gooseberry distal: Biological Overview | Evolutionary Homologs | Regulation | Developmental Biology | Effects of Mutation | References

Gene name - gooseberry

Synonyms - gsb-d

Cytological map position - 60F1

Function - transcription factor

Keyword(s) - segment polarity

Symbol - gsb

FlyBase ID:FBgn0001148

Genetic map position - 2-107.6

Classification - homeodomain and paired domain

Cellular location - nuclear

NCBI links: Precomputed BLAST | Entrez Gene

gooseberry-distal is a segment polarity gene induced by pair rule genes. gooseberry is primarily involved in a hedgehog-independent wingless autoregulatory loop. In other words, gooseberry acts to establish the maintanence of the crucial wingless signal, on which the segmental structure of the fly depends. In return, wingless acts to regulate gsb. gsb also regulates the synthesis of its sister gene, gooseberry-proximal. They function as a team in neurogenesis to specify and differentiate a subset of neurons (Gutjahr, 1993 and Li, 1993a).

Despite considerable divergence in their coding sequences, the two gooseberry genes and their proteins have conserved the same function. The finding that the essential difference between genes may reside in their cis-regulatory regions exemplifies an important evolutionary mechanism of how function diversifies after gene duplication (Li, 1994b).

The pathway regulating gsb in determination of neural fate is worth examining in detail because of the complexity of the interactions. During neurogenesis, the transmembrane protein Patched promotes a wingless-mediated specification of a neuronal precursor cell, NB4-2. Wg, secreted by row 5 cells promotes wingless expression in adjacent row 4 cells; Wg in turn represses gooseberry. Novel interactions of these genes with engrailed and invected during neurogenesis have been uncovered. While in row 4 cells Ptc represses gsb and wg, in row 5 cells en/inv relieve Ptc repression of gsb by a non-autonomous mechanism that does not involve hedgehog. The non-autonomous mechanism originates in Row 6/7 cells where en/inv engender hedgehog and another unknown secreted signal which acts in turn on adjacent row 5 cells to heighten wingless, and consequently, the expression of gooseberry. This differential regulation of gsb leads to the specification of NB5-3 and NB4-2 identities to two distinct neuroblasts. The row 5, NB5-3, neuroblasts are specified by high levels of gsb, expressed autonomously in row 5. The fate of row 4, NB4.2, requires an absence of gooseberry, assured by Patched repression and Wingless signaling from adjacent row 5 cells. The uncoupling of the ptc-gsb regulatory circuit by hedgehog and the unknown secreted signal from row 6/7 cells enables gsb to promote Wg expression in row 5 cells (Bhat, 1997).

Successive specification of Drosophila neuroblasts NB 6-4 and NB 7-3 depends on interaction of the segment polarity genes wingless, gooseberry and naked cuticle

Mechanisms have been studied leading to the fate specification of a set of late delaminating neuroblasts, NB 6-4 and NB 7-3, both of which arise from the engrailed (en) expression domain, with NB 6-4 delaminating first. No evidence is found for a direct role of hedgehog in the process of NB 7-3 specification. NB 7-3 normally requires Hh only for maintenance of Wg expression, which in turn leads to En maintenance. Evidence is presented to show that the interplay of the segmentation genes naked cuticle (nkd) and gooseberry (gsb), both of which are targets of wingless (wg) activity, leads to differential commitment to NB 6-4 and NB 7-3 cell fate. In the absence of either nkd or gsb, one NB fate is replaced by the other. However, the temporal sequence of delamination is maintained, suggesting that formation and specification of these two NBs are under independent control (Deshpande, 2001).

In the En domain Wg plays a role both in NB formation and NB specification. The homeodomain transcription factor En is a prerequisite for the formation of the NBs 6-4 and 7-3, because in its absence both NBs fail to form. Since Wg signaling is necessary for maintaining En expression, it is also essential for the formation of these two NBs. Hh is co-expressed in the En domain and En maintains Hh expression in rows 6 and 7, and Hh in turn is essential for Wg expression in row 5, thereby constituting a maintenance loop. Thus, for late NBs in row 6 and 7, the expression of En is crucial and Hh is required to maintain En expression via Wg. However, for the separate specification of NB 6-4 and NB 7-3, differential regulation of two Wg targets, nkd and gsb, is essential (Deshpande, 2001).

Wg is a diffusible molecule expressed in row 5 and acts on neighboring rows, which include rows 6 and 7. However, row 6 differs from row 7 because it expresses gsb, which is, as stated above, a target of Wg signaling. The fact that row 7 does not express gsb, despite being under the influence of Wg raises the question of how this differential regulation is brought about. In this work it is shown that Nkd is essential for this regulation. Nkd is a negative regulator of the Wg signal transduction pathway, itself being a target of this pathway. In the absence of Nkd, Gsb is derepressed, owing to Wg hyperactivity in row 7, leading to the generation of an ectopic NB 6-4 like fate. Thus, the distinct identities of NB 6-4 and NB 7-3 are brought about by the interplay of Gsb and Nkd. For NB 6-4 specification, Gsb is an essential factor. In the absence of Gsb NB 6-4 fails to be specified and instead takes the identity of NB 7-3 fate. Conversely, for NB 7-3 specification, a Gsb-free environment, which is created by the activity of Nkd, is essential. In summary, NB 6-4 needs the expression of Gsb and En, whereas NB 7-3 needs En but the absence of Gsb (Deshpande, 2001).

However, the fact that gsb as well as nkd are targets of Wg signaling makes it difficult to explain why gsb is repressed by nkd only in the posterior region of the En stripe. The posterior En domain is further away from the Wg source than the anterior En domain and therefore should receive a lower signaling input when compared with the anterior region. As a consequence, this should lead to higher Nkd activity in the anterior En cells, leading to a stronger Gsb repression in this region -- the opposite of what was observed. A careful analysis of the expression pattern on the transcriptional level does not give any obvious clues to solve this apparent paradox. During early germ band extension (stage 8-9) nkd transcription is nearly ubiquitous with higher RNA levels in the two to four cell rows posterior to the En stripe. At late phase of germ band extension, nkd expression is most abundant anterior to the En stripe and lower just posterior to the En-stripe. No significant difference between the anterior and posterior En domain could be detected. One explanation for the differential regulation of gsb could be that, owing to earlier pair rule gene activity of paired, the level of Gsb protein at the time of NB 6-4 delamination in the anterior En region is high enough to override repression by Nkd activity. Alternatively, a direct differential regulation of the two Wg targets that is due to the different levels of Wg signaling could be responsible for the observed regulatory differences. It could be that the regulation is such that the amount of Wg signaling within the En stripe causes a relatively homogenous level of nkd expression in this region. At the same time, the transcriptional activation of gsb could be more sensitive to Wg signaling levels, resulting in a very strong activation, especially near to the Wg-expressing cells. As a result, the relatively low Nkd activity in the whole En stripe might be able to inhibit gsb expression in the region of low gsb activation only: the posterior En domain. A hint that a differential regulation of Wg targets indeed exists comes from the Wg-dependent En regulation: it seems that a lower Nkd activity is sufficient to repress gsb but not to inhibit en expression. This conclusion was drawn from the finding that overexpression of nkd within the En stripe using an EnGal4 driver line leads to a selective repression of gsb with no obvious effect on en expression itself. Clearly, additional work has to be carried out to clarify these points (Deshpande, 2001).

Besides row 6 neuroectoderm, row 3 neuroectoderm also has the potential to generate an ectopic NB 7-3. It has been shown previously that in embryos mutant for ptc, neuroectodermal cells in the area of row 3 begin to express En and additional serotonergic neurons can be found in these mutant embryos, which suggests the presence of an ectopic NB 7-3 like fate. Additionally, when En is ubiquitously expressed, only row 3 has the ability to give rise to an ectopic NB 7-3 fate. In all cases, this occurs at the cost of row 3 NBs such as NB 3-3. It is thought that this might reflect that row 3 neuroectoderm, which is right in the middle of the segment, represents something like a 'ground state' in the neuroectoderm: in this area neither Hh nor Wg signaling may take place. Therefore the decision to specify late row 3 or late row 7 NBs seems to be only dependent on the absence or presence of En, respectively (Deshpande, 2001).

Previous work has indicated that genes expressed in proneural clusters are involved in specifying the individual fates of NBs that develop from these clusters. The finding that NB 6-4 and NB 7-3 can be mutually transformed while the sequence of birth does not change suggests that the mechanism for the timing of late NB delamination is independent from mechanisms that regulate NB identity. This might be reminiscent of early NBs. Initiation of S1 NB formation requires the activity of proneural genes that have been shown to be dependent on pair-rule genes. The identity of the NBs delaminating from these clusters, however, is dictated by the activity of segment polarity genes. Thus, the control of proneural gene expression that enables NB formation and the control of segmentation genes conferring NB identity occurs in parallel. At later stages, pair-rule gene expression vanishes and can no longer be responsible for NB formation. How is NB formation regulated in the following segregation waves? One possibility is that after the first segregation wave, NB formation and identity are more tightly linked; the finding that specific NBs like NB 4-2 are sometimes not transformed but missing in wg mutant embryos seems to support this idea. However, the finding that the transformed NB 6-4 and NB 7-3 are delaminating according to the 'old identity' shows that, at least in these cases, NB formation and specification is independent. The results favour the idea that the timing of the formation of proneural clusters within the neuroectoderm is generally independent of the segment polarity genes investigated here. This does not exclude permissive functions, such as those of En, which enable the proneural cluster formation as such. According to this hypothesis, intrinsic or extrinsic factors present in the position of the proneural cluster at the time of delamination govern the identities of the NBs. This might be not only true for the positional regulation of NB identity but also for the determination of NB identity along the temporal axis. Indeed, heterochronic transplantation experiments strongly support the possibility that one or more extrinsic factors exist that lead to stage specific NB identities. It will be a challenge for the future to identify these factors, and to investigate whether similar mechanisms exist in higher organisms (Deshpande, 2001).

gooseberry and gooseberry neuro in the central nervous system and segmentation of the Drosophila embryo

The gooseberry locus of Drosophila consists of two homologous Pax genes, gooseberry neuro (gsbn) and gooseberry (gsb). Homologous recombination was obtained of null mutants of either gene as well as a deficiency inactivating only gsbn and gsb. This analysis shows that (1) gsbn null mutants are subviable while all surviving males and most females are sterile; (2) gsb and gsbn share overlapping functions in segmentation and the CNS, in which gsbn largely, but not completely depends on the transcriptional activation by the product of gsb; (3) as a consequence, in the absence of gsbn, gsb becomes haploinsufficient for its function in the CNS, and gsbn-/-gsb-/+ mutants die as larvae. Such mutants display defects in the proper specification of the SNa branch of the segmental nerve, which appears intact in gsbn-/- mutants. Lineage analysis in the embryonic CNS showed that gsbn is expressed in the entire lineage derived from NB5-4, which generates 4 or 5 motoneurons whose axons are part of the SNa branch and all of which except one also express BarH1. Analysis of gsbn-/-gsb-/+ clones originating from NB5-4 suggests that together the genes specify the SNa fate and concomitantly repress the SNc fate in this lineage and that their products activate BarH1 transcription. Specification of the SNa fate by Gsb and Gsbn occurs mainly at the NB and GMC stage. However, the SNa mutant phenotype can be rescued by providing Gsbn as late as at the postmitotic stage. A model is proposed how selection for both genes occurred after their duplication during evolution (He, 2013).

Although many transcription factors have been identified that control the cell fates of the ISN subtypes, little is known about the factors that determine the SN subtypes. The results demonstrate that both gsb and gsbn play essential roles in the specification of SNa MNs. The fact that in gsbn-/-gsb-/+ embryos ectopic SNc projections are induced at the expense of the normal SNa projection suggests that, in the absence of a functional gsbn gene and the presence of only one functional gsb gene, MNs of the NB5-4 lineage acquire the SNc identity. It follows that gsb and gsbn specify the SNa fate and concomitantly repress the SNc fate, thereby subdividing the SN into SNa and SNc. This hypothesis is testable, once regulators of the SNc pathway are known, and plausible, as an analogous situation may exist for the subdivision between ISN, ISNb, and ISNd. In lim3 mutants, ISNb MNs fail to innervate their normal target muscles but project ectopically in ISNd. In addition, in eve mutants - even-skipped (eve) and dHb9 display non-overlapping expression patterns in the VNC and MNs - dHb9 is derepressed in MNs that in wild-type embryos innervate dorsal muscles through ISN. Conversely, in dHb9 mutants, eve is ectopically expressed in several MNs that normally express dHb9 and innervate ventral muscles through ISNb but now fail to innervate the ventral muscles and ectopically project with ISN to the dorsal muscle field. Therefore, mutual repression between motoneuronal subtypes might be a general mechanism by which subtype identities are determined. The expression patterns of both dHb9, which is not expressed in Gsbn-positive SNa MNs, and eve in MNs seem not to be changed in gsbn-/-gsb-/+ mutants. Hence, it remains unclear whether mutual repression mechanisms act only as switch between motoneuronal subtypes or also between ISN and SN identities (He, 2013).

Although both gsbn and gsb are involved in the specification of the SNa MNs, it appears that gsb plays a more important role, as evident from the fact that while two doses of gsb/gsbn are required to properly determine the SNa fate, one of them has to be gsb whereas gsbn is dispensable in the presence of two functional copies of gsb. As gsb is no longer expressed in the CNS after stage 13, this further implies that the fate of the SNa MNs derived from NB5-4 is specified between stages 11 and 13. It has been demonstrated that, despite the considerably diverged C-terminal moieties of the coding sequences of gsb, gsbn, and paired (prd), their proteins can still largely substitute for each other's functions and their specific developmental roles reside in the difference between their cis-regulatory regions acquired during evolution. Accordingly, the differential roles of gsb and gsbn most probably result from the dependence on gsb of gsbn transcription in SNa MNs. Once gsbn is expressed in the NB5-4 lineage, it functions redundantly with gsb to activate downstream target genes, as no SNa phenotype is observed in the absence of Gsbn. The redundancy of gsbn with gsb becomes evident only when, in addition to the homozygous deletion of gsbn, a copy of gsb is removed, which results in the SNa phenotype and in turn can be rescued by the addition of a gsbn transgene. (He, 2013).

While gsbn and gsb are the only transcription factors shown to specify the identity of SNa, at least another transcription factor is involved in the determination of the SNa pathway as at least two SNa MNs express neither gsbn nor gsb but BarH1. Since all BarH1-positive motor axons project in the SNa nerves (innervating muscles 21–24, 8 and/or 5), while the 5 Gsbn-positive MNs derived from NB5-4 innervate muscles 22–24, 5 and perhaps 8 (but not muscle 21), muscle 21 is innervated by one of the two BarH1 SNa MNs that do not express Gsbn (He, 2013).

The expression of gsbn and gsb in the NBs of rows 5 and 6 raises the question of whether the SNa phenotype observed in gsbn-/-gsb-/+ mutants simply results from a change in NB5-4 identity. A clue to this question provides the observation that in the absence of a functional gsbn gene, no SNa phenotype is observed if two functional copies of gsb are present. Since these are expressed mainly at the NB and GMC stages, but only at low levels or not at all in MNs, it is probable that the SNa fate is specified before the MN stage. Indeed the level of gsb provided by only one gsb gene in gsbn-/-gsb-/+ mutants is sufficient to support the wild-type SNa fate in 10% of the NB5-4 clones analyzed, while in many of the remaining clones at least one SNa MN maintains its identity and properly innervates its target muscle. However, the SNa fate can also be largely rescued in gsbn-/-gsb-/+ mutants by gsbn at the postmitotic stage in MNs. Remarkably, all neurons of the NB5-4 lineage are motoneuronal. It is thought that these observations are best explained by a model in which NB5-4 MNs are specified during two critical periods. At the NB stage, when the gsb protein level is high, the fate of the progeny of NB5-4 is specified by gsb to be motoneuronal such that their axons will extend outside of the VNC and project through the SN. This explains why in gsbn-/-gsb-/+ mutants the progeny of NB5-4 still are MNs although they mostly target the wrong muscles. During the second, much longer period, extending from the GMC to the MN stage, the subtype identity, SNa versus SNc, is specified. As long as levels of gsb and/or gsbn protein are sufficiently high at the GMC or MN stage, the subtype identity is properly specified. Thus, sufficient gsb protein is provided in gsbn-/-gsb+/+ embryos during the GMC and perhaps early MN stage, whereas in gsbn-/-gsb-/+ embryos rescued by elav-Gal4 UAS-gsbn, levels of gsbn provided at the MN stage are sufficient to properly specify the subtype identity (He, 2013).

The increase in cell number of the NB5-4 lineage in gsbn-/-gsb-/+ mutants could result from a deregulation of the proliferation of NB5-4 and/or GMCs or from a decrease in apoptosis of the neurons, which raises the question of whether a defect in motor axons is somehow coupled to a defect in the control of cell number. In 25% of the clones that exhibit ectopic SNc projections, however, the cell number is similar to that in the wild type, in the range of 4-5. Moreover, in all three clones that show normal SNa and no ectopic SNc projection, the cell number is increased, ranging from 7 to 9. Therefore, the two phenotypes do not seem to be correlated, which suggests that the two functions in the NB5-4 lineage of gsb and gsbn are separable. Consistent with this conclusion, it has been reported that in embryos deficient for programmed cell death, NB5-4 clones consist of 10-13 neurons but display a wild type-like projection pattern (He, 2013).

For historical reasons, the strong cuticular phenotype of Df(2R)IIX62 was assumed as gold standard for a null allele of gsb despite the fact that it was soon discovered that this deficiency deletes two duplicated genes at the gsb locus, gsb and gsbn. That no gsb alleles with a strong segment-polarity phenotype like that of Df(2R)IIX62 were found was puzzling but attributed to ineffective mutagenesis screens. Accordingly, gsb alleles exhibiting weak cuticular phenotypes were assumed to be hypomorphic. The results presented in this study strongly suggest a different explanation for this long-standing problem: gsbn and gsb fulfill partially redundant functions in embryonic segmentation. They further explain why previously strong cuticular phenotypes similar to that of Df(2R)IIX62 could be obtained only by combining a gsb null allele, like the large deficiency Df(2R)SB1 that uncovers gsb but not gsbn, with a large deficiency uncovering both gsb and gsbn, thus generating gsbn-/+gsb-/- mutants (He, 2013).

Thus, the situation for the specification of the epidermis between mid stage 9 and mid stage 11 (Li and Noll, 1993) by the segment-polarity functions of gsb and gsbn is similar to that for the determination of the SNa MNs derived from NB5-4 by their CNS functions during stages 11-13. The SNa phenotype of gsbn-/-gsb-/+ embryos can be rescued if gsbn is provided at the MN stage. Similarly, the cuticular segment-polarity phenotype of gsb null mutants is very weak because it is rescued after stage 10 by the presence of gsbn in the epidermis. If the gsbn dose is reduced in gsb-/- embryos, the rescue effect by gsbn is diminished and the cuticular phenotype enhanced (He, 2013).

Since it has been shown that gsbn and gsb share several overlapping functions, one should be cautious when attributing phenotypes observed in gsbn-/+gsb-/- mutants solely to the loss of gsb functions. The two strong or null alleles of gsb generated in this study, gsbs252 and gsbgsbJ46, will offer the opportunity to study the functions of gsb without the discussed genetic complications (He, 2013).

A common feature of many gene pairs in Drosophila is 'enhancer sharing': two duplicated neighboring genes share the same enhancer(s) and hence are expressed in similar patterns. As a consequence, one of the two genes could be completely redundant. Indeed, this is the case for the invected gene of Drosophila at the engrailed-invected locus. By contrast, the enhancers of gsb and gsbn cannot activate the heterologous promoters properly. Moreover, the two gooseberry genes are linked in a regulatory hierarchy: transcription of gsbn partially depends on the expression of gsb. These features of the gsb locus suggest that aspects of their functional interactions differ from those of other gene pairs. The phenotype of gsbn mutants is mild compared to that of gsb mutants. Nevertheless, gsbn and gsb are still intimately linked, notably revealed by the haploinsufficiency of gsb in gsbn mutants. The detailed analysis of the SNa phenotype in gsbn-/-gsb-/+ mutants suggests that the functional link between gsb and gsbn is established mainly through the strong requirement for full activation of gsbn by a functional gsb protein. Once gsbn is expressed, Gsb, together with Gsbn, activates their target genes in a redundant fashion, probably because of their two highly conserved DNA-binding domains, the paired-domain and the paired-type homeodomain. It appears that the gsb-gsbn hierarchy at least ensures normal development of the dorsal SNa MNs if one copy of gsb is lost as long as one functional copy of gsbn is preserved. Moreover, gsbn-/-gsb-/+ larvae die during first instar, most probably due to defects in the CNS. Therefore, the establishment of a gsb-gsbn hierarchy avoids deleterious effects caused by the gsb haploinsufficiency (He, 2013).

The gsb locus provides an interesting paradigm of how selection for both genes may occur after duplication of their ancestral gene during evolution. A possible, though speculative, scenario explaining the selection for both genes, gsb and gsbn, could be the following. Let the general case first considered of a gene that becomes haploinsufficient for a specific function during evolution. This is a disadvantage against which evolution selects by favoring higher transcription levels in cells dependent on this haploinsufficient function. It is envisaged that such increased levels could result (1) from mutations in the enhancers that regulate transcription in the tissues subject to the haploinsufficiency, or (2) from duplication of the gene including its enhancer responsible for the haploinsufficient function. The second scenario, however, is possible only if a quadruple dose is neither lethal nor causes other disadvantages against which evolution would select. In the case of the ancestral gene of gsb/gsbn, it is argued that haploinsufficiency for a certain function, for example that required for proper development of SNa MNs, was overcome by a gene duplication and inversion, resulting in two genes that are divergently transcribed, gsbn and gsb. It is further argued that selection for both genes occurs if one of the genes is partly dependent on the other in cells that required the haploinsufficient function before gene duplication. This is conceivable if the enhancer of the ancestral gene responding to the segment-polarity gene Wg was included in the duplication. While the gsb gene remained dependent on Prd and Wg, gsbn became partly dependent on gsb that maintained Wg expression. Thus, as one duplicated gene, gsb, retained some of the original enhancers and the other, gsbn, became partially dependent on gsb in some tissues, the expression of the two genes overlapped in these tissues during development. As a consequence, gsb was no longer haploinsufficient because when one of its copies was inactivated by mutation, gsbn provided sufficient protein substituting for the missing gsb protein. This explains why gsbn, which is partly dependent on gsb, does not show an SNa phenotype when inactivated in both copies, as long as two copies of gsb remain present. This mechanism would have selected for both gsb genes after their duplication, and each genes was thus able to acquire additional enhancers and functions during evolution. Consistent with this hypothetical scenario, the haploinsufficiency of the ancestral gene is still apparent today in gsb when both copies of gsbn are inactivated by mutation. By this scenario, haploinsufficiency for a particular function of the ancestral gene can be overcome by gene duplication. However, this is not possible if twice the level of the ancestral gene product is deleterious for any of its essential functions (He, 2013).

Interestingly, a similar regulatory link may have been established during the much earlier duplication that gave rise to prd and the ancestor of gsb/gsbn. Here, the ancestral gene of gsb/gsbn may have retained most functions of its ancestor and became dependent on the activation by prd, which retained the enhancer that activated its ancestor initially. By the acquisition of additional enhancers, prd evolved as pair-rule gene activating the ancestor of gsb/gsbn in alternating segments, while other evolving pair-rule genes activated gsb/gsbn in the remaining segments. Why prd is eventually activated in a segment-polarity pattern is still a conundrum, as so far no functions of prd have been demonstrated in every other segment. A possible explanation is that prd performs in these segments an epidermal function that is redundant with that of gsb, similar to the redundancy demonstrated in this study for functions of gsb and gsbn (He, 2013).

In addition to a hierarchical regulatory link, there seems to be a structural link between the duplicated genes, which may also hold the clue to the question why their enhancers have become so selective for their cognate promoters (He, 2013).


The gooseberry (gsb) locus contains two closely linked genes, gsb (often referred to as goosberry distal (gsb-d) and gooseberry-neuro, transcribed in opposite directions. They are separated by about 10kb of a common upstream region.

cDNA clone length - 1564

Bases in 5' UTR - 130

Exons - two


Amino Acids - 427

Structural Domains

Both gooseberry genes are structurally related to one another and to the paired (prd) gene. The two proteins have considerably divergent coding sequences. The structural homology between these two proteins and the Paired protein consists essentially of two domains forming most of the amino-terminal halves of the proteins: the PRD domain of 128 amino acids and a PRD-type homeodomain of 60 amino acids, plus an additional 18 amino acids at its amino-terminal end (Baumgartner, 1987). Unlike Paired, there is no C-terminal PRD domain.

gooseberry distal: Evolutionary Homologs | Regulation | Developmental Biology | Effects of Mutation | References

date revised: 3 July 97 

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.