In both vertebrates and invertebrates, members of the LIM-homeodomain (LIM-HD) family of transcription factors act in combinatorial codes to specify motoneuron subclass identities. In the developing Drosophila embryo, the LIM-HD factors Islet (Tailup) and Lim3, specify the set of motoneuron subclasses that innervate ventral muscle targets. However, since several subclasses express both Islet and Lim3, this combinatorial code alone cannot explain how these motoneuron groups are further differentiated. To identify additional factors that may act to refine this LIM-HD code, the expression of POU genes in the Drosophila embryonic nerve cord was analyzed. The class III POU protein, Drifter (Ventral veinless), is co-expressed with Islet and Lim3 specifically in the ISNb motoneuron subclass. Loss-of-function and misexpression studies demonstrate that the LIM-HD combinatorial code requires Drifter to confer target specificity between the ISNb and TN motoneuron subclasses. To begin to elucidate molecules downstream of the LIM-HD code, the involvement of the Beaten path (Beat) family of immunoglobulin-containing cell-adhesion molecules was analyzed. beat Ic genetically interacts with islet and Lim3 in the transverse nerve (TN) motoneuron subclass and can also rescue the TN fasciculation defects observed in islet and Lim3 mutants. These results suggest that in the TN motoneuron context, Islet and Lim3 may specify axon target selection through the actions of IgSF call-adhesion molecules (Certel, 2004).

In each abdominal hemisegment of the Drosophila embryo, the axons of ~40 motoneurons exit the ventral nerve cord and specifically synapse with 30 identified muscle fibers. The axons of these motoneurons form six discrete fascicles and exit the VNC together, extending into the periphery to innervate specific muscle groups (or fields). The ventral muscles are innervated by the motoneurons of three subclasses: the transverse nerve (TN), intersegmental nerve b (ISNb) and intersegmental nerve d (ISNd). In addition, the segmental nerve c (SNc) innervates a set of externally located ventral muscles. The two TN motoneurons contact ventral muscle 25 or the ventral process of the lateral bipolar dendritic neuron (LBD). The ISNb motoneurons innervate the ventral muscles 6, 7, 12, 13, 14, 28 and 30, and the ISNd motoneurons muscles 15, 16 and 17. The Drosophila LIM-HD genes, islet and Lim3, are co-expressed in a subset of CNS neurons, including several classes of motoneurons. Although Isl and Lim3 are required in these distinct motoneurons to specific neuronal identity, this two member 'LIM-code' cannot alone be responsible for the unique differentiation and function of each subclass. More specifically, since the TN and ISNb motoneuron subclasses both express a combination of Isl and Lim3, it is likely that other factors act to discriminate between these two subclasses (Certel, 2004).

To identify factors that act to further differentiate these motoneuron subgroups, the expression pattern of previously characterized transcription factors was examined. Isl and Lim3 are co-expressed with the class III POU factor, Drifter (Dfr), in a limited number of embryonic VNC neurons. Immunohistochemical experiments using double transgenic animals reporting on Isl and Lim3 expression and antibodies directed against Dfr reveal restricted triple expression specifically in the ISNb motoneuron subclass. Dfr, Isl and Lim3 expression is clearly observed in the RP1, RP3, RP4 and RP5 motoneurons that project out the ISNb fascicle to innervate ventral muscles 6, 7, 12 and 13. Dfr expression is not detected in the pair of Isl- and Lim3-expressing TN neurons. To confirm that the lateral co-expressing neurons belong to the ISNb subclass, embryos carrying the Lim3A-tau-myc transgene were double-labeled. This reporter construct drives expression in Isl-expressing motoneurons that project via the ISNb but not the ISNd fascicle. The lateral Dfr-expressing neurons do belong to the ISNb subclass since they also express the Tau-myc fusion protein. The co-expression of Dfr and Isl was further analyzed. In addition to the ISNb neurons, Dfr and Islet co-expression is also observed in two serotonergic EW neurons (Certel, 2004).

To verify the lack of Dfr expression in the ISNd subgroup, focus was placed on the well-characterized GW/7-3M motoneuron that innervates muscle 15 via the ISNd pathway. To identify this ISNd motoneuron, the enhancer-trap line, eg^P289, was used that drives lacZ expression in the 7-3M progeny. Although Dfr expression is seen in the EW interneurons, the ISNd GW/7-3M motoneuron does not express Dfr. The results of these labeling experiments demonstrate that Dfr expression can be further used to subdivide the ISNb from the TN and ISNd neuronal classes (Certel, 2004).

The ISNb fascicle contains motor axons from at least eight motoneurons innervating ventral muscles 6, 7, 12, 13, 14, 28 and 30. Loss of isl or Lim3 function in ISNb motoneurons affects axon targeting resulting in a reduction of target muscle innervation. The most common phenotype in both isl and Lim3 mutants is a failure to innervate the cleft between muscles 12 and 13. In Lim3 mutants, muscles 12/13 were not innervated by motor axons in 46% of hemisegments compared with 3% in wild-type hemisegments. In isl mutants, the lack of muscle contacts was also coupled with the ISNb motor axons, leaving the ventral muscle field and joining the TN (26% of hemisegments in isl mutants versus 0% in wild type) (Certel, 2004).

Dfr is functionally required in a variety of tissues, including the midline glia, trachea, wing veins, sensory neurons and antennal lobe projection neurons. The early lethality of null dfr alleles, and the VNC perturbations caused by midline glia defects made analyzing the requirement for Dfr function in ISNb neurons difficult. To circumvent this problem, and to examine whether Isl, Lim3 and Dfr are required to specify similar aspects of motor axon targeting, the ISNb fascicle was analyzed in trans-heterozygous combinations. Removing a single copy of isl, Lim3 and dfr results in motor axon targeting defects characterized by significant reductions in muscle innervation. The failure to innervate muscles 12 and 13 observed in both isl^37Aa/+;dfr^B129/+ and Lim3^Bd1/+;dfr^B129/+ trans-heterozygotes is easily quantifiable and observed in isl and Lim3 mutant embryos. Reducing the levels of Isl and Dfr also results in ISNb motor axons leaving the ventral muscle field and targeting the TN fascicle. Neither Dfr nor Isl is expressed in motoneurons that innervate the dorsal muscles or the externally located ventral muscles. In line with their restricted expression, no evidence was found of non-cell autonomous affects upon the targeting of the ISN and SNc motor axons in dfr, isl trans-heterozygotes (Certel, 2004).

In an attempt to address the specificity of the genetic interactions between POU and LIM-HD genes, genetic interactions were tested between dfr and another key regulator of ISNb identity, the Drosophila exex/hb9 gene. However, no evidence of ISNb axon pathfinding defects were found in embryos trans-heterozygotes for dfr and hb9 (dfr^B129/hb9^KK30). Therefore, results from genetic interaction studies indicate that Dfr may be required specifically in combination with Isl and Lim3 to specify the ISNb motoneuron subclass (Certel, 2004).

isl and Lim3 mutants display axon targeting defects manifested by a reduction in muscle innervation. Results from the trans-heterozygous genetic interaction studies suggest that Dfr is necessary in combination with Isl and Lim3 to specify ISNb axon targeting. If this hypothesis is correct, then a loss of Dfr function in ISNb motoneurons should also result in a reduction in muscle innervation. To address the possible role of Dfr in ISNb specification, RNA interference was used to reduce or eliminate the expression of Dfr protein specifically in neurons, without affecting its expression and function in midline glia. Transgenic flies were generated expressing double-stranded dfr RNA (UAS-dsdfr) under the control of neuronal-specific Gal4 drivers. To further assist in the removal of Dfr function, two independent UAS-dsdfr insertions were recombined onto a chromosome containing a null dfr allele, dfr^B129. Although the protein produced by this allele is detected by the Dfr antiserum, it is non-functional because of a premature stop codon located before the DNA-binding POU domain. Several Gal4 drivers were used to analyze the effectiveness of the UAS-dsdfr transgenes. Using both the C155(elav)-Gal4 and the Lim3B-Gal4 drivers, Dfr protein was reduced or eliminated in the majority of ISNb motoneurons, including the RP motoneurons. As a control, the Dfr-expressing midline glia showed no loss of Dfr protein and, accordingly, no loss of the midline glia-specific marker Slit. This shows that UAS-dsdfr could be used together with neuronal-specific Gal4 drivers to address how loss of Dfr function affects ISNb neuron specification (Certel, 2004).

Using both early (ftz_ng-Gal4.20) and post-mitotic (Lim3B-Gal4) drivers to reduce Dfr activity, two axon outgrowth phenotypes were observed. (1) ISNb motor axons can now leave the ventral muscle field and contact or target the TN fascicle. This re-targeting suggests that losing Dfr leaves the ISNb neurons in an Isl/Lim3-only specified state generating a TN fate. (2) Reducing Dfr function causes an overall decrease in muscle innervation by the ISNb motor axons similar to defects observed in isl mutants. Both of these axon targeting and innervation phenotypes indicate that Dfr plays an important role in ISNb neuronal specification (Certel, 2004).

If Dfr functions as part of a LIM/POU combinatorial code to specify ISNb motoneurons, then ectopically expressing Dfr in the Isl/Lim3-expressing TN motoneurons would be predicted to alter TN axon pathfinding toward an ISNb-like behavior. To test this, Dfr was misexpressed in postmitotic Isl/Lim3-expressing TN neurons using the Lim3B-Gal4 driver. In wild-type development, the transverse nerve forms from the fasciculation of a sensory nerve axon (the lateral bipolar dendrite or LBD neuron) and the TMNp neuron. At late stage 15/16, these growth cones contact each other on the ventral interior muscle surfaces. The TN fascicle is on a different focal plane and in wild-type embryos does not come in contact with the ISNb fascicle (Certel, 2004).

Misexpression of Dfr protein in Lim3B-Gal4;UAS-dfr double transgenic embryos did not have any effect on TN axons exiting the ventral nerve cord or lead to random innervation in the periphery. Instead, adding Dfr function to the TN neurons caused the motor axons to target the ISNb muscle field in a significant number of hemisegments (52%). TN motor axons either send collaterals into ISNb muscle targets or in some cases even innervated ISNb muscles. This change in motor axon targeting appears to be specific to the TN motoneurons. Misexpressing Dfr in most, if not all, motoneurons via the ftz_ng-Gal4.20 did not result in any targeting changes or defects in the SNc or ISN fascicles; it did, however, result in a similar percentage of TMNp targeting defects (Certel, 2004).

The ability of the TN motoneurons to be respecified by Dfr misexpression was tested in isl mutant embryos. Without Isl function, the retargeting of TN motor axons did not occur, suggesting possible cooperative actions between these transcription factors. These results indicate that it is the addition of Dfr specifically to the TN Isl/Lim3 LIM-HD code that allows this motoneuron subclass to exhibit ISNb motoneuron characteristics (Certel, 2004).

In C. elegans touch receptor neurons, the POU protein UNC-86 directly regulates the expression of the LIM-HD gene, MEC-3. And in vertebrates, abLIM is a transcriptional target of the POU factor, Brn3.2. To determine whether Dfr, Isl and/or Lim3 are possible transcriptional targets of one another, the individual expression patterns were examined in each mutant background. The results indicate that in Drosophila Dfr, Isl and Lim3 must function at the same hierarchical level in ISNb motoneuron specification (Certel, 2004).

Studies in various model systems have led to the identification of a number of transcription factors with highly restricted expression. These factors are important for different aspects of neuronal differentiation, including axon pathfinding. By contrast, many other molecules, such as cell adhesion molecules, receptor protein tyrosine phosphatases and semaphorins, also play a crucial role in axon guidance, yet these molecules are often, at least in Drosophila , more broadly expressed. This apparent disjunction could indicate that the downstream genes regulated by highly restricted transcription factors such as Dfr and Isl are thus far uncharacterized molecules. Therefore, to identify possible targets of the LIM/POU or LIM-HD combinatorial codes, specific molecules were sought that are differentially expressed between the ISNb and TN motoneuron subclasses. Restricted expression is exhibitied by 14 Drosophila Beat-like members of the immunoglobulin superfamily (IgSF) of CAMs. Members of the Beat family in Drosophila and the Zig family in C. elegans contain two Ig domains and most members have been shown to be highly restricted in their expression pattern, largely confined to subsets of neurons (Certel, 2004).

beat Ic is of particular interest because of its expression in a small number of embryonic neurons that include the TN motoneurons but not the ISNb neurons. In situ hybridization and immunohistochemistry was used to verify the TN expression of beat Ic. beat Ic transcripts are expressed in lateral cell clusters that contain the TN neurons but not in the ISNb RP neurons. In addition to restricted TN expression, removal of beat Ic function through overlapping deficiencies results in TN phenotypes indistinguishable from those observed in isl and Lim3 mutants. Embryos that lack beat Ic display errors in TN fasciculation. The TMNp axon fails to completely fasciculate with the LBD projection, resulting in bifurcation and aberrant ventral muscle exploration. This TN fasciculation in beat Ic mutants is improved by elav-Gal4 driven UAS-beat Ic expression. In isl mutants, the TN axons either do not exit the VNC or fail to fasciculate with the LBD projection, also leading to aberrant ventral muscle exploration. This failure of the TMNp and LBD axons to adhere is also observed in a significant number of Lim3 mutant hemisegments (Certel, 2004).

The similarity of TN mutant phenotypes suggests that Isl, Lim3 and Beat Ic are required to specify analogous aspects of motor axon targeting. To test this hypothesis, the TN fascicle was analyzed in trans-heterozygous combinations. Mutations in beat Ic are not available, therefore, defined deficiencies were used. Embryos that lack one copy of isl, Lim3 and beat Ic display a significant number of TN fasciculation defects. As in individual mutants, the TMNp and LBD axons in several trans-heterozygous combinations fail to fasciculate, thus leading to bifurcation and/or aberrant axon outgrowth. Two chromosomal deficiencies removing beat Ic, beat Ic^Df and beat Ic^Df1 were used to demonstrate that other genes removed by an individual deficiency did not influence the TN phenotype. In addition, at least two different isl and Lim3 alleles were used to verify that the observed genetic interactions were not due to specific chromosomes or alleles. Another Beat family member, beat Ib, is also expressed in the CNS; however, a lack of Beat Ib does not result in any defects in neuronal development. It was not possible to test for genetic interactions with any other Beat family members because neither overlapping deficiencies nor individual mutations are available for these genes. However, to assess the specificity of these interactions among Ig domain-containing molecules, genetic interactions were examined between isl, Lim3 and the CAM Fas2, which contains five Ig-like domains. No fascicle defects were detected in various trans-heterozygous combinations. These results raise the possibility that in TN motoneurons Isl and Lim3 may specify axon target selection through the actions of specific IgSF CAM members (Certel, 2004).

At least two possible explanations can be put forwards to explain the TN defects observed in isl and Lim3 mutants. First, the defects observed are because the TMNp and the LBD axons cannot fasciculate or adhere to one another. A second hypothesis is that these two axons do not recognize each other and therefore do not grow close enough together to fasciculate properly. If the defect lies in fasciculation, then increasing the levels of Beat Ic, a promoter of motor axon adhesion, should reduce the TN defects in isl and Lim3 mutants. To increase Beat Ic levels, the pan-neuronal driver elav-Gal4 and UAS-beat Ic transgenes were crossed into isl and Lim3 mutant backgrounds. As in previous experiments, strong isl and Lim3 alleles were crossed to isl^Df and lim^Df deficiencies to create embryos null for isl and Lim3, respectively (Certel, 2004).

Increasing the levels of Beat Ic through one copy of UAS-beat Ic and the elav-Gal4 driver significantly reduces the percentage of TN defects in isl mutants from 56% to 22%. In addition, the severity of TN adhesion defects decreased in the remaining 22% of affected hemisegments. Likewise, increasing Beat Ic levels in Lim3 mutants also significantly decreased the occurrence of TN defects from 55% to 31%. These results suggest that the TN defects observed in isl and Lim3 mutants are a result of a decrease in the adhesive properties between the TN motor axon and the LBD projection. This suggests that Beat Ic may be a direct transcriptional target of the LIM code (Certel, 2004).

At this time, this hypothesis could not be tested because of the unavailability of a Beat Ic antibody. Attempts were made to analyze beat Ic transcript accumulation in isl and Lim3 double mutants but it was not possible to achieve cellular resolution in the mutant ventral nerve cords. However, DNA-binding site pattern searches using previously described LIM-HD binding motifs indicate that there are significant clusters of LIM-HD sites surrounding and within the beat Ic locus. The Isl1 consensus site (CTAATG) is found 15 times in the chromosomal region encompassing the beat Ic locus; the Lhx3-binding site (AATTAATTA) is found nine times and 21 Isl 2.2 sites (YTAAGTG) have been identified. Although functional analyses will be needed to determine if these sites are necessary and/or sufficient for beat Ic expression, searching the entire genome with the Isl 2.2 site places the beat Ic locus seventh out of the first 10 identified. Furthermore, studies in C. elegans indicate that two LIM-HD genes, the Lmx-class gene lim-6 and the Lhx3-class gene, ceh-14, are required for the expression of at least four IgSF zig genes (Certel, 2004).

Therefore beat Ic is transcribed in a small number of cells in the embryonic nerve cord, including the transverse motoneurons. Loss-of-function experiments indicate Beat Ic is required in a pro-adhesive manner for the proper recognition/or fasciculation of the TMNp motor axon and the LBD fascicle. These TN defects are identical to the fasciculation defects observed in isl and Lim3 mutants. Trans-heterozygous combinations reveal strong genetic interactions between isl, Lim3 and beat Ic, and furthermore, increasing Beat Ic expression in isl and Lim3 mutants significantly rescues these TN axon fasciculation defects (Certel, 2004).

LIM-HD proteins participate in a number of unique complexes through protein-protein interactions mediated by their LIM domains. For example, LIM domains can interact with the widely expressed co-factor, NLI(Ldb1/CLIM-2) in mice or Chip in Drosophila. The NLI/CHIP proteins homodimerize and generate a bridge between two LIM-HD proteins, thereby leading to the formation of tetrameric complexes. Although this complex is functional in vivo, it has been found that other proteins also participate to generate further tissue specificity. In Drosophila, the newly identified Ssdp protein interacts with Chip to modify the activity of complexes comprising Chip and the LIM-HD protein Apterous in the wing. In vertebrates, the bHLH factors Ngn2 and NeuroM functionally interact with the NLI, Isl1 and Lhx3 complex to initiate motoneuron differentiation. These results suggest that the formation of further specialized combinations could be used to confer not only tissue but also cellular specificity (Certel, 2004).

In Drosophila , Isl and Lim3 are co-expressed in a subset of CNS neurons including two neuron subclasses, the TN and ISNb motoneurons. Although Isl and Lim3 are required in these distinct motoneurons to specific motor axon pathway choice neuronal identity, these two factors alone cannot be responsible for the unique differentiation of each subclass. In this study, evidence has been provided that the class III POU domain protein Dfr functions in combination with this Isl/Lim3 LIM code, to specify the ISNb motoneuron class. Loss-of-function analyses indicate each of these transcription factors is required in the ISNb neurons for the specification of motor axon target selection. Without Dfr, Isl or Lim3, these motor axons fail to correctly innervate their designated muscle targets. In addition, genetic interaction studies suggest that this phenotype indicates a common aspect of motoneuron designation has been altered (Certel, 2004).

How might this LIM/POU code function in ISNb neurons? In C. elegans touch receptor neurons, the LIM-HD factor, MEC-3 and the POU protein UNC-86, physically interact to control specification. In the pituitary, the LIM domain of Lhx3 (P-Lim) specifically interacts with the Pit1 POU domain and is required for synergistic interactions with Pit1. Whether Dfr, Isl and or Lim3 physically interact to regulate ISNb motor axon target selection has not been determined. However, misexpression experiments indicate that the re-specification of transverse motoneurons by the addition of Dfr does require functional Isl protein. Although, this result does not distinguish between the possibilities of direct interactions between these proteins or the binding of a common transcriptional target, it does indicate that a functioning 'LIM code' is required for the re-specification of the TN neurons (Certel, 2004).

A second finding of the Dfr misexpression studies is that the target selection of postmitotic neurons can be robustly re-specified. The Lim3B-Gal4 line was used to add Dfr to the LIM-only transverse motoneurons. This Gal4 line does not activate reporter construct expression until stage 14 -- a post-mitotic stage even for the late developing transverse motoneurons. At this stage, the TN motor axons have exited the CNS and are navigating the periphery, although the TN motor axon and LBD fascicle have not come into contact. Misexpressing Dfr even at this relatively late stage of TN motoneuron differentiation can clearly alter axon pathfinding, and in a significant percentage of hemisegments, TN motor axons actually appear to ectopically innervate ISNb muscle targets. This result shows that these motoneurons remain plastic, even after becoming postmitotic and further indicate that the LIM/POU code may be acting directly on genes involved in axon targeting (Certel, 2004).

The combinatorial expression of LIM-HD transcription factors confers motoneuron subtypes with the ability to direct their axons to reach distinct muscle targets. If more than one subgroup of motoneurons use a LIM code, how does subtype-specific motor axon pathfinding occur? Presumably, it is the downstream targets of each LIM code that confer the ability of individual or groups of motor axons to find their correct innervation targets. What might be the target(s) of the LIM code in Drosophila transverse neurons? Studies in vertebrates and invertebrates have demonstrated that members of the IgSF class of CAMs play important roles in cell-cell recognition and communication -- processes that are crucial for nervous system wiring. The ability of Ig-domains to form linear rods when deployed in series, and their propensity to bind specifically to other proteins, has made these molecules ideal for functioning as cell-surface receptors and/or CAMs. Furthermore, IgSF molecules have dramatically increased the number of cell-cell recognition molecules through family expansion, the generation of multiple variants through alternative splicing, receptor multimerization and cross-talking intracellular signaling pathways (Certel, 2004).

A recent genomic analysis indicates the Drosophila immunoglobulin superfamily repertoire consists of about 150 proteins; in C. elegans, 80 immunoglobulin superfamily molecules are predicted. Members of the Beat family in Drosophila and the zig family in C. elegans are located in gene clusters, have restricted expression patterns and share the same domain architecture. Each protein is comprised exclusively of two Ig modules with either a transmembrane domain [beat Ic, beat Ib, beat IIa, beat VI (GPI), zig-1] or secreted signals (10 Beat genes: zig-2 to zig-8). The Beat family members that have been functionally characterized appear to control fasciculation through anti- and pro-adhesive properties. Furthermore, unlike many previously described CAMs, several members of these families are expressed in subsets of CNS neurons (Certel, 2004).