Here we demonstrate, that two out of three predicted Vari proteins are in fact expressed in the embryo. Both proteins are also present in imaginal discs of third instar larvae and heads of adult flies (data not shown), while in ovaries only Vari-short could be detected. In the embryo, the two isoforms are differentially expressed, the smaller one being expressed earlier and much more abundant than the larger form. Localised Varicose protein is detected even later, after stage 12. This, together with the analysis of vari transcripts and proteins in the different mutants suggests that there are several levels of gene expression regulation. For example, in vari03953bthe longer transcript is highly upregulated, which is not reflected at the protein level. Here, the near absence of the small isoform seems to have an effect on the synthesis and/or stability of the longer isoform. In vari38EFa2 mutant embryos the truncated vari-long transcript is very abundant, which is not associated with more (truncated) Vari-long protein, pointing to less efficient translation an/or reduced stability of the mutant protein. In contrast, the truncated vari-short transcript is strongly reduced in abundance, which is not paralleled by a reduction in protein levels. In embryos mutant for this allele, properly localised protein can be detected, though in lower amounts (data not shown). Since the Vari antibody used does not allow discriminating between the two proteins, we cannot determine whether both isoforms are correctly localised. Although in variMD109, vari327 or variR979 one or both isoforms are synthesised in normal size and amount, in no case these proteins are correctly localised at the membrane [this work and ].
Although not predicted by commonly used domain prediction programs, sequence comparison of Vari-long with its closest vertebrate orthologues, MPP2_b, MPP6_c, and other related MAGUKs, makes the presence of a second, more divergent N-terminal L27 domain in the longer Vari isoform very likely. This situation is similar to the MAGUK Stardust, the L27N domain of which also fits less to the canonical sequence [5, 36]. The expression of two isoforms of a MAGUK protein, which differ by the presence or absence of the L27 domain(s) is not unique. For example, one close vari orthologue, human MPP6, also encodes two isoforms, one of which, MPP6_a, lacks the two L27 domains (Fig. 1B). The human postsynaptic density (PSD)-95 protein and Drosophila Discs large also come in two variants, one with and one without a L27 domain, respectively [4, 37]. In the case of Discs large, the DlgA isoform is predominantly expressed in the embryo and Dlg-S97 in the adult brain, but both isoforms are co-expressed in the larval NMJ . In the case of Vari, both isoforms are expressed in embryonic epithelia. Since the antibody used here recognises an epitope common to both Vari isoforms, we cannot completely rule out the possibility that different embryonic epithelia express different Vari isoforms, although this seems unlikely due to the interdependence between Vari-long and Vari-short. Our data further suggest that they are localised at the septate junction, since both co-immunoprecipitate with NrxIV-GFP, which is localised in the septate junction. Targeting and/or stabilisation of Neurexin IV are probably mediated by direct interaction between the PDZ domain of Vari and the C-terminal amino acids of Neurexin IV. However, other partners of Vari, particularly those interacting with the L27 domains, are still elusive. Similar as human VAM-1/MPP6_c, which binds human Veli-1 in vitro , the L27 domain of Drosophila Vari can interact with the L27 domain of DLin-7 in vitro. The different localisation of the two proteins, at least in epithelia of the Drosophila embryo, however, makes their in vivo interaction in these cells very unlikely.
What could be the significance of expressing two Vari isoforms? L27 domains can mediate the targeting of the respective protein to a particular membrane compartment, such as the synapse or the adherens junction, or stabilise interacting proteins by directly binding to the L27 domain of the respective partner [37, 40]. On the other hand, MAGUK proteins without L27 domains can efficiently be brought to their proper site, using other targeting mechanisms, for example palmitoylation . This raises the question as to i) whether the two Vari proteins rely on different mechanisms for targeting to the septate junctions and ii) whether the two Vari isoforms have specific functions in the Drosophila embryo. Using Gal4-mediated overexpression, either of them is capable to rescue variF033 mutant embryos to viability. This allele has been classified as a null allele, based on its genetic behaviour  and the complete and nearly complete lack of Vari-long and Vari-short, respectively (this work). The rescuing capability of Vari-short indicates that the L27 domains are not essential for viability, but does not exclude any non-essential function, in the embryo or at later stages. Strikingly, the hypomorphic allele vari38EFa2, which still expresses the short Vari isoform, but a modified Vari-long protein, gives rise to weak, but viable and fertile adults. The deletion in this allele removes 51 amino acids in the N-terminus, which affects both L27 domains. The fact that the escapers do not exhibit any mutant eye or wing phenotype similar to that obtained upon RNAi-mediated knockdown of vari in imaginal discs suggests a more supportive function for the larger isoform. The predominant role of Vari-short is further highlighted by the fact that variMD109, in which Vari-short is absent, is lethal, despite expression of normal Vari-long proteins. Hence, physiological amounts of only Vari-long are not sufficient for viability, but excess levels and/or earlier expression (using daG32/daughterlessGal4) can restore viability in the absence of Vari-short.
Mutational analysis of vari327 uncovered a five amino acid deletion in the SH3-domain of Vari, which almost completely removes one of the four core β-strands present in all SH-3 domains, thus completely abolishing Vari function. Although both Vari isoforms are expressed in vari327 in wild-type amounts, this allele is a functional null and no localised protein could be detected. The fact that the mutant proteins are not localised suggests that either the SH3 domain is necessary for targeting, or that the overall structure of the protein is affected, preventing proper localisation. Structural and functional analysis of other MAGUKs, e. g. PSD-95 or hCASK, point to either intra- or intermolecular interactions between the SH3 and the GUK domain [42, 43]. In the MAGUK PSD-93, binding of a ligand to the PDZ domain releases intramolecular inhibition of the GUK domain by the SH3 domain .
A strong reduction of varicose function by RNAi in postembryonic stages also affects the normal development of eyes and wings. It is interesting, however, to note that the consequences of RNAi-mediated knockdown of vari in wing and eye imaginal discs seem to be different. Reduced vari function in wing imaginal discs attenuates N signaling, as revealed by lowered activity of a N reporter gene construct and disrupted expression of the N target gene wingless at the prospective margin of wing imaginal discs. In wild-type wing imaginal discs N is activated on both sides of the dorsal/ventral compartment boundary by its two ligands, dorsally expressed Serrate and ventrally expressed Delta. N activity in the wing margin activates downstream genes, like cut or wingless, which are involved in the regulation of growth and patterning [reviewed in ]. Reduction of N activity results in the formation of notches in the margin, as observed here. In contrast, the N reporter seems to be normally activated in eye imaginal discs upon vari RNAi induction. Eye imaginal discs with reduced vari function display abnormal folding and adult eyes are smaller and misshapen. Additionally, ocelli and bristles are sometimes replaced by bare head cuticle. This phenotype is reminiscent of eye phenotypes observed in certain allelic combinations of coracle, which exhibit roughening of the eyes due to abnormally spaced ommatidia, but without affecting the patterning of the photoreceptor cells, and often lack ocelli and bristles. In addition, some coracle mutations suppress the effects of hypermorphic mutations in the EGF-receptor .
This suggests that SJs are differentially required for normal signalling at various developmental stages, but may affect different signalling pathways in a tissue dependent way. Given that SJs are required throughout imaginal disc development as suggested by our results, their lack may affect various signalling pathways, which are spatially and developmentally regulated and integrated as shown for the EGF-receptor and Notch pathway . So far, nothing is known how SJs may affect signalling. They could be involved in the correct localisation of receptors, membrane-bound ligands and/or components involved in signal transduction. The stage and tissue-dependent differential contribution of various signalling pathways may explain the different phenotypes obtained upon RNAi-mediated vari knockdown in eye and wing discs.
Septate junctions in larval eye imaginal discs have been well documented before , but their exact role during postembryonic development is largely unknown. NrxIV has recently been shown to be required for septate junction formation between and among the cone and pigment cells and for ommatidial morphology and integrity . Some of the phenotypes observed in NrxIV mutant clones in the developing eye, which are reminiscent to those obtained by eyGal4-mediated overexpression of vari-RNAi, have been interpreted as the result of compromised adhesion. Based on the molecular and genetic interaction between NrxIV and Vari, it is tempting to speculate that vari has a comparable role during eye development. In addition, during morphogenetic movements, cell divisions and cell rearrangements, SJs have to be redistributed . Loss of vari may thus interfere with these processes, which could explain the abnormal folding of eyGal4>UAS vari-RNAi eye imaginal discs.
Our data suggest that the final vari mutant phenotype is the consequence of compromised SJ function at different stages of development, which, in turn, may affect several cell-cell signalling and adhesion processes. A detailed dissection of the complexity of the mutant phenotype is beyond the scope of this manuscript. In the future it provides, however, a well-suited system to study the postembryonic function of SJs.