-injected LZoep embryos is owing to loss of the notochord specifically, we removed portions of the midline by shield ablation. Portions of the shield, and cells on either side of the shield, were removed by suction with a needle (figure 5a; [96]). The damaged shield region only partially regenerates, resulting in embryos that often have gaps in the notochord and floorplate as can be detected by in situ hybridization of the midline marker sonic hedgehog (shh). To select against embryos with non-axial mesoderm tissue losses, only embryos that appeared to develop normally, but had gaps in midline shh expression, were included in the analysis (see §2).

Open in a separate window

Figure 5.

Shield ablations in wild-type and LZoep embryos. (a) Illustration of shield ablation experiment. A needle is used to remove the shield, and cells on either side of the shield. Dorsal views of pitx2 and shh expression in shield-ablated wild-type (b) and LZoep (c) embryos at the 18 somite stage. Anterior is up; left (L) and right (R) are indicated. Damage to the midline in (b,c) is observed as gaps in the normally continuous shh expression in the midline. Note that these images are focused on the shh expression in the embryos, thus the pitx2 expression in the Rohan-Beard neurons is not as visible. (d) Bar graph representing pitx2 expression in shield-ablated embryos. Expression categories and colour codes are described in figure 1. (e) Statistical significance of experiments conducted as described in figure 3.

When shields were ablated in wild-type embryos, the majority of embryos expressed pitx2 bilaterally in the LPM, indicating that damage to the midline affects pitx2 expression as expected (figure 5b,d). Shield ablation in LZoep embryos increased the number of embryos that weakly expressed pitx2 and restored pitx2 expression in the LPM to strong levels in some embryos, results that were statistically significantly different from non-ablated LZoep embryos (figure 5c–e). Compared with ntla MO injection, the extent of midline removal is considerably less in shield-ablated LZoep embryos and, accordingly, the proportion of LZoep embryos expressing strong levels of pitx2 was lower. Taken together, the effects of ntla MO and shield ablation in LZoep embryos demonstrate that disruption of the notochord can restore LPM pitx2 expression to strong levels in LZoep embryos.

(e) Blocking lefty1 activity in LZoep embryos can restore pitx2 expression

lefty1, the feedback inhibitor of Nodal, is expressed in the notochord of zebrafish embryos [82,84], and knockdown of lefty1 results in bilateral expression of Nodal targets in the LPM [80,104]. We note that zebrafish lefty1 (lft1) is expressed during early somitogenesis in the posterior midline, just anterior to KV, prior to the initiation of asymmetric gene expression in the LPM (this study and see [84]). At the 3 somite stage, 11/11 wild-type embryos expressed lft1 in the posterior notochord region. Asymmetric pitx2 expression begins at the 10 somite stage in the posterior LPM, just lateral to the lft1 expression domain in the midline (figure 2e,f). We also detect lft1 expression in the midline of LZoep embryos at early and late somitogenesis stages, but sometimes at lower levels (figure 2h,j).

To test if lft1 plays a role in the midline-dependent repression we observe, we injected MO against lft1 into wild-type and LZoep embryos. Injection of low levels of the lft1 MO into wild-type embryos does not affect morphogenesis through somitogenesis (figure 6a,b; also see [80,104]), but results in predominantly bilateral expression of pitx2 (figure 6c,f). This pattern of expression is statistically no different from that observed upon midline removal using the ntla MO (figure 6f,g). Indeed, injection of lft1 MO into LZoep embryos can restore pitx2 expression to strong levels in the LPM, strikingly similar to what is observed when the midline is ablated in LZoep embryos (figure 6d–g). These results indicate that lft1 is required for the repression of LPM pitx2 expression in LZoep mutants.

Open in a separate window

Figure 6.

lft1 MO injections in wild-type and LZoep embryos. Lateral views of 18–20 somite stage LZoep embryos injected with a control MO (a) or the lft1 MO (b). lft1-injected embryos developed without apparent defects, but were often slightly delayed compared with uninjected controls. Dorsal views of pitx2 expression in lft1 MO-injected wild-type (c) or LZoep (d,e) embryos at the 18 somite stage. Anterior is up; left (L) and right (R) are indicated. (f) Bar graph representing pitx2 expression in MO-injected embryos. Genotype of injected embryos is listed first, followed by the injected MO. The ntla MO results from figure 4 are shown here to aid in comparison with the lft1 MO results. Expression categories and colour codes are described in figure 1. (g) Statistical significance of experiments conducted as described in figure 3.

(a) A model for Lefty and Oep function prior to asymmetric gene expression during the establishment of L–R asymmetry: the midline as an ‘activator’ and ‘repressor’

Based on our results, and previously proposed models for the establishment of L–R asymmetry, we propose the following model for L–R patterning in zebrafish (figure 7a–c). Both the left and right LPM have the potential to express left side-specific genes such as pitx2, which typically occurs at the 10 somite stage. lft1 expression (dark blue; figure 7a) is induced in the posterior notochord prior to 10 somites, most likely by an earlier Nodal signalling event. Indeed, at the 3 somite stage, robust lft1 can be observed in this domain. Lefty1 from this midline domain increases the amount of Nodal ligand required to induce LPM-nodal (i.e. raises the threshold required for robust Nodal activation) and thus represses pitx2 transcription (black inhibition arrows; figure 7a). This occurs symmetrically and prevents LPM Nodal activation prematurely, before the symmetry-breaking event at KV has occurred. An oep-dependent process in the midline weakens repression by Lefty1 (grey inhibition arrows; figure 7b). As Lefty can bind to Oep, and overexpression of Oep can suppress the effects of overexpressing Lefty [8], we suggest that physical interaction between Lefty1 and Oep at the midline sequesters some Lefty1, thereby reducing the effective amount of Lefty1 able to repress LPM Nodal signals. This antagonism of Lefty1 by Oep would thus allow Nodal signals to activate transcription in the LPM (light purple; figure 7b).

Open in a separate window

Figure 7.

Model depicting antagonism of Lefty by Oep prior to the emergence of asymmetric gene expression in the zebrafish LPM. All renderings are of the notochord and surrounding LPM. The notochord terminates into KV (oval). (a) lft1 is expressed in the posterior notochord before asymmetric gene expression is detected (dark blue). We postulate that Lefty acts to repress Nodal signalling to the LPM (black repression arrows) and thus prevents activation of Nodal target genes in the LPM prior to the L–R biasing event that occurs at KV. (b) An oep-dependent signalling event overcomes/weakens the midline repressor, Lefty (weaker repression depicted by grey repression arrows). This weakened repressor can still prevent Nodal activity from occurring in the LPM initially. A biasing mechanism, flow generated by monocilia in KV, leads to downregulation of the Nodal inhibitor charon around KV (not depicted) and allows Nodal signalling to initiate in the left LPM at around the 10 somite stage (light purple). (c) Nodal autoregulation then leads to the robust expression of left side-specific genes including pitx2. Nodal signalling to the right LPM continues to be repressed by the higher level of Charon on the right of KV in cooperation with Lefty from the midline. (d) In LZoep embryos, the early posterior notochord lft1 repressor is active and not weakened owing to the later absence of Oep. (e) Nodal signalling does not reach the threshold to overcome the repressors and robust transcription in the LPM does not take place. (f) In lft1 MO-injected LZoep embryos, the early posterior notochord repressor is removed, thus the threshold for Nodal to signal to the left and the right LPM is lower, and both sides can initiate transcription of left side-specific genes. (g) In ntla MO-injected LZoep embryos, midline structures are absent and thus, there is no midline repressor. Nodal signalling can thus initiate transcription of left side-specific genes bilaterally even in LZoep embryos as Oep is not required to overcome the repressor.

While the early interaction between oep and lft1 that our experiments have uncovered is important for allowing Nodal signals to initiate in the LPM, it may be less critical for setting up the asymmetry of these signals. After all, the early domain of lft1 is L–R symmetrical and we would expect this to dampen Nodal signalling in both the left and right LPM equally. Instead, asymmetries in LPM Nodal activity are generated downstream of asymmetric flow in KV and the resulting left-sided downregulation of the Nodal repressor charon. Flow-dependent downregulation of charon and oep-dependent dampening of Lefty in the midline would thus cooperate to generate a signalling threshold in which Nodal signals are able to robustly activate in the left LPM around the 10 somite stage but not in the right LPM. Subsequently, Nodal signals spread anteriorly in the left LPM and induce further lft1 expression in the notochord (figure 7c), which then acts to maintain asymmetry by preventing Nodal signals from activating and spreading in the right LPM.

This model can explain the phenotypes we observe in LZoep embryos as well as in the various treatments (ablations, transplants and MO injections) performed in this study. In LZoep embryos (figure 7d–g), lft1 is still present in the posterior notochord (figure 7d). This expression of lft1 is likely owing to residual Oep protein from the 1-cell oep mRNA injection at the time when lft1 is induced, several hours before pitx2 is initiated in the LPM. However, in these same embryos, later oep function is greatly reduced. This late reduction in Oep in LZoep mutants, therefore, results in more active Lefty repressor in the posterior notochord; thus, Nodal signals are repressed at a level that prevents them from initiating target gene expression in the LPM (figure 7e). Removing the repressor by loss of the midline (ntla MO (figure 7g) or shield ablation) or by reducing lft1 itself (figure 7f) is sufficient to allow for signalling to occur in the LPM of LZoep mutants because these treatments replace the requirement of Oep to repress Lefty. Importantly, these treatments also suggest that there is sufficient functioning Oep in the LPM of LZoep embryos to allow for robust Nodal activity; thus, the lost or weakened pitx2 expression in LZoep is not exclusively a consequence of loss of Oep in the LPM. This agrees with our finding that adding Oep-positive cells to the LPM of LZoep mutants cannot rescue high levels of LPM pitx2 expression. Furthermore, although the potential for rescue is clear, we suggest that the rescue of LPM pitx2 in LZoep mutants, either after wild-type shield transplants or injection of ntla MO, is only partial because the LPM still has limited functional Oep remaining from the early mRNA oep injection.

This model is consistent with the timing and location of lft1, pitx2 and oep expression. Zebrafish lft1 is expressed in the posterior notochord preceding pitx2 and spaw expression in the LPM. pitx2 expression is observed later (beginning around the 10 somite stage) in the posterior-most LPM, just lateral to the posterior notochord and KV (figure 2). lft1 expression in the midline later extends from the posterior notochord to the anterior, induced by Nodal from the LPM [87,88]. Our findings provide a previously unappreciated role for early lft1 in establishing LPM Nodal signals.

Moreover, the model explains the results of our midline ablation experiments in wild-type and LZoep embryos. Midline ablations result in bilateral expression of pitx2 in wild-type embryos and can restore pitx2 to strong levels in LZoep mutants. In our model, the source of the repressor (the early domain of lft1), the notochord, is removed in these experiments, allowing Nodal signals to activate pitx2 in the LPM strongly and bilaterally. It is worth noting that the biasing mechanism occurring at KV should be unaffected by these treatments as our transplants did not contribute to the dorsal forerunner or KV populations. Nevertheless, we likely see bilateral pitx2 expression because loss of the early lft1 repressor allows Nodal activity at both the left and right LPM to induce Nodal targets, regardless of subtle differences in Nodal levels caused by the biasing mechanism at KV. However, it is worth noting that KV function may not strictly be required for Nodal signalling in the LPM. In ntla mutants or embryos injected with ntla MO, ciliated cells are present, but KV does not form. Moreover, spaw expression around KV is lost in ntla mutants, suggesting that the activating signal for Nodal in the LPM in this mutant must come from another source that remains elusive at this point. It is possible that this LPM-spaw activation in the absence of KV-spaw occurs because lft1 and charon, which both encode Spaw repressors, are also not expressed in ntla mutants, and therefore, the threshold for LPM-spaw activation is likely markedly reduced. Nevertheless, as KV-spaw is absent in ntla mutants, this observation calls into question the current major model that LRC Nodal signals induce the LPM Nodal cascade. In the mouse, node-derived Nodal is thought travel to the LPM and initiate Nodal expression there; indeed, Nodal mutants fail to activate LPM Nodal expression [24]. However, it is yet to be determined whether node-derived Nodal protein alone is sufficient to induce LPM Nodal in the endogenous context or whether node-Nodal is able to active the LPM Nodal cascade in the absence of inhibitors like Cerl-2. It is worth noting that, unlike what we report in zebrafish, it is thought that LPM Nodal signals are responsible for inducing floor-plate expression of Lefty1 [85], and therefore, removing Lefty1 from Nodal mutants would likely have no impact on whether the Nodal cascade is activated in the LPM.

This work was conducted when R.D.B. was a postdoctoral fellow in the laboratory of Alexander F. Schier, and I thank Alex for his mentorship and support. I thank members of the Schier and Yelon laboratories for discussions; Derek Stemple for teaching me the shield transplantation procedure; and S. Zimmerman, T. Bruno and N. Dillon for fish care.