ABSTRACT
LIM-homeodomain transcription factors are expressed in subsets of neurons and are required for correct axon guidance and neurotransmitter identity. The LIM-homeodomain family member Apterous requires the LIM-binding protein Chip to execute patterned outgrowth of the Drosophila wing. To determine whether Chip is a general cofactor for diverse LIM-homeodomain functions in vivo, we studied its role in the embryonic nervous system. Loss-of-function Chip mutations cause defects in neurotransmitter production that mimic apterous and islet mutants. Chip is also required cell-autonomously by Apterous-expressing neurons for proper axon guidance, and requires both a homodimerization domain and a LIM interaction domain to function appropriately. Using a Chip/Apterous chimeric molecule lacking domains normally required for their interaction, we reconstituted the complex and rescued the axon guidance defects of apterous mutants, of Chip mutants and of embryos doubly mutant for both apterous and Chip. Our results indicate that Chip participates in a range of developmental programs controlled by LIM-homeodomain proteins and that a tetrameric complex comprising two Apterous molecules bridged by a Chip homodimer is the functional unit through which Apterous acts during neuronal differentiation.
INTRODUCTION
Even simple nervous systems contain remarkably diverse neuronal cell types, each characterized by distinct properties such as axon trajectory and the type of neurotransmitter synthesized. The acquisition of these properties is determined in part by activation of cell-specific repertoires of gene expression. Studies of diverse species have shown that members of the LIM-homeodomain (LIM-HD) family of transcription factors are expressed in subsets of neurons and are required for key aspects of neuronal differentiation (Way and Chalfie, 1988; Lundgren et al., 1995; Pfaff et al., 1996; Hobert et al., 1997; Thor and Thomas, 1997; Benveniste et al., 1998; Sharma et al., 1998; Hobert et al., 1999). In the spinal cord of vertebrates and in the ventral nerve cord (VNC) of Drosophila, the expression of specific combinations of LIM-HD family members constitutes a functional “LIM-HD combinatorial code” that dictates axon pathway selection of motor neurons (Sharma et al., 1998; Thor et al., 1999). The Drosophila LIM-HD genes islet (isl), lim3 and apterous (ap) direct axon pathfinding and neurotransmitter identity in subsets of postmitotic neurons of the embryonic VNC. isl is required for transmitter production by dopaminergic neurons (Thor and Thomas, 1997), and both isl and lim3 control motor neuron pathway selection in the intersegmental nerve (ISN) (Thor et al., 1999). In addition to its role in wing morphogenesis (Cohen et al., 1992; Diaz-Benjumea and Cohen, 1993; Blair et al., 1994), ap is required within the VNC for correct axon guidance of a subset of interneurons and for expression of the neurotransmitter dFMRFamide (dFMRFa) in neuroendocrine cells (Lundgren et al., 1995; Benveniste et al., 1998).
LIM-HD proteins contain a DNA-binding homeodomain plus two cysteine-rich LIM domains (Freyd et al., 1990; Karlsson et al., 1990) that mediate interactions with other proteins (Schmeichel and Beckerle, 1994; Arber and Caroni, 1996). The LIM domains of all LIM-HD proteins, as well as those of nuclear LIM-only (LMO) proteins, are bound by Nuclear LIM Interactor (NLI), first identified in mouse and also known as Ldb1 or CLIM-2 (Agulnick et al., 1996; Jurata et al., 1996; Bach et al., 1997). NLI homodimerizes in vitro and thereby coordinates LIM-HD and LMO transcription factors into homomeric and heteromeric complexes by forming a bridge between two LIM-HD proteins (Jurata et al., 1998). NLI is expressed widely throughout embryonic development of the mouse, with elevated levels in the CNS where broad domains of NLI expression overlap with those of the LIM-HD proteins Islet-1, Lhx-3 and Lhx-2 among others (Jurata et al., 1996; Bach et al., 1997). Analysis of mice mutant for NLI has revealed that genetic dissection of its function in the CNS will be complicated by an early requirement for NLI at gastrulation (A. Agulnick, personal communication).
Experiments with the Drosophila ortholog of NLI, called Chip (dLdb), provided the initial evidence that interactions between these cofactors and LIM domains are required in normal development. Chip interacts with the LIM domains of Apterous (Ap), and genetic experiments indicated that Chip supports Ap activity during wing development (Morcillo et al., 1997; Fernandez-Funez et al., 1998). In vitro biochemical assays and further genetic experiments in vivo suggest that the active form of Chip and Ap in the wing is a tetramer comprising two Ap molecules bridged by a Chip homodimer (Milan and Cohen, 1999; van Meyel et al., 1999). This complex is required for Ap activity in dorsoventral patterning and outgrowth of the wing and is subject to disruption by dLMO, a nuclear LIM-only protein that can compete with LIM-HDs for binding to Chip (Milan et al., 1998; Shoresh et al., 1998; Milan and Cohen, 1999). In the wing, dLMO expression is upregulated by Ap, providing a mechanism for negative feedback upon Chip/Ap tetrameric complexes and modulation of Ap activity (Milan et al., 1998).
Like NLI, Chip is expressed widely during embryogenesis and interacts with all LIM-HD proteins in vitro, suggesting it may be a requisite component of many transcription regulatory complexes involving LIM-HD factors (van Meyel et al., 1999). To determine whether Chip is a general cofactor for diverse LIM-HD functions in vivo, we studied the role of Chip in Apterous and Islet functions in the embryonic nervous system. Here we show that loss-of-function Chip mutants have defects similar to ap and isl mutants including loss of neurotransmitter production in Ap and Isl cells and aberrant axon pathfinding of the Ap neurons. Chip acts cell autonomously during axon guidance of the Ap neurons, and requires intact domains for self-dimerization and for interaction with LIM domains, suggesting that, as in wing development, Chip and Ap are required to interact physically and form a tetrameric complex in Ap neurons. To test this model in the nervous system, we reconstituted the complex using a Chip/Ap chimera lacking domains normally required for their interaction. The Chip/Ap chimera rescues the axon guidance defects of ap mutants, of Chip mutants and of embryos doubly mutant for both ap and Chip. Our results provide the first functional evidence in vivo that Chip participates in a range of developmental programs controlled by LIM-HD proteins, and that a Chip/LIM-HD complex is the primary functional unit through which LIM-HD proteins control transcriptional regulation of neuronal differentiation.
MATERIALS AND METHODS
Fly strains and genetics
A description of all UAS constructs and the generation of transgenic Drosophila lines has been published previously (van Meyel et al., 1999). Multiple lines were generated for each UAS transgene and those that exhibited strongest GAL4-directed c-myc expression were selected for analysis and recombined with a UAS-tau-lacZ reporter for axon tracing (Callahan and Thomas, 1994). For mutant analyses, we recombined Chipe5.5 onto each of the apGAL4 and apP44 chromosomes, then introduced UAS transgenes into these compound heterozygotes and intercrossed them to achieve the appropriate mutant background. Activation of dFMRFamide expression was assayed using a Tv-lacZ reporter on the X chromosome (Benveniste et al., 1998). All fly crosses and embryo collections were carried out at 25°C.
Immunohistochemistry
Embryo dissection and HRP immunostaining were performed as described previously (Callahan and Thomas, 1994; Thor and Thomas, 1997; O’Keefe et al., 1998). For fluorescence immunostaining to detect Chip and Ap simultaneously, dissected embryos were incubated at 4°C overnight with rat anti-Ap antibody diluted 1:1000 (Lundgren et al., 1995). They were then rinsed, and incubated simultaneously with affinity-purified rabbit anti-Chip antibody (diluted 1:75) and a biotin-conjugated anti-rat secondary antibody (Vector; diluted 1:500) at room temperature for 2 hours. Subsequently, FITC-conjugated streptavidin (Jackson Immunoresearch Labs; diluted 1:400) and Cy-3-conjugated anti-rabbit antibody (Jackson Immunoresearch, diluted 1:500) were incubated for 1 hour at room temperature, rinsed and mounted for microscopy. To simultaneously localize Ap- and dLMO-expressing cells, we double labelled apGAL4/+; UAS-tau-lacZ/+ embryos with mouse anti-dLMO sera (diluted 1:100) and rabbit anti-β-galactosidase (diluted 1:1000) as above.
Quantification of axon pathfinding errors
To compare the occurrence of Ap interneuron pathfinding errors among embryos of different genotypes, one observer (D.D.O’K.) was blinded to the genotype of each embryo and scored abdominal segments A1-A7 for the presence or absence of pathfinding errors. The mean percentage of wild-type segments per embryo was calculated (average n=10 embryos per genotype), and for each genotype the mean value was compared with the others in a multiple comparison analysis of variance (ANOVA) using the Student-Neuman-Keuls test (P<0.05).
RESULTS
Chip expression in the CNS
Chip is expressed in most, if not all, embryonic and larval tissues (Morcillo et al., 1997). In wild-type embryos, we found strong, nuclear Chip expression throughout the developing VNC with no apparent subclasses of neurons excluded (Fig. 1A). A substantial fraction of embryonic Chip is contributed maternally during oogenesis, and this maternally derived expression is required for early embryonic segmentation (Morcillo et al., 1997). To estimate the relative contribution of zygotic and maternally derived Chip to the embryonic VNC, we examined homozygous embryos mutant for Chipe5.5, a null allele encoding a truncated protein, and Chipγ230, a deficiency that completely removes Chip. Derived from an intercross of heterozygous parents, these mutants are expected to retain half of the maternal and no zygotic Chip expression. We observed little reduction of staining in mutant embryos relative to Chip/+ heterozygotes (data not shown). Thus it appears a substantial fraction of Chip in the VNC is provided maternally.
Co-labelling embryos with anti-Ap and anti-Chip antibodies revealed that Chip expression overlaps with all the Ap neurons of the developing VNC (Fig. 1A). Based upon previous studies of biochemical interactions and wing development, we propose that, in the nervous system, Ap could function in a tetrameric complex with Chip. However, it is also possible that Ap could act in a Chip-independent fashion if Chip were sequestered in Ap cells by co-expression of dLMO. Using anti-dLMO antibodies (Milan et al., 1998), we found that dLMO is not expressed in the Ap neurons of the developing VNC (Fig. 1B), suggesting that Chip is free to participate in complex formation and Ap function (Fig. 1C).
Chip is required for neurotransmitter identities regulated by Ap and Isl
If Chip were required for Ap function, elimination of Chip might be expected to result in an ap-like phenotype. The requirement of maternally supplied Chip in segmentation (Morcillo et al., 1997) precluded an examination of the effects of eliminating both maternal and zygotic Chip on neuronal development. Thus, we assayed neurotransmitter expression and axon guidance in Chipe5.5 mutants in which half of the maternal and all of the zygotic Chip expression were absent.
In each thoracic hemisegment of the VNC, ap is expressed in a lateral cluster of four neurons, one of which is the Tv neuroendocrine cell that expresses the neurotransmitter dFMRFa (Schneider et al., 1993). In wild-type embryos, there are a total of six Tv cells, one in each thoracic hemisegment. In ap mutants, the Tv neurons are present, but half of all Tv neurons stochastically fail to express dFMRFa (Benveniste et al., 1998). This regulation of dFMRFa by ap is transcriptional, since expression of a fusion transgene comprising a 446 bp Tv neuron-specific enhancer of the dFMRFa gene driving β-galactosidase (Tv-lacZ, Fig. 2A) is similarly reduced in ap mutants. Ap binds in vitro to each of three sequences within the enhancer, and mutagenesis of these sites confirmed that these sequences are important for Tv-lacZ expression in vivo (Benveniste et al., 1998).
To determine whether reduction of Chip results in an ap-like reduction in transcriptional activation of dFMRFa, expression of the Tv-lacZ reporter transgene was assayed in wild-type, apP44 and Chip mutant embryos (Fig. 2). Both apP44 and Chipe5.5 mutant embryos show decreased Tv-lacZ activity in Tv neurons relative to wild-type controls (Fig. 2B-D,F), implicating Chip in the establishment of this Ap-regulated aspect of neuronal differentiation. The reduction of Tv-lacZ activity was less severe in Chip null mutants than ap null mutants, probably because of the maternally supplied Chip remaining in Chip mutants. In embryos homozygous for the antimorphic Chipγ96.1 mutation (Torigoi et al., 2000), Tv-lacZ expression was reduced further than Chip null mutants but not to the level of ap mutants (Fig. 2E,F).
Like Ap, the LIM HD protein Isl also regulates neurotransmitter identity of embryonic neurons. There are three dopaminergic cells per segment of the VNC, one unpaired midline cell and a pair of dorsal lateral cells (Fig. 3A,B), all of which express Isl protein and thus represent a subset of the isl interneurons. isl mutants show loss of expression of tyrosine hydroxylase (TH), a rate-limiting enzyme in the synthesis of dopamine (Thor and Thomas, 1997). Double-labelling immunochemistry showed that dLMO is not expressed in the isl neurons of the embryonic VNC (data not shown), suggesting that, in these neurons, Chip is free to associate with Isl. To test the role of Chip in the expression of TH, we stained late-stage wild-type and Chip mutant embryos with anti-TH antibodies. Homozygous Chipe5.5 mutant embryos retain TH expression in the ventral unpaired midline cells, but few of the dorsal lateral cells express TH, and in those that do, TH levels are significantly reduced relative to wild-type (Fig. 3C,D). In embryos homozygous for the Chipγ96.1 antimorph, TH expression was greatly diminished in both the ventral midline and dorsal lateral dopaminergic neurons (Fig. 3E,F). While it is clear that the paired dorsal TH cells are more sensitive to the reduction in Chip dosage than the unpaired ventral cells, the effects of the antimorphic Chipγ96.1 allele suggest that TH production in the latter cells is also dependent on Chip. From these results, together with the above results on the expression of FMRFamide, we conclude that Chip is required for both Ap- and Islet-regulated neurotransmitter production in the CNS.
Chip is required for correct axon pathway selection of Ap interneurons
For a detailed analysis of Chip’s role in neuronal differentiation, we studied axon guidance of the Ap-expressing interneurons. To visualize the axon trajectories of the Ap neurons, we used apGAL4, a P[GAL4] insertion in the ap locus, to drive expression of the axon tracing reporter UAS-tau-lacZ in the ap cells of the CNS (Calleja et al., 1996; O’Keefe et al., 1998). Within each abdominal hemisegment of wild-type embryos, Ap is expressed by three interneurons that extend axons ipsilaterally and anteriorly within each longitudinal connective (Lundgren et al., 1995). Upon reaching the adjacent homologs, forming a discrete axon bundle running the length of the VNC. apGAL4 is a mutant allele of ap, but apGAL4/+ heterozygotes display few pathfinding errors, with 93% of segments showing no defects (Fig. 4B,H). In ap mutants (apGAL4/apP44), all Ap interneurons survive and extend axons, but these axons display pathfinding errors in nearly every segment, including defasciculation of axon bundles, aberrant pathway selection within the longitudinal connectives, occasional misrouting across the midline, and stalling (Fig. 4C,H).
To determine whether reduction of Chip causes ap-like defects of axon guidance, we first examined embryos transheterozygous for apGAL4 and Chipe5.5. We detected no increase in axon guidance defects over Chipe5.5/+ or apGAL4/+ individuals. We next examined homozygous Chipe5.5 mutants, in which the levels of Chip should be further reduced, and found they did exhibit the same array of pathfinding errors as ap mutants (Fig. 4D). We quantified the incidence of pathfinding errors relative to apGAL4/+ controls and found fewer errors than occur in ap mutants (Fig. 4H). Thus, axon guidance defects, like the reduction of Tv-lacZ activity, are more moderate in Chipe5.5 mutants than in ap mutants, likely due to maternally supplied Chip. Pathfinding errors in Chipe5.5 mutants occurred in the absence of any discernible disorganization of neurons or neuroblast lineages, as determined by immunochemical staining with antibodies to Engrailed, Even-skipped and Prospero (data not shown). Furthermore, as assayed with an antibody to Fasciclin II (Fas II) (Van Vactor et al., 1993), we detected no defects in the pathfinding of Fas II-expressing cells indicating that errors of axon guidance of the Ap neurons occurred without concomitant errors among large subsets of other neurons.
Introduction of full-length Chip (ChipFL) specifically into Ap interneurons of Chipe5.5 mutant embryos using a UAS-ChipFL transgene rescued pathfinding errors completely to the level of apGAL4/+ controls (Fig. 4E), showing that Chip activity is required cell-autonomously within Ap neurons for correct axon pathfinding. In contrast, mutant Chip proteins lacking either the LIM Interaction Domain (ChipΔLID, Fig. 4F) or the Dimerization Domain (ChipΔDD, Fig. 4G) failed to rescue the pathfinding errors, despite being expressed at high levels and properly localized to nuclei. This indicates that binding to LIM domains and self-dimerization are both required for Chip function. ChipΔDD, which acts in a dominant negative fashion in wing development (van Meyel et al., 1999), increased the incidence of pathfinding errors (Fig. 4H), presumably because it prevents Ap from associating with maternally supplied endogenous Chip.
To further test the role of Chip in LIM-HD-mediated axon guidance, we examined the Isl-expressing ISN motor neurons in Chipe5.5 and Chipγ96.1 homozygous mutant embryos using the anti-Fas II antibody. In isl mutants, these motor neurons fail to innervate their appropriate muscle targets (Thor and Thomas, 1997). In contrast to the Ap neurons, we detected no clear alterations in projections of the Isl motor neurons in Chip mutants (data not shown). This suggests that the levels of Chip may not be reduced sufficiently to affect these neurons, similar to the reduced sensitivity of the unpaired ventral TH-expressing Isl cells. Alternatively, Isl might function during pathway selection of these neurons in a Chip-independent fashion, a possibility consistent with our previous finding that Isl, but not Ap, is uniquely able to form heterodimers directly with Lim3 and that these Isl/Lim3 heterodimers may be favored over Isl/Chip complexes (van Meyel et al., 1999).
Chip and Ap are required to physically interact for function in axon guidance
In Drosophila wing development, the active form of Ap appears to be a tetramer, in which the LIM domains of Ap and the LID of Chip permit the homeodomains of two Ap molecules to be spanned by a Chip homodimer (Milan and Cohen, 1999; van Meyel et al., 1999). To determine whether Ap function in the nervous system also requires formation of a complex with Chip, we removed the domains responsible for this interaction and tethered the remainder of each protein together, yielding the chimera ChipΔLID:ApΔLIM. This chimera is capable of rescuing wing defects of ap mutants (van Meyel et al., 1999). Using apGAL4 to drive expression of a UAS-ChipΔLID:ApΔLIM transgene, we tested whether the chimera could also rescue axon guidance defects in ap mutants. Axon fasciculation and correct pathway selection were restored in a majority of segments examined from ap mutant embryos carrying the chimera transgene, indicating that the axon guidance defects of ap mutants were rescued by the chimera (Fig. 5B,F). In contrast, simultaneous introduction of unfused UAS-ApΔLIM and UAS-ChipΔLID transgenes failed to rescue the defects in ap mutants (data not shown).
The observation that reduced Chip expression in the VNC causes defects in axon guidance of the Ap interneurons provided us with an opportunity to assess the ability of the ChipΔLID:ApΔLIM chimera to rescue a Chip mutant phenotype. We found the chimera rescued axon defects in homozygous Chipe5.5 embryos to a level equal to that achieved by full-length Chip and to that of apGAL4/+ controls (Fig. 5C,F). This result confirms that the sole function of the 32 amino acid LID domain of Chip is to bind Ap, since the effects of its removal are negligible when the remainder of Chip is tethered directly to Ap.
Although the sole function of the Chip LID appears to be to bind the LIM domains of Ap, the fact that the chimera only partially rescues the axon guidance defects of ap mutants suggested a major but not exclusive role for the Ap LIM domains in binding Chip. Therefore we hypothesized that the chimera should also only partially rescue axon pathfinding of embryos doubly mutant for both ap and Chip. Double mutants exhibit the same types of pathfinding errors as single mutants, and these occur with a frequency equal to that of ap mutants (Fig. 5D,F). We indeed observed that the ChipΔLID:ApΔLIM chimera rescued double mutants but, unexpectedly, this rescue was to the level of apGAL4/+ controls, significantly higher than that achieved in rescued ap single mutants (Fig. 5E,F). We suspect that the chimera cannot completely rescue the ap phenotype because there remain wild-type levels of endogenous Chip in ap mutants and that this pool of Chip is free to dimerize with the chimera and form non-functional complexes with vacant LID sites. Regardless of the mechanism, the complete rescue of double mutants reveals that the sole function of the Ap LIM domains in the nervous system is to mediate a complex with Chip.
DISCUSSION
Control of Drosophila wing patterning by Apterous (Ap) is accomplished through complex formation of Ap with dimers of the LIM-binding protein Chip (Milan and Cohen, 1999; van Meyel et al., 1999). Here we provide evidence that formation of this complex is also required for two distinct functions of Ap in the CNS. First, Chip is required to activate transcription from a small Tv neuron-specific enhancer of the FMRFamide gene in vivo. Second, physical interaction between Ap and Chip is required for correct axon pathway selection of Ap interneurons. Since we have found that Chip is a necessary cofactor for Ap not only in the wing but also in the establishment of neurotransmitter identity and axon guidance, we hypothesize that Chip is an obligate cofactor for all Ap activities. We also found that Chip is required for at least one function of Islet in the CNS, the activation of tyrosine hydroxylase expression in dopaminergic neurons. Given the widespread expression of Chip in Drosophila and of NLI in mouse, and given their high-affinity interactions with all LIM-HD proteins, Chip/NLI family members are likely cofactors for a wide range of LIM-HD activities in the transcriptional control of developmental programs.
Chip was first identified in a screen for mutations that reduce the activity of a remote wing margin enhancer in the cut locus, and subsequent genetic experiments imply that Chip mediates communication between enhancers and promoters over extremely long distances (Morcillo et al., 1996, 1997). The direct downstream effectors of LIM-HD transcription factors are largely unknown, as are the details of recruitment and coordination of LIM-HD proteins onto tissue-specific enhancers in vivo. We found that expression of the Tv-lacZ reporter transgene was disrupted in Chip mutant embryos, similar to ap mutants. These results suggest that Chip participates in the elaboration of a specific neuronal phenotype by cooperation with Ap to regulate expression directed by the 446 bp Tv neuron-specific enhancer of the dFMRFa gene. That such a small enhancer responds to Chip indicates that Chip can act locally to mediate protein interactions over short distances, presumably by bridging Ap molecules bound to the Ap response elements identified in vitro.
Additional components of Chip/LIM-HD complexes
It has been proposed that a single LIM domain can serve as a specific protein-binding interface, and that proteins with multiple LIM domains could function as scaffolds for the assembly of protein complexes (Schmeichel and Beckerle, 1994; Arber and Caroni, 1996). Indeed, biochemical assays and yeast interaction experiments have shown that one of the two LIM domains of all LIM-HD and LMO proteins is sufficient to interact with NLI/Ldb1 or the related protein CLIM-1a (Jurata et al., 1996; Bach et al., 1997), but that both LIM domains are required for high-affinity interactions (Jurata and Gill, 1997; Breen et al., 1998). For wing development and axon guidance, we have shown that the sole function of the Ap LIM domains is to bind Chip, and conclude that Ap function in these developmental processes does not entail LIM-mediated interactions with other proteins. However, it is curious to note that the LIM domains of Ap are more closely related to the LIM domains of its vertebrate orthologs Lhx-2 and Lhx-9 than they are to those of other Drosophila proteins like Isl or Lim3. This evolutionary conservation is characteristic of LIM-HDs and may reflect specific activities of LIM domains in cells or processes not yet analyzed. In support of this notion, the specificity of the interaction between LIM-HDs and certain transcription factors of other classes resides within the LIM domains. For example, the bHLH transcription factor E47 binds the LIM domains of Lmx-1 but not Islet-1 (Johnson et al., 1997), and Islet-1 binds the LIM domains of Lhx-3 but not other LIM-HDs (Jurata et al., 1998).
We previously demonstrated that, in Ap neurons, the N terminus of Ap, including the LIM domains, is not completely interchangeable with that of another LIM-HD family member (O’Keefe et al., 1998). This stands in contrast with Ap function in the wing disc where the N terminus is interchangeable. Barring technical issues related to protein stability or expression levels in the CNS, this result suggests that, in addition to binding Chip, Ap function in the CNS requires specific interactions with at least one other factor. Since we completely rescued the double mutant pathfinding errors with a chimera lacking Ap LIM domains, we speculate that such specific interactions must involve non-LIM domain sequences in the Ap N terminus (amino acids 1-147). In addition to these Ap-mediated interactions, it is likely that Chip also recruits other proteins of other classes into this multimolecular transcriptional regulatory complex (Bach et al., 1997, 1999).
The Chip/LIM-HD complex: a source of diversity and a substrate for modulation
Although it is clear that the formation of Chip/Ap complexes is critical in the CNS, as it is in the wing, our results suggest that there are at least two aspects in which control of their formation differs between these two tissues. First, the relative stoichiometry of Ap and Chip is an important determinant of wing patterning and outgrowth, as overexpression of Chip leads to defects that can be suppressed by co-overexpression of Ap (Fernandez-Funez et al., 1998; Milan and Cohen, 1999; van Meyel et al., 1999). In contrast, axon guidance and neurotransmitter expression in Drosophila embryos are insensitive to Chip overexpression (data not shown), and we show here that ChipΔLID does not influence axon guidance in a dominant negative fashion, in contrast to ChipΔDD. This suggests that, in the embryonic CNS, where Chip levels are high due to maternal expression, the crucial determinant of sufficient Chip/Ap complexes is the absolute levels of Ap rather than the relative stoichiometry of the two proteins.
Second, in the developing wing disc, upregulation of the LIM-only protein dLMO by Ap serves as a negative feedback mechanism by which Chip/Ap complexes are disrupted. However, dLMO is not expressed in the Ap neurons of the VNC, at least during embryonic axon pathfinding. It is possible, however, that another Drosophila LMO protein fulfills this task, or perhaps dLMO expression in Ap neurons commences later during larval or pupal development. Alternatively, downregulation of Chip/Ap activity might not be required in the CNS as it is in the wing. This latter alternative appears likely since the Chip/Ap fusion chimera, which is not susceptible to modulation by LMO factors, rescues perfectly the axon guidance errors of double mutants for Chip and ap. In contrast, in the wing, where dLMO interferes with Chip/Ap complex formation, the dLMO-insensitive Chip/Ap chimera appears to be hyperactive, phenocopying dLMO loss-of-function mutants (Milan and Cohen, 1999; van Meyel et al., 1999).
In Ap cells of the wing disc and CNS, there are as yet no other LIM-HD proteins known to be expressed. This would suggest that the relevant tetramer in these cells consists of two molecules of Ap bridged by two dimerized molecules of Chip. In the dopaminergic cells of the VNC, Isl is the only LIM-HD known to be expressed. It is not yet known whether TH is a direct target for transcriptional regulation by Isl, but the relevant complex for the regulation of TH expression, whether direct or not, may consist of two Chip-bridged molecules of Isl. However, Isl and Lim3 are co-expressed in the motor neurons of the intersegmental nerve branch b (ISNb), analagous to the combinatorial expression of LIM-HDs in vertebrate motor neuron pools (Sharma et al., 1998; Thor et al., 1999). Thus, in ISNb motor neurons, there are several possible combinations, including Chip-bridged complexes of two molecules of Isl, or two molecules of Lim3, or one molecule of each. In addition, Drosophila Lim3 and Isl, like their vertebrate counterparts, are uniquely capable of forming heterodimers in the absence of Chip, suggesting they can participate in both Chip-dependent and Chip-independent heterodimeric complexes (Jurata et al., 1998; van Meyel et al., 1999). Thus at least four possible combinations exist in ISNb motor neurons expressing only these two LIM-HD proteins and one cofactor. It is possible that each of these combinations has a unique specificity for target sequence recognition and that, from these four possible complexes, those that are actually formed during development may determine the repertoire of gene expression that gives these motor neurons their particular identities. In fact, biochemical studies in vitro suggest that Isl binds Lim3 with higher affinity than Chip and thus these direct Isl/Lim3 heterodimers may be favored over Isl/Chip complexes (van Meyel et al., 1999). Consistent with this possibility is that, in Chip mutant embryos, we detect no clear isl-like defects in pathfinding of the Isl motor neurons.
In vertebrates, the complexity of combinatorial expression patterns of LIM-HD proteins and the availability of potential cofactors suggests that the LIM-HD code is a major contributor to neuronal diversity. Recent results, including those presented here, indicate that the complexity of the LIM-HD combinatorial code is determined not only by the availability, concentration and relative affinities of LIM-HDs, but also of Chip/NLI, LMO proteins and possibly other cofactors.
Acknowlegments
We thank Dale Dorsett, Patrick Morcillo and Katie Gonzalez for generous provision of the Chip cDNA, antibody and mutant stocks, Paul Taghert and Ron Benveniste for Tv-lacZ flies, Steve Cohen and Marco Milan for the dLMO antibody, and members of the Thomas lab for insight and helpful discussions. This work was supported by Fellowships from the Medical Research Council of Canada and the Hoffmann Foundation to D. vM.; NIH Grant DK13149 to G. G. and NIH Grants to J. B. T.