Distinct functions of α-Spectrin and β-Spectrin during axonal pathfinding

The development of multicellular animals is tightly linked to the evolution of a dynamic cell capable of forming and stabilizing manifold types of differentiation. This is most beautifully seen in the developing nervous system. Here, neurons migrate to their final destinations while, at the same time, the growth cones of the developing axons migrate towards their targets. In addition, dendritic specializations are formed and are kept in a dynamic equilibrium.

The dynamic changes of cell shape, as well as the subsequent stabilization of a specific form, demand a molecular machinery that can sense and transmit extracellular signals to the cytoskeleton. An important structural element that links cell adhesion proteins in the cell membrane to the F-actin cytoskeleton is the sub-membranous Spectrin network. Spectrins were first identified as important determinants defining the biconcave shape of erythrocytes. Erythrocytes that lack the Spectrin-based cytoskeleton loose their shape and stability, resulting in severe anemia in humans(Gallagher, 2004; Tse and Lux, 1999). Now,Spectrins are recognized as a large class of proteins ubiquitously expressed during development. The Spectrin proteins organize an extended protein network just below the plasma membrane by linking different actin fibers and many other proteins by numerous interaction motifs, such as the SH3 domain inα-Spectrin (α-Spec) (Bialkowska et al., 2005; Nedrelow et al.,2003). Furthermore, a pleckstrin-homology (PH) domain inβ-Spectrin (β-Spec) allows its direct binding to membrane lipids(Williams et al., 2004). One of the best-characterized adaptor proteins that binds to Spectrin is Ankyrin,which can mediate interaction with other cellmembrane-associated receptors or channel proteins (Bennett and Chen,2001; De Matteis and Morrow,2000).

Generally, Spectrins are long rod-like structural proteins and have been found in all metazoan species. α- and β-Spectrin form antiparallel dimers that associate in a head-to-head fashion to form an(αβ)₂ hetero-tetramer(Bennett and Baines, 2001). In humans, Spectrins are found in all cells, and several different α- andβ-Spectrin isoforms exist (Bennett and Baines, 2001; Berghs et al.,2000; Dhermy,1991). In Drosophila, only one α-spectringene, one β-spectrin and oneβ _H-spectrin gene have been described(Dubreuil et al., 1989; Dubreuil et al., 1990; Lee et al., 1993; Thomas and Kiehart, 1994; Thomas et al., 1998). The Drosophila α- and β-Spectrins share about 60% sequence identity with their human homologs, whereas the β_H-Spectrin isoform shares only 34% with its human homolog. β_H-Spectrin expression is found only in epithelial cells, whereas α- andβ-Spectrin are ubiquitously expressed during development.

Within Drosophila epithelia, the(αβ)₂-Spectrin tetramer is found at basolateral membranes, whereas the (αβ_H)₂-Spectrin tetramer is localized to the apical membrane domain only(de Cuevas et al., 1996; Dubreuil et al., 1997; Lee et al., 1997; Pesacreta et al., 1989). At cell-cell contact zones, the basolateral (αβ)₂-Spectrin is recruited to Neuroglian, a Drosophila homolog of the L1-cell adhesion molecule, via the adaptor protein Ankyrin. An additional protein recruited to the basolateral cell membrane is the Na⁺/K⁺-ATPase(Dubreuil et al., 1996; Dubreuil et al., 1997; Dubreuil et al., 2000; Nelson and Veshnock, 1987). A functional correlate of these interactions has recently been demonstrated in the neuromuscular junction. Here, reduction of both α- andβ-Spectrin not only leads to a mis-localization of Neuroglian and Fasciclin II, another cell adhesion protein, but finally results in a destabilization and retraction of the synapse(Featherstone et al., 2001; Pielage et al., 2005).

In Drosophila, mutations in all spectrin genes have been identified. Flies lacking β_H-spectrin function show reduced viability only and surviving flies exhibit relatively mild phenotypes,arguing against an essential function in determining epithelial-cell polarity(Zarnescu and Thomas, 1999). In contrast to β_H-spectrin, α-spectrinand β-spectrin are both essential genes. Only 50% of the embryos lacking zygotic α-spectrin expression reach the larval stages,and the ones that do die in the first-instar stage(Dubreuil et al., 2000; Lee et al., 1993). The survival rate of homozygous-mutant β-spectrin animals is even further reduced, and less than 10% of the mutant animals are able to leave the egg shells (Dubreuil et al.,2000). Clonal analyses have revealed essential functions ofα -spectrin in the polarity of follicle cells(Lee et al., 1997; Thomas et al., 1998). Althoughα-Spectrin and β-Spectrin are both found in a common protein complex, β-spectrin-specific functions have been described(Dubreuil et al., 2000). Epithelial cells of the midgut lacking α-spectrin normally position the Na⁺/K⁺-ATPase in their cell membrane. By contrast, β-spectrin mutants show an abnormal Na⁺/K⁺-ATPase distribution, suggesting thatβ-Spectrin can function independent of α-Spectrin.

Both α- and β-Spectrin are required during neuronal development. In C. elegans, it has been shown that β-spectrin is required for normal axonal outgrowth and fasciculation(Hammarlund et al., 2000). In mammalian axons, the Spectrin proteins are required to stabilize transmembrane proteins at the nodes of Ranvier(Lacas-Gervais et al., 2004; Yang et al., 2004). In Drosophila, α- and β-Spectrin have been shown to be involved in synapse organization and stability(Featherstone et al., 2001; Pielage et al., 2005). Here,we report the characterization of two mutants that were previously identified in a large phenotypic screen for genes affecting axonal pattern formation in the Drosophila embryo (Hummel et al., 1999a; Hummel et al.,1999b). We demonstrate that klötzchen and karussell (kus) encode α-Spectrin and β-Spectrin proteins, respectively.

The kus (β-spectrin) phenotype is characterized by a slit-sensitive crossing of the CNS midline by Fasciclin II-expressing axon fibers. Most prominently, we detected enlarged growth cones in single-cell analyses. Cell-type-specific genetic-rescue experiments demonstrated a requirement for β-Spectrin in cortical neurons. This requirement, in part, may include non-autonomous effects, because single neurons cannot be rescued in homozygous-mutant kus embryos. We demonstrate that α-Spectrin protein levels are tightly coupled to the levels of β-Spectrin by post-translational mechanisms, suggesting thatα -spectrin mutants may share β-spectrin-mutant phenotypes. However, within the nervous system, β-Spectrin appears to act independently of α-Spectrin.

We further demonstrate that klötzchen mutants, which were initially selected based on a phenotype distinct from β-spectrin mutants,affect the α-spectrin locus. We show that a reduction inα-Spectrin levels render the animal very sensitive to background mutations and temperature. klötzchen-mutant flies that do not carry a background mutation did not show a mutant CNS phenotype and,similarly, germline clones of hypomorphic alleles did not result in an abnormal nervous system phenotype. Thus, within the Drosophilanervous system, β-spectrin appears to act independently ofα -spectrin to stabilize neuronal growth cones.

To generate β-Spectrin-specific antibodies, we cloned 1686 bp starting from the ATG of β-spectrin into pQE31. The resulting fusion protein was expressed and purified according to the manufacturer's instructions (Qiagen),and was used to immunize rabbits (Davids, Regensburg).

To generate a UAS::β-spectrin construct we used the cDNA clone AT24411 (BDGP), which contains the 5′ region of theβ -spectrin mRNA. The 3′ third of the mRNA was cloned via a RT-PCR approach (details are available on request). The resulting clones were subcloned in pUAST and the sequence was verified by sequencing. Subsequent germline transformation was performed according to standard procedures. Several independent insertion lines were tested, which all showed similar effects. An UAS::β-spectrin^dsRNA construct was made using a 500 bp fragment from the 5′ region of theβ -spectrin open reading frame (ORF). Several independent transgenic lines were established. Only lines that led to a lethal phenotype when crossed to daughterless-Gal4 were used in this study.

EMS mutagenesis of isogenic chromosomes has been described by Hummel et al.(Hummel et al., 1999a). All complementation analyses were performed at 25°C under standard conditions. The duplication Dp(1;3)B^S3iD2 (provided by R. Dubreuil,University of Illinois, Chicago, IL), which rescues the lethality of five kus mutant chromosomes, was used. To remove the background lethal mutations from the α-spectrin^E2-26 mutation, we first exchanged most of the third chromosome using the rucuca multi-marker chromosome. Subsequently, all recessive mutations of the rucuca marker were removed following recombination. The resulting E2-26 chromosome could be rescued to full viability using an ubi:α-spectrinmini-gene (Lee et al., 1993). The following alleles were used: α-spec^lm88,α -spec^rg41 (Lee et al., 1993); α-spec^E2-26,α -spec^D4-65, α-spec^N2-141,β -spec^G113, β-spec^E175,β -spec^E292, β-spec^H127,β -spec^L105, β-spec^M046,β -spec^S012(Hummel et al.1999a);α -spec^N-2, α-spec^P-2,α -spec^S-1, α-spec^1.3,α -spec^1.2.1 (this work);β -spec^em6, β-spec^em15,β -spec^em21(Dubreuil et al., 2000); slit^B1-32 (Hummel et al., 1999a); and gcm^P1(Jones et al., 1995). To identify mutant animals, we employed GFP- or lacZ-labeled balancer chromosomes. The FRT elements ubi::GFP FRT19A and rsp17⁴ P[white⁺]70C FRT80B were used,and mitotic recombination was induced by an ey::Flp transgene(Bloomington stock center). The following Gal4 strains were used: ptc::Gal4, da::Gal4, ap::Gal4, elav::Gal4(Bloomington Stock Center); sim::Gal4(Scholz et al., 1997); and repo::Gal4 (provided by B. Jones, University of Mississippi,USA).

Immunohistochemistry was performed as previously described(Hummel et al., 1999a). The following antibodies were used: mouse anti-Wrapper(Noordermeer, 1998);anti-α-Spectrin, anti-Repo, BP102, anti-Fasciclin II (Developmental Studies Hybridoma Bank); rabbit anti-β-Galactosidase (Cappel); anti-myc 9E10 (Santa Cruz); anti-GFP (Invitrogen); anti-Kette(Bogdan and Klambt, 2003);anti-HRP-Cy5 (Dianova); and rabbit anti-β-Spectrin (this work). DiI labeling of individual neuroblast cell clones was performed as previously described (Bossing and Technau, 1994; Bossing et al., 1996).

We have previously identified seven independently induced mutations in the Drosophila gene kus in a large-scale mutagenesis screen for defects in embryonic axon pattern formation(Hummel et al., 1999a; Hummel et al., 1999b). All kus mutants were characterized by their having a similar unique CNS phenotype that was not found in any other complementation group(Hummel et al., 1999a; Hummel et al., 1999b). In particular, stage-14 kus-mutant embryos were characterized by axon structures that loop out into the CNS cortex, a phenotype that was never observed in wild-type embryos (Fig. 1C). Interestingly, the circular axon pattern phenotype is resolved in older embryos and, at the end of embryogenesis, commissures appear to be fused, which is often indicative of CNS-midline defects(Klämbt et al., 1991). In addition, the longitudinal axon tracts are found in closer proximity to the CNS midline and are often thinner as compared to the wild type(Hummel et al., 1999a; Hummel et al., 1999b)(Fig. 1D). All kusmutants that have been identified lead to very similar mutant phenotypes.

We were able to rescue the lethality of five kus alleles using the chromosomal duplication Dp(1;3)B^S3iD2, which affects the cytological interval 16A-D. The subsequent complementation assays that we performed showed that kus alleles cannot complement the lethality of the previously described β-spectrin alleles em15 and em21 (Dubreuil et al.,2000). To further confirm that kus encodesβ-Spectrin, we sequenced the first 4000 bp of theβ -spectrin ORF in three mutant kus alleles and found a mutation resulting in a stop codon at position 538 of the deducedβ -spectrin ORF in the allele S012(Fig. 2A). In addition, we assayed β-Spectrin protein expression in homozygous-mutant kusembryos that were selected using a twist-GFP FM7 balancer chromosome. When using a newly generated antiserum against the N-terminus ofβ-Spectrin, most kus mutants revealed altered β-Spectrin protein expression in western blot experiments(Fig. 2B). The previously described allele em15 and our allele E292 encode proteins that presumably lack the C-terminal PH domain, but may retain the Ankyrin-binding domain. The strongest previously knownβ -spectrin allele, em6, generates a 190 kDa large protein and thus may not be a complete null. The proteins encoded by G113,L015 and M046 are all significantly shorter compared with the wild-type protein. In the L015 mutant, the remaining protein appeared to encompass only two or three Spectrin repeats. Interestingly, all truncatedβ-Spectrin proteins were relatively stable.

In protein extracts of hemizygous-mutant E175, H127 and S012 mutants, we could not detect any β-Spectrin protein. Sequence analysis predicted that S012 mutants could generate a 20 kDa large protein fragment, which, however, was not detectable in western blot experiments. In summary, these data show that kus mutants affect theβ -spectrin locus. Despite these differences in β-Spectrin protein sizes in the different kus and/or β-spectrinmutations, we did not observe significant qualitative differences in the axonal phenotypes between these alleles. This indicates that a functional,full-length β-Spectrin protein is required for normal nervous system development.

It was previously shown that β-Spectrin is expressed ubiquitously during Drosophila development(Dubreuil et al., 2000). Within the developing nervous system, β-Spectrin is expressed in all neurons (see Fig. S1A in the supplementary material). Superficially,expression in the axonal compartment appeared somewhat lower when compared with the neuronal cell bodies. However, relatively high levels of β-Spectrin expression can be found on specific axonal fascicles in both the connectives and the commissures (Fig. 3A, arrowheads). Within the commissures, the level ofβ-Spectrin expression is further modulated and appears highest at the CNS midline (Fig. 3A, arrow). Co-expression with Wrapper, a specific marker for the midline glial cells(Noordermeer et al., 1998),demonstrated an overlap of this β-Spectrin expression domain with the midline glial cells (see Fig. S1C,C′ in the supplementary material). Unfortunately, the resolution of the confocal microscope did not allow us to determine whether the enhanced levels of β-Spectrin expression at the midline were due to axonal or glial β-spectrin expression.

Co-expression with the glial marker Repo(Campbell et al., 1994; Halter et al., 1995; Xiong et al., 1994)demonstrated that β-Spectrin is also expressed in the longitudinal glia(see Fig. S1D,D′ in the supplementary material). As glial processes invade the longitudinal connectives, β-Spectrin expression within the longitudinal connectives might be due to glial expression. To discriminate between glial versus axonal expression, we analyzed gcm mutants,which have no lateral glial cells (Hosoya et al., 1995; Jones et al.,1995; Vincent et al.,1996). gcm-mutant embryos still showed a fasciculatedβ-Spectrin staining pattern, indicating that axons and not glial cells are the major source for β-Spectrin staining within the longitudinal connectives (Fig. 3A,B).

To further test whether expression in neurons or glia is required for theβ -spectrin^kus axon phenotype, we performed cell-type-specific rescue experiments using Gal4/UAS-mediated expression ofβ -spectrin. Ubiquitous expression of a UAS-β-spectrin construct induced by a daughterless::Gal4 driver rescued hemizygousβ -spectrin^kus-mutant animals to full viability (data not shown). Following expression of β-spectrin in all lateral glial cells using a repo::Gal4 driver, we did not observe any alterations in the phenotypic strength (data not shown). Similarly, whenβ -spectrin was expressed in all CNS-midline cells using the sim::Gal4 driver, the β-spectrin^kus-mutant phenotype was not rescued (data not shown). However, when we expressedβ -spectrin in all postmitotic neurons using the elav::Gal4 driver, we noted a complete rescue of the axonal patterning defects (Fig. 1). However, neuronal expression could not rescue the lethality associated withβ -spectrin^kus mutants, demonstrating thatβ -spectrin has additional essential functions outside of the nervous system (Dubreuil et al.,2000). To further test the requirement ofβ -spectrin^kus for Slit-Roundabout (Robo) signaling,we performed single-cell rescue experiments using both sim::Gal4 and apterous::Gal4 (ap::Gal4) driver strains. When ap::Gal4 was used to express β-Spectrin no phenotypic rescue of their trajectory is observed in only a few cortical neurons and theβ-Spectrin-expressing fascicles are located in a unchanged position compared with the kus mutant (Fig. 4B-C′). Similarly, we failed to obtain cell-specific rescue in the MP1 fascicle following the expression of β-spectrin using the sim::Gal4 driver (Fig. 4D-F′).

These results suggest that β-spectrin^kus is required in cortical neurons for normal pathfinding and, furthermore, indicate that community effects are important in steering growth cones to their correct targets.

To understand better how β-Spectrin affects axonal pattern formation,we employed additional single-cell markers that reveal contralateral projections. In wild-type embryos, the SemaIIb:τmyc marker(Rajagopalan et al., 2000) is expressed in only one neuron per hemineuromere; this neuron projects its axon across the midline and the axon then follows a specific path within the longitudinal connectives (Fig. 3C,C′). In β-spectrin^kus mutants,specification of the SemaIIb neurons is not affected and the overall axonal trajectories are unchanged. However, the structure of the SemaIIb:τmyc-positive axon fascicle is severely altered in these mutants and the precision of axonal pathfinding is disrupted(Fig. 3D,D′). Although the normally straight axonal projection across the midline appeared irregular,ectopic crossings of the CNS midline were never observed(Fig. 3D,D′). In addition, the position of the SemaIIb:τmyc-expressing cell bodies was often shifted towards the CNS midline.

Very similar observations were made when analyzing the progeny of single labeled neuroblasts in stage-16 embryos(Fig. 5). Following the labeling of 210 individual DiI-labeled neuroblasts in 70β -spectrin^kus- mutant embryos, we found mostly normal projection patterns and did not observe contralateral-projecting axons that ectopically crossed or illegitimately re-crossed the CNS midline. However,axons displayed abnormal varicosities and additional small, ectopic side-branches (Fig. 5). Most prominently, we noted alterations in the structure of the growth cones, which appeared enlarged with sometimes extensive, filopodia-like processes(Fig. 5A′,C′;arrowheads). Thus, β-Spectrin appears to be required to establish or maintain the structure of growth cones needed for precise axonal patterning.

A hallmark of the β-spectrin^kus phenotype is that the longitudinal connectives are located closer to the CNS midline in mature embryos (Hummel et al., 1999a; Hummel et al., 1999b)(Fig. 1). Correlating with the slight collapse at the CNS midline, we noted mild axon-crossing defects in hemizygous-mutant β-spectrin^kus animals. Whereas, in wild-type embryos, Fasciclin II-positive axon tracts never cross the CNS midline, we observed rare ectopic crosses of the midline inβ -spectrin^kus-mutant embryos(Fig. 6). As midline crossing of Fasciclin II-positive axons is often sensitive to the dose of the midline repellent Slit, we performed gene-dosage experiments. Heterozygous slit-mutant animals, as well asβ -spectrin^kus/+; sli/+double-heterozygous female embryos, never showed ectopic crossing of Fasciclin II-positive axon tracts (Fig. 6). When we removed one copy of slit in a hemizygousβ -spectrin^kus embryo, the number of neuromeres with ectopic midline crosses increased significantly (P<0.001; t-test), suggesting that kus function is somehow integrated in the Slitsignaling pathway (Fig. 5). Similar results were obtained for the kus^G113 mutant, in which a truncated β-Spectrin protein was expressed (Fig. 2;data not shown).

In addition to affecting β-spectrin expression, all of theβ -spectrin^kus alleles affectα -spectrin protein levels(Fig. 2C). Interestingly, the effects on α-Spectrin levels do not correlate with the extent of the C-terminal β-Spectrin deletions. This is in agreement with the notion that the C-terminal domain of β-Spectrin is required to bind, and thus presumably stabilize, α-Spectrin(Deng et al., 1995; Yan et al., 1993). To determine the subcellular localization of α-Spectrin and the truncatedβ-Spectrin proteins we stained whole-mountβ -spectrin^kus-mutant embryos and, in addition,generated homozygous-mutant β-spectrin^kus tissue in the eye antennal imaginal discs using the Flp/FRT system. When we generated mutant clones using the two hypomorphic β-spectrin^kusalleles G113 or E292, we could detect reduced levels ofβ-Spectrin expression that was still correctly localized at the cell membrane and a concomitant reduction in α-Spectrin protein levels(Fig. 7A,B). To test whetherα-Spectrin is similarly required for the stability of β-Spectrin protein, we generated homozygous-mutant eye discs lackingα -spectrin expression. In these experiments we observed the loss of α-Spectrin in the mutant clone but did not detect any change in the level of β-Spectrin that was still correctly localized to the subcortical region of the cell (Fig. 7C). Thus, in contrast to α-Spectrin, β-Spectrin can localize independently to the cell membrane. To further analyze the close regulatory interaction between these spectrin genes, we performed gain-of-function studies and expressed full-length β-Spectrin in wild-type wing imaginal discs. This not only led to a clear up-regulation ofβ-Spectrin protein expression but also to a concomitant increase in the levels of α-Spectrin protein (Fig. 7D). As RNA levels were not affected (data not shown), we conclude that there exists a post-translational mechanism stabilizing α-Spectrin protein. Together, these experiments demonstrate an intimate regulation ofα- and β-Spectrin, and show thatβ -spectrin^kus mutants are functionalα - and β-spectrin double mutants.

Our data demonstrate that β-spectrin^kus-mutant embryos are impaired in both α-Spectrin and β-Spectrin expression. In our recent large-scale EMS mutagenesis that led to the identification ofβ -spectrin^kus mutants, we identified several other mutations affecting the formation of the segmental commissures(Hummel et al., 1999a; Hummel et al., 1999b). One of the complementation groups identified was klötzchen, whose members show a commissural phenotype distinct to that of kus mutants(Hummel et al., 1999a)(Fig. 8B). Using standard genetic-mapping techniques, we localized the common lethality of five independent alleles to the chromosomal interval 61F-62A and subsequently showed that these mutants affect the α-spectrin locus. Sequence analysis of the klötzchen allele E2-26 revealed a A→T mutation at position 250 of the α-spectrin ORF that leads to the termination of translation after 83 amino acids (instead of 2416 amino acids for the full-length α-Spectrin protein). Thus, E2-26 probably represents a null allele. Interestingly, heterozygousα -spectrin-null mutations have relatively normal levels ofα-Spectrin protein, indicating a strict regulation of the expression levels (Fig. 8G, compare lanes w¹¹¹⁸ and Df(3L)aprt/+).

klötzchen mutants were initially isolated based on their embryonic CNS phenotype; however, following the removal of all lethal background mutations on the klötzchen^E2-26 chromosome by meiotic recombination (see Materials and methods), no obvious axonal phenotypes were detected (Fig. 8C). Similarly, the previously describedα -spectrin-null mutation rg41(Lee et al., 1993) does not lead to an embryonic CNS phenotype. In addition, we did not see any Fasciclin II-positive axonal tracts crossing the CNS midline in homozygous E2-26 mutants (40 embryos scored, Fig. 6C). The distribution ofβ-Spectrin protein in the longitudinal axon tracts did appear slightly altered, which might be due to fasciculation defects(Fig. 3A; Fig. 8E,F). Thus, zygoticα-Spectrin is not required for normal axonal patterning, but loss ofα -spectrin renders neurons sensitive to background mutations. In agreement with this notion, we noted ectopic midline-crossing of Fasciclin II-positive axons when we removed one copy of slit in a homozygousα -spectrin-mutant background(Fig. 6C).

In an independent experiment that aimed to clarify the genetic organization at the genomic interval spanning the α-spectrin-discs lost region, we generated 21 additional EMS inducedα -spectrin alleles. All mutants were lethal in trans to deficiencies of the region (Df(3L)Aprt32 or Df(3L)My10) and most animals died during the first-larval-instar stage. Two alleles, P-2 and 1.3, appeared to have normal levels ofα-Spectrin protein (Fig. 8G) and were lethal during the early third-larval-instar stage,indicating that they are hypomorphic alleles. Interestingly, some of theseα -spectrin mutants showed temperature-dependent intragenic complementation (see Table 1). Similarly, the axonal phenotype of α-spectrin alleles appeared temperature-sensitive (Fig. 8D). As the temperature sensitivity is not due to an altered stability of the α-Spectrin protein in these animals (see Fig. S2 in the supplementary material), the Spectrin network itself might be temperature labile.

The formation of the neuronal network requires a multitude of cellular interactions for precise axonal pathfinding and the establishment of specific synaptic connections (Dickson,2002). In an attempt to identify some of the key regulators that control axonal pattern formation, we have isolated mutants that have specific defects in the developing embryonic nervous system of Drosophila(Hummel et al., 1999a; Hummel et al., 1999b). Here,we present the identification and further characterization of two of these mutations, kus and klötzchen.

Rescue experiments and direct sequence analysis demonstrate that kus encodes the Drosophila β-Spectrin protein. kus mutants were initially isolated due to a distinct axonal phenotype, including ectopic CNS-midline crossing of Fasciclin II-positive axons. To further analyze this phenotype, we labeled the progeny of single neuroblasts in kus-mutant embryos but, despite the large number of labeled clones, we were unable to detect any aberrant midline crossings. Similarly, when we employed cell-type-specific Gal4 drivers, we could not see clear pathfinding defects across the midline. It is therefore likely that the observed phenotype is a result of inappropriate contact between medial Fasciclin II-expressing axons from both sides of the midline mimicking ectopic midline crossings. Because wild-type Slit levels are required to position the longitudinal fascicles, a reduction of slit gene dosage results, as expected, in a further medial positioning of the longitudinal connectives,explaining the increase in the number of ectopic midline crosses in these animals. Similar phenotypes were also observed by Garbe et al.(Garbe et al., 2007).

Interestingly, we found defects in the architecture of the neuronal growth cones in β-spectrin^kus-mutant animals, which may explain the general sensitivity ofβ -spectrin^kus-mutant neurons to guidance signals such as Slit. The enlarged growth cones detected inβ -spectrin^kus mutants correlate nicely with data on growth cone formation after axotomy(Gitler and Spira, 1998);axonal injury leads to an increased activity of the protease calpain, which cleaves Spectrin and results in the removal of the submembranous Spectrin meshwork prior to the regeneration and growth of the growth cone(Gitler and Spira, 1998). In secretory cells, the submembranous Spectrin cytoskeleton prevents the premature fusion of vesicles with the plasma membrane(Aunis and Bader, 1988; Perrin et al., 1992). Similarly, Spectrins may function to regulate the fusion of intracellular membrane vesicles needed to enlarge and advance the growth cone(Gitler and Spira, 1998),which could explain the enlarged growth cones that we detected inβ -spectrin mutants (Fig. 4).

Within the Drosophila nervous system, α- and β-Spectrin are the only Spectrins that are expressed. These two proteins form a heterodimer in which β-Spectrin appears to be the key determinant,because α-Spectrin protein is only stable in the presence ofβ-Spectrin and ectopic expression of β-Spectrin leads to a concomitant increase in the level of α-Spectrin protein. To test whether this regulation occurs at the level of RNA or protein, we determined the expression of the corresponding transcripts, but noted no alteration (data not shown). It is possible that the association of α- and β-Spectrin blocks ubiquitination of α-Spectrin and its subsequent degradation via the proteasome. Ubiquitination has been previously reported forα-Spectrin (Corsi et al.,1995; Galluzzi et al.,2001) and may thus help to define the correct protein-expression levels.

Despite the intimate coupling of the two expression profiles, it has been demonstrated that β-spectrin can function independently ofα-Spectrin. During the development of the midgut, the correct localization of the Na⁺/K⁺-ATPase requires onlyβ -spectrin, but not α-spectrin, function(Dubreuil et al., 2000). Similarly, the phenotypes associated with the different spectrin mutants isolated in this study are distinct. Whereas β-spectrin leads to a typical looping of CNS axons during stage 13, no abnormal axonal phenotypes could be detected for α-spectrin alleles. Similarly, we failed to detect any midline phenotypes for Fasciclin II-positive axons inα -spectrin^E2-26-null mutants. A possible explanation to this phenotypic discrepancy may be the maternal contribution ofα -spectrin; however, a similarly strong maternal component has been described for β-spectrin. Attempts to generateα -spectrin germline clones using the null allele rg41failed because of an essential function of α-spectrin during oogenesis (de Cuevas et al.,1996). To circumvent this maternal α-spectrinfunction, we employed the hypomorphic α-spectrin-mutant allelesα -spec^N-2, α-spec^P-2 or a-spec^1.3 to generate germline clones using the ovo^D/FRT system. However, embryos with both impaired maternal and zygotic α-spectrin expression displayed no nervous system phenotype, supporting the notion that β-spectrin acts independent of α-Spectrin protein.

α-spectrin mutations turned out to be sensitive to background mutations and temperature effects. In addition to the phenotypic effects of uncharacterized background mutations, we detected a temperature dependence of the spectrin-mutant phenotypes and temperature-dependent intragenic complementation of hypomorphic α-spectrin alleles. It is well-known that microtubule dynamics depend on temperature and that microtubules depolymerize in the cold(Osborn and Weber, 1976). As microtubule stability is already compromised in α-spectrinmutants (Pielage et al.,2005), any further destabilization might have significant effects on (neuronal) development (Dent and Gertler, 2003). Alternatively, temperature sensitivity might reflect differences in the efficacy of endocytosis. Although we cannot pinpoint the molecular mechanism underlying the temperature sensitivity of spectrin mutants, we can conclude that Spectrins act as a global stabilizing protein network that coordinates a large variety of membrane receptors, including the Robo receptor that is needed to sense the Slit protein.

We thank G. Bashaw for communication and discussion of results prior to publication; and L. Goldstein, B. Dickson, R. Dubreuil, D. Branton, B. Jones and the Bloomington Stock Center for sending flies and antibodies. J.P. was supported by a pre-doctoral grant of the Boehringer-Ingelheim foundation. This work was funded by grants of the DFG (SPP 1111).