loco encodes an RGS protein required for Drosophila glial differentiation

Within the Drosophila nervous system, glial cells perform a variety of important functions. They are capable of controlling proliferation rates of neighboring neuroblasts (Ebens et al., 1993), they ensheath and electrically insulate axon bundles (Auld et al., 1995) and they are likely to secrete neurotrophic factors (Xiong and Montell, 1995). Furthermore, during embryonic development, neuron-glia interactions are required for the correct establishment of the axonal scaffold (Jacobs and Goodman, 1989; Klämbt and Goodman, 1991; Goodman and Doe, 1993; Hidalgo et al., 1995; Hosoya et al., 1995; Jones et al., 1995; Klämbt et al., 1996).

Based on position and gene expression, embryonic CNS glial cells can be divided into two major classes, the midline glia and the lateral glia (Klämbt et al., 1996). The lateral glial cells are a heterogeneous group of cells, characterised by the expression of a number of genes encoding transcription factors. The lineage relationships of these cells have been described (Jacobs et al., 1989; Bossing et al., 1996; Schmidt et al., 1997). Lateral glia development is initiated by the activity of the gene glial cells missing (gcm), which acts as a master regulatory gene of glial cell development (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). gcm encodes a novel transcription factor with an 8-nucleotide DNA-binding site (Akiyama et al., 1996; Schreiber et al., 1997). In its absence, glial-specific gene expression is abolished and glial cells are transformed into neuronal cell types (Hosoya et al., 1995; Jones et al., 1995; Bernardoni et al., 1997). gcm governs glial development by activating transcription factors such as TRAMTRACK, POINTED and REPO, which in turn control glial differentiation by two different mechanisms (Hosoya et al., 1995; Jones et al., 1995; Giesen et al., 1997). On the one hand, inappropriate, neuronal differentiation appears to be suppressed in developing glial cells by the BTB-Zn-finger domain protein TRAMTRACKP69 (Giesen et al., 1997). On the other hand, glial differentiation needs to be activated by transcription factors such as the POINTED ETS domain proteins or the REPO homeodomain protein (Campbell et al., 1994; Klaes et al., 1994; Xiong et al., 1994; Halter et al., 1995).

pointed encodes two proteins, PNTP1 and PNTP2, which share a common ETS DNA-binding domain. Within the embryonic CNS, both POINTED proteins are specifically expressed in non-overlapping sets of glial cells, with PNTP1 being expressed in many lateral glia and pntP2 being restricted to the midline glial cells (Klämbt, 1993; Scholz et al., 1997). In pointed mutant embryos, glial cells fail to differentiate and do not properly ensheath the neuropile (Klaes et al., 1994; Giesen et al., 1997). Ectopic expression and single cell transplantation experiments demonstrated that pointedP1 is not only required but also sufficient for several aspects of glial cell differentiation (Klaes et al., 1994). POINTEDP1 binds directly to DNA through a conserved consensus site and subsequently acts as a transcriptional activator (O’Neill et al., 1994; Albagli et al., 1996).

To understand pointed function during glial cell development, it is necessary to identify its glial-specific target genes. A number of approaches have been employed to identify direct targets of transcription factors (Gould et al., 1990; Wagner-Bernholz et al., 1991). Here we have utilized the enhancer trap technique to isolate a candidate target gene of pointed. The integration of a modified P-element in the vicinity of genomic enhancer regions often leads to a tissue- or cell-specific expression pattern of the lacZ reporter gene (Bellen et al., 1989; Bier et al., 1989). In many cases, lacZ expression precisely mimics the expression of a nearby endogenous gene. Thus genes can be isolated solely based on their expression pattern irrespective of their mutant phenotypes.

We have screened 11,000 enhancer trap insertion lines for expression in the developing CNS. About 30 lines lead to specific expression in various subsets of CNS glial cells. Among this collection, P-element insertions into the genes gcm, pointed, tramtrack and gliotactin have been identified (Klämbt, 1993; Auld et al., 1995; Jones et al., 1995; Giesen et al., 1997). Here we report on two P-element integrations at the cytological interval 94B/C. Within the CNS, β-GALACTOSIDASE expression pattern conferred by the lines 3-109 and rC56 is restricted to lateral glial cells (Klämbt and Goodman, 1991; Winberg et al., 1992) and corresponds well to the expression pattern of an endogenous gene near the P-element integration sites named loco. loco encodes the first two known Drosophila members of the RGS protein family, whose members are known to regulate G-protein signalling. One LOCO isoform is specifically expressed in glial cells. During glial cell development, loco is required for the correct extension of cell processes. loco-deficient glial cells fail to ensheath axons as well as failing to perform normal glia-glia cell interaction resulting in a loss of the blood-brain barrier. In a yeast two-hybrid screen, a specific interaction of LOCO with Gαi could be demonstrated. This interaction can also be mediated by protein domains outside the RGS domain.

Two P-element insertions into the loco gene were isolated in a large enhancer trap screen (C. K., A. Nose and C. S. G., unpublished data), loco^3-109 and loco^rC56, both of which carry a P[rosy, lacZ] insertion at the cytological position 94B/C. Both P-element insertion lines are homozygous viable. Four X-ray-induced alleles reverting the rosy marker carried by the rC56 enhancer trap insertion in the loco gene were recovered: Df(3R)23D1 deleting the cytological region 93F-94E; Df(3R)5C1 deleting the cytological region 93E/F-94C/D; Df(3R)15CE1 deleting the cytological region 93F-94C/D; Df(3R)17D1 deleting the cytological region 93E/F-94B/C.

Lethal loco alleles were recovered as transposase-induced revertants of the loco^rC56 allele. From 700 independent excision strains analyzed, 2 lines were lethal over the X-ray-induced deficiencies and failed to complement each other. The genomic breakpoints of loco^Δ13 were cloned by inverse PCR: 35 cycles of 10 seconds 94°C/1 minute 54°C/4 minutes 72°C) using primers PZ-3547r (5′-GTCGCCACCAATCCCCAT-3′) and PZ-3588 (5′-TGGAGCCCGTCAGTATCG-3′). EMS-induced loco alleles were generated following standard mutagenesis procedures (Lewis and Bacher, 1968) on an isogenised st e chromosome and identified by non-complementation of loco^Δ13 at 29°C.

All loco alleles were kept over a TM3 Sb balancer chromosome carrying a P[elav-lacZ] construct in order to allow the unambiguous identification of mutant embryos.

Immunohistochemistry and electron-microscopic analyses were performed as described previously (Klämbt et al., 1991; Stollewerk et al., 1996).

Genomic loco DNA sequences were isolated by plasmid rescue and used to screen a genomic EMBL4 library, kindly provided by M. Noll, Zürich. Genomic phages were hybridised to polytene chromosomes to verify the cloning of the correct chromosomal region at 94B/C. cDNA clones were isolated from a λgt11 library (Zinn et al., 1988). Nested deletions were constructed after DNase I treatment (Sambrook et al., 1989) and subsequently sequenced using a T7 based sequencing kit (Pharmacia) or an ABIPRISM 310 automated sequencer. The sequence of the longest cDNA clone for each transcript class (p109c1 corresponding to the glia transcript and p109c2 corresponding to the PNS transcript) was determined and confirmed by sequencing corresponding genomic clones. The clone p109c1 is 2918 bp in length. At base 207, there is an ATG in frame with a long open reading frame (ORF) coding for a hydrophilic protein of 829 amino acids, with a deduced molecular weight of approximately 97 kDa. The 3918 bp clone p109c2 corresponds to the loco-c2 transcript and encodes a protein of 1175 amino acids. The relative molecular weight of the deduced LOCO-c2 protein is 135 kDa. Sequence analysis was performed using the Lasergene software package.

Several differences between cDNA and genomic sequences were noted. The ones resulting in amino acid substitutions are (G204→C204; R18S) (C206→G206; N19K) (A316→G316; E56G) (A435→G435; I95V) (A718→G718; N190S) (C1101→T1101; L317W) (C3675→T3675; R1175stop). Homology searches were performed using FASTA (Altschul et al., 1990).

Yeast two-hybrid screen: strains, plasmids and procedures used for the two-hybrid screen were essentially as described (Golemis and Brent, 1997). Construction of Gαi bait: the Gαi-coding region was fused to the lexA-coding region, by cloning a BamHI-XhoI fragment containing the coding region of a Gαi cDNA clone into the bait vector pEG202-PL1, to make the Gαi bait plasmid pEG202GAiwt. pEG202-PL1 was based on the bait vector pEG202 (Gyuris et al., 1993; Golemis and Brent, 1997), but with some alterations to permit more flexibility in inserting bait sequences in frame with the upstream lexA sequence. pEG202 was cut with EcoRI and XhoI, and a fragment made from two 25-mer oligonucleotides that had been annealed to form a double-stranded linker with overhangs that could ligate to the cut pEG202, was inserted. This gave a polylinker sequence (codons shown in frame with upstream lexA) of CTG GAA TTC CGC GGA TCC ATG GCG GCC GCT CGA GTC GAC CTG AGC, and a vector with unique EcoRI, SacII, BamHI, NcoI, NotI, XhoI and SalI sites, with all sites except EcoRI in a different reading frame from pEG202.

The Gαi-coding region fragment was prepared as follows. (1) An EcoRI-XhoI C-terminal fragment of a Gαi cDNA clone, containing DNA from the EcoRI site at codon 157 to a HincII site 180 bp 3′ to the stop codon (Provost et al., 1988) was subcloned into these sites of pBluescript, to give pBSKS-GAiwt-RI-XhoI. (2) The 5′ end of the gene was amplified by PCR, using a left-hand primer that included the ATG initiation codon and added a BamHI site and an NdeI site that included the ATG, and a right-hand primer 3′ of the EcoRI site at codon 157. (3) The BamHI-EcoRI fragment of this PCR product was subcloned into pBSKS-GAiwt-RI-XhoI to give pBSKS-GAiwt-ATG-XhoI. (4) A BamHI-XhoI fragment, carrying the whole of the Gi gene, was cut out of this plasmid and cloned between the BamHI and XhoI sites of pEG202-PL1, to give pEG202GAiwt and putting Gαi in frame with lexA.

Library screening: the Drosophila ovary cDNA library RFLY3 (Finley et al., 1996) in vector pJG4-5 (Gyuris et al., 1993) was obtained from Russ Finley. This library is estimated to contain 3.2×10⁶ independent cDNA clones. An aliquot was used to generate over 10⁷ transformants in E. coli strain XL1-Blue, and these transformants were pooled and used to prepare library DNA. About 3×10⁷ transformants of yeast strain EGY48 (pSH18-34) that carried pEG202GAiwt were generated using this library preparation, of which about 350 had a LEU2+ phenotype that indicated a possible interaction. PCR amplification and restriction digestion was used to identify duplicate clones and 25 unique clones were identified. One clone of each type was used to retransform yeast strain EGY48 (pSH18-34) that carried pEG202-GAiwt. 16 clones conferred a LEU2+ phenotype that suggested a bona fide interaction; 9 clones did not. Neither the interacting clones nor the LexA-Gi fusion alone activated LEU2 expression. None of the loco clones showed any interaction with a number of other control LexA bait constructs, including a fusion of LexA to Drosophila neuronal synaptobrevin (DiAntonio et al., 1993).

As the ovary cDNA library had been oligo(dT)-primed (Finley et al., 1996), DNA sequencing of the 3′ end of each clone, using a primer that annealed to downstream pEG202-PL1 vector sequences, was used to sort 14 of the 16 clones into 6 different genes. The 5′ boundary of the loco fragment in each clone was determined by one or more sequencing reactions using a primer that annealed to upstream pEG202-PL1 vector sequences, followed by alignment with the existing loco sequence.

DNA probes were labelled by random priming. Non-radioactive in situ hybridisation experiments using digoxigenin-labelled probes were performed as described (Tautz and Pfeifle, 1989). The stained embryos were embedded in GMM (Ashburner, 1989) and photographed on a Zeiss Axiophot.

The ETS transcription factors encoded by pointed control CNS glial cell differentiation presumably through activation of target genes. In order to identify possible glia-specific target genes, we made use of the enhancer trap technique. Two enhancer trap lines, 3-109 (Klämbt and Goodman, 1991) and rC56, were selected based on their specific β-GALACTOSIDASE expression in the lateral CNS glia (Fig. 1). Both lines show identical β-GALACTOSIDASE expression patterns and carry a P-element insertion at the cytological position 94B/C. In embryos carrying the rC56 enhancer trap insertion, first β-GALACTOSIDASE expression can be detected in early stage 12 in cells which, based on their position, appear to be the progeny of the lateral glioblast (Fig. 1A). Interestingly, at this early stage these cells appeared to be already different. The anterior pair of progeny expresses elevated levels of β-GALACTOSIDASE. As CNS development continues, these cells migrate medially and divide (Jacobs et al., 1989; Halter et al., 1995). By the end of embryogenesis, most glial cells except the midline glia express β-GALACTOSIDASE. In addition, β-GALACTOSIDASE expression can be detected in the dorsal leading edge cells in the lateral ectoderm (data not shown). To test whether CNS glia expression depends on the function of pointed, we assayed β-GALACTOSIDASE expression directed by the rC56 enhancer in a pointed^8B74 mutant background (Fig. 1) or in rC56 pnt^8B74/pnt^Δ88 heterozygous embryos (data not shown). First expression directed by the rC56 element at early stage 12 in the progeny of the longitudinal glioblast appears to be slightly weaker in the mutant. During later stages of CNS development, a reduction in the expression level as well as in the number of cells expressing the rC56 reporter gene can be detected. In stage 16 mutant embryos, the activity of the rC56 enhancer is most prominently reduced or absent in the longitudinal glial cells, whereas A and B glial cells and the VUM glial cells appear relatively unaffected (Fig. 1F, see also Fig. 3G). The influence of pointed on the rC56 enhancer is less pronounced in the posteriormost 2-3 neuromeres. Furthermore, the rC56 reporter can be ectopically activated by ectopic expression of pointedP1 (Klaes et al., 1994). This suggests that the lacZ gene located in the rC56 P-element is under at least partial control of a pointed-dependent genomic enhancer element, which encouraged us to look for a corresponding gene in the vicinity of the P-element insertion.

35 kb of genomic DNA sequences flanking the 3-109 P-element insertion site were isolated. The location of the P-elements rC56 and 3-109 is indicated in Fig. 2. The mapping of exons by restriction analysis and genomic sequencing revealed two different variants differing in their 5′ ends (c1 and c2) of a gene that we named loco (locomotion defects, see below) (Fig. 2, see below). In situ hybridisation experiments with transcript-specific digoxigenin-labelled cDNA probes showed that both loco RNA classes are expressed in the embryo (Fig. 3). loco-c1 transcription is very weak and is detected only after prolonged incubation (6-12 hours) in the staining solution. Using a 200 bp loco-c1-specific probe, expression can be first detected in late stage 12 embryos (Fig. 3D). In stage 16 embryos, loco-c1 RNA is found in the leading edge cells (Fig. 3F, arrowhead), in the tracheal cells (Fig. 3D,F, asterisk) and in the lateral glial cells within the CNS (Fig. 3E, arrow). Except for loco expression in tracheal cells, this corresponds well with the β-GALACTOSIDASE expression pattern observed for the two P-element insertions in the loco gene. loco-c2 transcripts are found only in scattered cells in the lateral ectoderm. Based on their position, these cells might correspond to PNS progenitor cells (Fig. 3A-C). No expression can be detected in the CNS.

The sequences of the different cDNA clones were determined and compared to genomic sequences to deduce the exon-intron structure. We observed a number of differences between genomic and cDNA sequences (see Materials and Methods). The glia-specific exon of loco-c1 encodes only the initiator methionine, such that the remaining cDNA sequences are shared with the loco-c2 transcript. A second, in-frame ATG is found 46 codons 3′ of the c1 initiator. The introns between exons 2/3 and 3/4 are 67 and 72 bp in size.

The deduced protein sequences of both transcripts were compared with those in the EMBL database using the FASTA program (Altschul et al., 1990). A conserved domain of 125 amino acids, encoded by exon 2, identifies LOCO as a member of the RGS (Regulators of G-Protein signalling) protein family (Figs 2, 4, 5). To date about 16 different members of the family have been described. loco represents the first Drosophila RGS family member. Within the RGS domain, LOCO is most similar (>40% homology) to rat RGS12 and rat RGS14 (Snow et al., 1997). Interestingly, homology to rRGS12 and rRGS14 extends beyond the RGS domain to the C terminus of the deduced LOCO sequences. Three additional regions of homology were designated as B (48 of 160 aa are identical in LOCO and rRGS12 which corresponds to 30% identity), C (14/59 identical, 26% identity), and D (20/51 identical, 39% identity) (see Figs 4, 5). No homologies were found in the LOCO-c2-specific domain.

To determine the function of loco during glia development, we first generated chromosomal deficiencies removing the loco gene (see Materials and Methods). Subsequently, we mobilized the P[rosy, lacZ] insertion rC56 and generated 700 independent P-element excision lines. Among these lines, two non-complementing lethal mutations, loco^Δ13 and loco^Δ293, were recovered. These mutations also failed to complement the X-ray-induced deficiencies, indicating that the lethal hit was in the vicinity of the P-element insertion. The extent of the genomic deletions associated with the two alleles was determined by Southern blot analyses and is indicated in Fig. 2. In loco^Δ13, the proximal deletion breakpoint lies within the P-element leaving the lacZ gene intact. However, the relative level of β-GALACTOSIDASE expression in different glial cells appears to be altered and, in loco^Δ13 embryos, the longitudinal glial cells express only low levels of β-GALACTOSIDASE (data not shown). This indicates that glia-specific enhancer elements reside upstream of the rC56 enhancer trap insertion whereas a transcriptional activator acting specifically in the longitudinal glial cells must reside 3′ of the rC56 insertion. The breakpoint in loco^Δ13 was cloned and is at least 7 kb downstream of the loco gene. A small inversion as well as a deletion of about 2 kb of genomic sequence is associated with the loco^Δ293 allele (Fig. 2). Here, putative promotor sequences as well as the first exon are deleted. Both mutations are homozygous embryonic lethal.

Additional loco alleles were obtained following EMS mutagenesis (see Materials and Methods). From 3.088 chromosomes analysed, we obtained five mutations, defining a single complementation group (see Table 1 for details). They are lethal in trans to both loco excision mutations at 29°C. Based on the complementation analyses, they were placed into the following allelic series: loco^L1>loco^F1>loco^T1≈loco^A1>loco^M1.

At 25°C, loco^M1 but none of the other EMS-induced alleles produces adult escapers when in trans to locoΔ13 or loco^Δ293. About 10% of the expected numbers of transheterozygous adults appear. They often fail to eclose from the opened puparium. Eclosed flies show a paralytic phenotype and drop into the food and die. If such flies are rescued from the food, they show a severe impairment of spontaneous locomotor activity and display a ‘shaking’ phenotype. Response to mechanical stimulation (e.g. after stimulation of thoracic bristles) is weak in locoM1/locoΔ293 and undetectable in locoM1/locoΔ13, indicating that Δ13 is a stronger allele than Δ293. Similar phenotypes, albeit with lower expressivity, are seen in flies heterozygous for loco^M1 and other EMS-induced alleles at lower temperatures (see Table 1 for details). All adult escapers die after a maximum of 2 days.

The strongest loco allele is represented by the embryonic lethal mutation loco^Δ13. No abnormal CNS axon pattern phenotype was detected using the mAb BP102, which labels the overall axon pattern. mAb 1D4 recognizes the FASCICLIN II protein, which is expressed on a subset of longitudinal fascicles (Fig. 6A). In mutant loco^Δ13 but not in mutant loco^Δ293 embryos, a slight defasciculation of axons can be found. In addition, we observed an occasional crossing of fasciclin II-positive axons within the longitudinal connective (Fig. 6B).

To analyse the different glial cells in mutant loco embryos, we used anti-REPO antibodies (Halter et al., 1995) which label most lateral glial cells. No gross defects were detected in the number and position of these cells. This indicates that birth and migration of the lateral glial cells do not depend on loco function. To analyse terminal differentiation of glial cells, we used the M84 enhancer trap marker (Klämbt and Goodman, 1991). In wild-type stage 16/17 embryos, a regular pattern of evenly spaced glial cells can be detected (Fig. 6). In loco^Δ293 as well as in loco^L1 mutant embryos defects in the positioning of some of the M84-positive cells were observed. In particular, β-GALACTOSIDASE expression is reduced specifically in the A and B glial cells compared to more laterally positioned glial cells (Fig. 6C,D, arrows).

To obtain a higher level resolution, we performed an electron microscopic analysis of homozygous loco^Δ293, loco^Δ13 embryos as well as of loco^Δ13/Df(3R)15E1 embryos (Fig. 7). In wild-type embryos, the longitudinal glial cells cover the longitudinal axon tracts and extend processes into the connectives. Furthermore the entire cortex is covered by a thin perineurial glia sheath (Fig. 7A,C arrowheads). In loco mutant embryos, the differentiation of both glial cell types appears to be affected. In loco^Δ13/Df(3R)15E1 mutant embryos, the perineurial sheath is virtually absent (Fig. 7 compare A,C to B,F). This is most clearly seen at the CNS midline. In wild-type embryos, glial cells form a continuous barrier preventing the contact of midline cells with the hemolymph (Fig. 7A,B arrowheads; see G,H for enlargements).

A similar phenotype can be seen for the longitudinal glial cells. The nuclei are found at relatively normal positions, but no glial cell processes can be detected within the connectives. In addition, the intimate glial-glial cell contact, that is observed in wild-type embryos is severely disrupted in loco mutant embryos. Often we find axons on the dorsal surface of longitudinal glial cells, apparently in direct contact with the hemolymph (Fig. 7D-F, arrowhead).

In summary, the loco mutant phenotype can be described as a late glial cell differentiation defect, where the formation of glial cell processes enwrapping neuronal cell bodies and axons does not occur.

RGS domains directly interact with G-protein alpha subunits, displaying a remarkable degree of specificity (De Vries et al., 1995; Berman et al., 1996; Druey et al., 1996; Hunt et al., 1996; Watson et al., 1996). If LOCO indeed functions as a regulator of G-protein signalling, we would anticipate the presence of a G-protein in the lateral glial cells. We therefore analysed the expression of Gαs, Gαi and Gαo RNAs in the embryonic nerve cord and found that the Gαi subunit appears to be specifically expressed in the glial cells (Wolfgang et al., 1991; S. G. et al., unpublished data).

Further evidence for interaction of LOCO and Gαi was found in a yeast two-hybrid screen (Gyuris et al., 1993). A cDNA clone of Drosophila Gαi (Provost et al., 1988) was used as ‘bait’ for interacting proteins. The Gαi gene was fused in frame at its N terminus to a gene encoding a LexA DNA-binding domain. Yeast that expressed this fusion was transformed with a library carrying Drosophila cDNAs fused to a gene for a transcriptional activation domain (Finley et al., 1996). Clones that encoded putative Gαi-interacting proteins were identified by the ability of the transformed yeast colonies to express a LEU2 gene that contained LexA-dependent regulatory elements and the interaction was confirmed by reintroducing the putative positive clones into yeast that carried the LexA-Gαi fusion.

Six non-overlapping sets of interacting clones were identified. Four non-identical loco clones were recovered, with C-terminal fragments of various lengths fused to the lexA gene (Figs 2, 5). The longest fragment began at residue 443 of the predicted LOCO c2 protein and included the RGS domain; the shortest encoded only 199 amino acid residues that extended C-terminal from residue 977 of the predicted LOCO c2 protein and included the final 43 amino acids of the conserved region D closest to the C terminus (Figs 2, 5). None of the loco clones showed any interaction with several control LexA fusions (see Materials and Methods). Thus LOCO appears to be an RGS domain protein specific for Gαi.

In the present paper, we describe the Drosophila gene loco and show for the first time that RGS domain proteins are required for correct glial differentiation. loco is expressed in lateral glial cells throughout development and appears to be a target gene of pointed. loco is required for the proper formation of the blood-brain barrier and the ensheathment of neuronal cell bodies and axons. It is coexpressed and able to specifically interact with Gαi suggesting a role for G-protein signalling in glial development.

Although glial cells are an important component in any complex nervous system, not much is known about the molecular mechanisms underlying glial development. In Drosophila, a number of gene functions and mechanisms required during glial development are emerging. Following lineage specification (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996), terminal differentiation of glial cells is mediated by transcription factors encoded by repo and pointed (Klaes et al., 1994; Halter et al., 1995).

The identification of genes activated by pointed in glial cells should provide new insights in the molecular mechanisms underlying glial differentiation. loco, identified by an enhancer trap approach, might represent such a pointed target gene. Analysis of the loco promotor region revealed the presence of GCM- and ETS-binding sites suggesting that loco might be a direct target of gcm as well (S. G., unpublished data). loco promotor-lacZ fusion constructs revealed a small promotor fragment that is capable of directing lacZ expression in almost all loco-expressing glial cells. This promotor fragment is indeed dependent on pointed function and ectopic pointed expression as well as ectopic gcm expression result in a corresponding ectopic lacZ expression. Sequence analysis and in vitro mutagenesis revealed both GCM- and POINTED-binding sites within this element (S. G. and C. K., unpublished data). These data, as well as the phenotypes observed in loco and pointed mutant embryos (Klaes et al., 1994), suggest that loco is indeed a target of pointed. However, it is important to emphasise that loco expression in the tracheal system does not appear to depend on pointed function (Fig. 3).

In loco-deficient embryos, glial development initially proceeds normally, however, glial-glial cell-cell interactions appear to be defective. Axons within the longitudinal connectives are not completely enwrapped and are occasionally found even at the dorsal surface of the CNS (Fig. 7D-F, arrowheads). A similar phenotype can be observed following a fasII antibody staining (Fig. 6B). In addition, the blood-brain barrier is not established. Due to the high potassium concentration in the hemolymph, this is likely to result in a disruption of axonal conductance. The adult, paralytic phenotype of the weak EMS-induced loco alleles might be a consequence of such a defect as well.

Similar phenotypes were found in neurexin or gliotactin mutants (Auld et al., 1995; Baumgartner et al., 1996). Here too, the formation of the blood-brain barrier is defective and the animals are paralysed. GLIOTACTIN is a transmembrane protein expressed by a subset of glial cells, NEUREXIN is a transmembrane protein that is required for the formation of septate junctions (Baumgartner et al., 1996). In contrast to the above mentioned proteins, LOCO is likely to be localized within the cell. What causes the mutant phenotype found in loco mutant embryos?

LOCO is the first known Drosophila member of the recently described family of Regulators of G-protein Signalling (RGS) (Koelle, 1997; Arshavsky and Pugh, 1998) with highest homology to RGS12. RGS proteins were first described in yeast and C. elegans (De Vries et al., 1995; Druey et al., 1996; Koelle and Horvitz, 1996). RGS proteins stimulate the GTPase activity of different Gα subunits as much as 100-fold (Watson et al., 1996), accelerating the transition from the GTP-bound active to the inactive GDP-bound form and thereby terminating trimeric G-protein signalling. Different RGS proteins vary in their specificities for the Gαi and Gαo subunits. To date, no RGS protein binding to Gαs has been identified. The RGS domain itself is sufficient for both binding Gα and GTPase activation (De Vries et al., 1995). Here, we have shown that LOCO physically interacts with Gαi. Strikingly, the interaction with Gαi is not confined to the RGS domain but can also be mediated by C-terminal sequences, possibly by a stretch of 51 amino acids that is conserved between LOCO and RGS12 (see Fig. 4). It is interesting to note that rat RGS12 also interacts with Gαi (Snow et al., 1998).

Several G-proteins have been identified in Drosophila (Wolfgang et al., 1990; Parks and Wieschaus, 1991; Quan et al., 1993; Wolfgang and Forte, 1995). Beside their role in phototransduction and learning (Scott et al., 1995; Connolly et al., 1996; Zuker, 1996), only few functions have been associated to date to G-proteins (Parks and Wieschaus, 1991). Interestingly Gαi RNA (but not Gαs and Gαo) is expressed in dorsal CNS cells at a position typical of glial cells (Wolfgang et al., 1991), S. G., unpublished data). This, as well as the interaction data presented, suggests that loco function is required to regulate Gαi signalling in glial cells.

Taken together, our data argue for an important role of G-protein-mediated signalling in terminal glial cell differentiation. G-protein signalling is thought to be triggered by binding of a ligand to a seven transmembrane domain receptor. To date, no such receptor has been reported to be expressed in the Drosophila glial cells. Recently cross talk between receptor-tyrosine-kinases and G-proteins has been described (Weiss et al., 1997). Interestingly, the CNS expression of heartless, the Drosophila FGF-receptor2 gene, is restricted to glial cells (Beiman et al., 1996; Gisselbrecht et al., 1996; Shishido et al., 1997). heartless mutant embryos show a defect in lateral glial development. Based on immunostaining using anti-HEARTLESS antibodies, mutant glial cells appear rounded and are incapable of increasing their surface area (Shishido et al., 1997). This is reminiscent of the phenotype of mutant loco embryos described here. It is thus tempting to speculate that HEARTLESS and G-protein signalling involving LOCO act in concert to trigger glial cell shape changes in response to extracellular signals.

We are indebted to A. Klaes for her help with the in situ hybridisation analysis, to Ingrid Bunse and Marie-Laure Parmentier for help with the sequencing, and to Marika Charalambous for help with two-hybrid screening. We thank the Brent laboratory for two-hybrid advice and reagents, and Mike Forte for the Gi clone. This work was supported through grants by the HFSPO and the EC to C. K. and by Grant ICS00761 from the Biotechnology and Biological Sciences Research Council to C. J. O’K.