Comparative sequence analysis and tissue localization of members of the SLC6 family of transporters in adult Drosophila melanogaster

The SLC6 family of transporters, also referred to as the neurotransmitter:sodium symporter family (NSS) or sodium neurotransmitter transporter family (SNF), is composed of integral membrane transporter proteins. These transporters are critical for maintaining physiological homeostasis in animals by transferring compounds such as neurotransmitters,amino acids or osmolytes across the plasma membrane. SLC6 members transport compounds against their concentration gradient by coupling substrate uptake to the energy built up in the transmembrane Na⁺ gradient. Some of these transporters also depend on the presence of extracellular Cl^-(reviewed in Amara and Arriza,1993; Nelson,1998). Substrates for members of the SLC6 family include:neurotransmitters such as GABA, serotonin, norepinephrine and dopamine; the amino acids glycine and proline, which in mammals are also used as neurotransmitters; and the osmolytes taurine and betaine. Recently, the broad substrate profiles B^O+, Bo and IMINO systems have been attributed to SLC6 transporters (Sloan and Mager,1999; Broer et al.,2004; Takanaga et al.,2005). The SLC6 family also includes a group of transporters for which no substrates have been identified, known as the `orphan'neurotransmitter transporters (Amara and Arriza, 1993; Nelson,1998).

The SLC6 gene family is defined by conserved structural features, including a predicted twelve-transmembrane domain topology, highly conserved amino acid residues, particularly in transmembrane (TM) domains 1, 2 and 4-8, and a large extracellular loop between transmembrane domains 3 and 4 that is predicted to be glycosylated (Amara and Arriza,1993; Nelson,1998). Crystal structural data have shown that many of the conserved residues in TM1, TM6 and TM8 are involved in substrate and sodium binding in a bacterial leucine transporter(Yamashita et al., 2005). Mice null for individual SLC6 transporters show abnormal physiological phenotypes. For instance, disruption of the dopamine transporter (DAT) gene leads to persistence of dopamine in the extracellular space and a hyperlocomotive phenotype equivalent to the effects of cocaine and amphetamines, which are known to inhibit the DAT (Giros et al.,1996). When other SLC6 transporters lack function, it can result in death or disease (Gomeza et al.,2003; Heller-Stilb et al.,2002; Quan et al.,2004; Tsai et al.,2004).

Despite their importance in mammalian physiology, there is little information about these proteins in Drosophila. Thus far, only four genes from the SLC6 family have been cloned from Drosophila: the genes responsible for the selective uptake of serotonin (SerT)(Corey et al., 1994; Demchyshyn et al., 1994) and dopamine (DAT) (Porzgen et al.,2001) and two orphan transporters inebriated(ine) (Burg et al.,1996; Soehnge et al.,1996) and bloated tubules (blot)(Johnson et al., 1999). The cellular pattern of expression of SerT RNA in the embryo is similar to that of serotonin immunolabelling(Demchyshyn et al., 1994);likewise DAT is expressed in a cellular pattern in the larva similar to that of dopamine immunolabelling(Porzgen et al., 2001). The other two sequences, Ine and Blot, do not yet have identified substrates but both are expressed in a variety of cells inside and outside the central nervous system (CNS) (Burg et al.,1996; Huang et al.,2002; Johnson et al.,1999; Soehnge et al.,1996).

To begin to address the role of SLC6 transporters in Drosophila,we used a bioinformatics approach to identify 21 Drosophila genes with similarity to known SLC6 transporters. Phylogenetic analysis revealed that these Drosophila SLC6 transporters segregated into four of the five previously recognized SLC6 subfamilies(Nelson, 1998) and helped to define a sixth subfamily, the insect amino acid transporters (IAAT)(Soragna et al., 2004; Boudko et al., 2005). Using in situ hybridization, we focused on localizing SLC6 transporters in the CNS of adult flies. Probes that did not label CNS cells were tested on a variety of other fly tissues as positive confirmation of the reagents. In the course of this work we also localized a vesicular monoamine transporter and,curiously, found that it labelled glial cells at the distal margin of the lamina cell body layer in addition to the expected localization in neurons in the brain in the monoaminergic pattern.

Drosophila yw⁶⁷ or w¹¹¹⁸ mutants,both of which lack eye pigment, were used in all in situhybridizations.

To identify candidate Drosophila genes that could encode Na⁺/Cl^--dependent transporters, we initially performed BLASTP and TBLASTN (Altschul et al.,1997) searches of predicted proteins from the annotated Drosophila genome sequence (Adams et al., 2000). Well known Na⁺/Cl^--dependent monoamine transporters from several organisms, including SerT, DAT and the norepinephrine transporter, were used in sequence searches. A comprehensive group of known or predicted amino acid transporters from vertebrates, insects and worm were assembled, and the full peptide sequences were evaluated by multiple sequence alignment using CLUSTALX(Thompson et al., 1997). The alignment of 84 SLC6 family transporters was used to create neighbor-joining phylogenetic trees in order to depict relationships between various members of the SLC6 family across multiple species. One thousand bootstrap trials were performed to evaluate the significance of the branch node patterns. We used a bootstrap value of >75% to define subfamilies. A separate alignment of the 21 putative SLC6 from Drosophila and the leucine transporter from Aquifex aeolicus [for which the crystal structure was recently published (Yamashita et al.,2005)] was generated to highlight conserved regions in the transmembrane domains. The final alignments depicted herein were manually adjusted and shaded using GeneDoc software (Karl B. Nicholas and Hugh B. Nicholas, 1997).

Total RNA was isolated from Drosophila heads using the TRIzol reagent (Gibco, Carlsbad, CA, USA) and converted to cDNA using a poly-T primer and Superscript II reverse transcriptase (Gibco, Carlsbad, CA, USA). Portions of the cDNA for each candidate gene were amplified by polymerase chain reaction (PCR) and cloned into pGEM-T Easy (Promega, Madison, WI, USA). All clones were verified by sequencing (UNC sequencing facility). Digoxigenin(DIG)-labelled sense and anti-sense riboprobes were generated from linearized plasmids according to manufacturer's specifications using the DIG RNA labelling kit (Roche, Indianapolis, IN, USA). RNA probes averaging 1000 bp and ranging from 800-1200 bp were purified using three precipitations in 3×ethanol and 0.3 mol l^-1 LiCl incubated at -80 for 2 h between each precipitation and resuspended in DEPC-treated water or hybridization buffer.

Drosophila were anesthetized with CO₂. The back third of the abdomen was cut off and the proboscis was removed or, in the case of abdominal sections, the front half of the fly was cut off to allow complete penetration of fixative. The flies were submerged in ice-cold 4%paraformaldehyde (PFA) in phosphate-buffered saline (PBS), transferred to 10%,then to 20% sucrose in PBS, and incubated at 4°C overnight in each solution. Flies were then suspended in Optimal cutting temperature(Tissue-Tek, Torrance, CA, USA) reagent, frozen in isopentane cooled by liquid nitrogen, and sectioned on a cryostat microtome at 10-15 μm. Sections were collected on room temperature Superfrost Plus slides and stored at 4°C until in situ hybridization or antibody labelling could be performed.

The DIG method (Roche, Indianapolis, IN, USA) of in situhybridization was used following the procedure of Nowicki and Burke with minor modifications (Nowicki and Burke,2000). Briefly, slides containing sections of Drosophilaheads were washed and post-fixed in 4% PFA in PBS for 30 min then rinsed twice for 5 min in PBS. After two 2 min washes in 2× sodium chloride/sodium citrate pH 4.5 (SSC), sections were incubated for 30 min in Tris-glycine buffer. Riboprobes, suspended in hybridization buffer at 25-100 ngμl^-1, were applied to the slides and incubated overnight at 65°C in a humidified chamber.

Sections were rinsed 3 times for 20 min each in 5× SSC at room temperature. Sections were transferred to a preheated solution of 20%formamide and 0.5× SSC (sol B) and incubated at 60°C for 40 min. Sol B was replaced with preheated sol B and the temperature of the solution was allowed to cool to 37°C. Sol B was again replaced with pre-heated sol B and the slides were incubated at 60°C for 30 min. Slides were transferred to 2× SSC at room temperature for 30 min. They were then incubated in 2%Boerhenger Blocking Reagent in a maleic acid buffer (block) for a minimum of 10 min. Sections were then incubated overnight in a humidified chamber at 4°C in anti-DIG antibody diluted 1:5000 in block.

Antibody was rinsed away with four washes of 10 min and one wash of 20 min in Tris-buffered saline (TBS) at RT. Sections were then incubated for 10 min in 100 mmol l^-1 Tris with 500 μg ml^-1 levamisole and 0.1% Triton X-100. Sections were placed in 0.131 mg ml^-15-bromo-4-chloro-3′-indolyphosphate p-toluidine (BCIP) and 0.5 mg ml^-1 nitro-blue tetrazolium (NBT) in 10% polyvinyl alcohol for 1 h to 3 days. Sections were then rinsed in PBS and mounted with Glycergel (DAKO,Carpenteria, CA, USA). Sequential sections were probed by alternating between sense and anti-sense probes. Sense probes showed no specific label in the head but revealed non-specific labelling of the exterior margin of the eye.

Total RNA was isolated using the Qiagen RNeasy kit (Qiagen, Valencia, CA,USA) using the manufacturer's specifications. w¹¹¹⁸ flies were dissected into head, thorax and abdomen and immediately frozen on dry ice. Tissue was manually homogenized and further lysed using the QiaShredder spin column. Total RNA was treated with DNase using the DNA-free kit (Ambion,Austin, TX, USA). 2.5 μg of total RNA was used for reverse transcription reaction. PCRs were run for 30 cycles at optimal temperature for the respective primer pairs. Exact primer sequences can be provided upon request.

Flies were fixed as above and sectioned between 10-15 μm. Sections were permeabilized in PBS/0.5% Triton for 10 min and then washed in PBS. Endogenous fluorescence was quenched in 0.5% NaBH₄ in PBS for 10 min and the slides were washed in PBS. Sections were blocked in 5% normal goat serum (NGS)for 1 h, incubated overnight at 4°C in primary antibody in 5% NGS, rinsed in PBS, and incubated in an antigoat secondary antibody Alexa 488 (Molecular probes, Carlesbad, CA, USA) at a dilution of 1:2000 in 5% NGS for 1 h. and rinsed again in PBS and mounted in Gel/Mount (Biomeda, Forest City, CA, USA). Repo antibody (Developmental studies hybridoma bank, Ames, IA, USA) was used at 1:1 and Neurexin IV (NrxIV, generous gift of M. Bhat) was used at a dilution of 1:500. Sections labelled with fluorescent markers were visualized on a Nikon Eclipse E800 microscope with a 20× Plan Apo objective with a numerical aperture (NA) of 0.75 or a 40× Plan Apo objective with an NA of 0.95 and imaged with a Hammamatsu ORCA-ER camera (Bridgewater, NJ, USA) and visualized with the software package Metamorph (Universal Imaging,Downingtown, PA, USA). Images were optimized for visualization and publication using Adobe Photoshop (Adobe, San Jose, CA, USA).

We attempted to identify all SLC6 genes in the Drosophila genome using a bioinformatics approach with amino acid sequences for known SLC6 transporters as bait with which to search GenBank using the NCBI BLASTP and TBLASTN algorithms (Altschul et al., 1997). We identified a total of 21 genes that we consider to be putative members of the SLC6 family in Drosophila (listed in Fig. 1B and bolded and underlined in Fig. 2). All candidate sequences are predicted to have between ten and twelve TM domains (data not shown) and at least 20% identity to SerT or DAT. A multiple sequence alignment including each of the putative SLC6 family members in Drosophila with the amino acid sequence of a SNF6 family member from Aquifex aeolicus, LeuT_Aa, for which the crystal structure was recently published (Yamashita et al.,2005), reveals that each of the candidate sequences contains multiple residues that are absolutely or highly conserved among all of the Drosophila sequences (Fig. 1B). A complete alignment of all transporter sequences used in our phylogenetic analysis is available (supplementary material, Fig. S1).

The highly conserved region in the vicinity of TM domains 1 and 2(Lill and Nelson, 1998) was present in the majority of the candidate genes. For clarity in describing proteins occurring in multiple species, we will use a species prefix such as Dm for Drosophila melanogaster. Five of the candidate proteins(DmCG13793, DmCG13794, DmCG13795, DmCG13796 and DmCG33296) are more divergent at the N terminus, with DmCG13793 and DmGG33296 lacking TM1-TM2. The other three divergent candidate proteins have limited homology in the region of TM1 but have a significant degree of similarity to SLC6 members throughout the remainder of their sequences (this is also the case for several of the more divergent C. elegans SNF genes). Based on the crystal structure of LeuT_Aa, the first transmembrane domain is likely to play a pivotal role in substrate binding (Yamashita et al., 2005). Therefore, these highly divergent sequences may lack functional transporter activity or have alternative roles. One possible explanation for the differences in the N-terminal region of the proteins is that most of the candidate genes are predicted in silico from genomic sequence and therefore are subject to the limitations of exon scanning algorithms. However, cDNA sequences for three of these candidate genes(DmCG13794, DmCG13795 and DmCG13796) have been deposited in the GenBank database and we have therefore included them in this analysis as tentative SLC6 family members.

Structural features identified in the candidate proteins may provide clues as to their function. For example, three candidates (Blot, Ine and DmCG8291)are predicted to have long intracellular N-terminal domains and three candidates (Blot, DmCG5549 and DmCG8291) are predicted to have large intracellular C-terminal domains. DmCG5549 has a short extension of the intracellular linker sequence between TM 6 and TM 7, while the orphan neurotransmitter subfamily members (including DmCG5226 and DmCG10804) are predicted to have extended extracellular linker 4a sequences between TM 7 and TM 8 as well as a short extension of the extracellular linker sequence between TM 11 and TM 12 (see Fig. 1A). Long intracellular terminal domains or loops may indicate intracellular regulation by protein-protein interactions or phosphorylation(Melikian, 2004) while extracellular loops are likely to be involved in substrate binding, the transport process and inhibitor binding(Zomot and Kanner, 2003).

In order to infer functional information about the putative Drosophila SLC6 family members, we compared them to a large number of known SLC6 sequences using phylogenetic analysis(Fig. 2). We aligned the 21 Drosophila sequences with an assembled collection of 63 putative neurotransmitter or amino acid transporters from other organisms: 25 sequences from other insects, 19 from vertebrates, 13 from C. elegans and 5 from prokaryotes. Fig. 2displays a phylogenetic tree derived from the complete alignment of 84 SLC6 family members. Accession numbers for the sequences used are provided(supplementary material, Table S1) and the complete alignment is available(supplementary material, Fig. S1).

Some of the previously uncharacterized Drosophila sequences segregate into previously recognized subfamilies anchored by well-characterized transporters (Nelson,1998). For example, DmCG1732 and its Anopheles homolog AgCP8499 are likely to function as GABA transporters as they are closely related to human GAT-1, C. elegans SNF11 [a recently identified GABA transporter (Mullen et al.,2006)], and the GABA transporters from Trichoplusia ni(Gao et al., 1999) and Manduca sexta (Mbungu et al.,1995). DmCG5549 and its Anopheles homolog AgCP10735 are related to the amino acid transporter subfamily, which includes the amino acid transporters for glycine and proline and a transporter with a broader substrate capacity, B^O+ (Sloan and Mager, 1999). DmCG7075 and its Anopheles homolog,AgEbiG6595, are closely related to a proline transporter from Manduca(Sandhu et al., 2002), and these transporters segregate loosely with the glycine/proline subgroup and the monoamine transporters. Surprisingly, only the Drosophila DAT and SerT are present in the monoamine transporter subfamily. Since octopamine and histamine are also biogenic amines and the Trichplusia ni octopamine transporter is present in the monoamine subfamily, we expected that Drosophila candidate genes with these transport activities would be found in this subfamily. However, it appears that there are no close monoamine homologs in this subfamily.

Two candidate genes, DmCG5226 and DmCG10804 (and their Anopheleshomologs AgCP7001 and AgCP7501, respectively) group with the previously defined subfamily of orphan neurotransmitter transporters that includes NTT4 and NTT7-3 as well as two other newly predicted mammalian orphan transporters(Nash et al., 1998). Interestingly, this family also includes a neutral amino acid transporter(B°AT), mutations of which cause Hartnup disease(Broer et al., 2004; Kleta et al., 2004; Seow et al., 2004).

Our phylogenetic analysis identified a number of predicted genes that help define the novel insect amino acid transporter (IAAT) subfamily(Boudko et al., 2005), which to date is represented only in insects. Predicted sequences from Drosophila melanogaster (DmCG1698, DmCG3252, DmCG4476, DmCG8850,DmCG15088 and DmCG15279) and Anopheles gambie (AgCP10401, AgCP10412,AgCP10503 and AgCP10610) are clearly related to the Manduca sextatransporters KAAT1 (Castagna et al.,1998) and CAATCH1 (Feldman et al., 2000), which are competent to transport substrate amino acids using either K⁺ or Na⁺ as the driving ion. Sequence similarities among KAAT1, CAATCH1 and the other members of the IAAT subfamily suggest that many of the members of this subfamily possess this unique functional property.

A large number of predicted Drosophila and Anopheles SLC6 transporters segregate into poorly defined subgroups. For example, DmIne and AgCP4566 do not segregate nicely into a recognized subfamily. DmIne mediates a response to hypertonic solutions (Chiu et al., 2000) and has been proposed to comprise its own subfamily(Boudko et al., 2005). Other sequences, including a number of orphan transporters from C. elegans,vertebrate NTT-5 and putative Drosophila transporters (DmBlot,DmCG8291, DmCG13793, DmCG13794, DmCG13795 and DmCG33296) make up a loose assemblage of outliers, which do not appear to be affiliated with defined subfamilies. We have identified Anopheles homologs for the Drosophila Ine and Blot orphan transporters (AgCP4566 and AgCP2093,respectively) as well as for the other outlier sequences, suggesting a conserved, if still unknown, function for these transporters.

We used RT-PCR to determine a relative segmental expression profile for the SLC6 homologs in adult flies. mRNA was isolated from the head, thorax and abdomen of either males or females and amplified by RT-PCR to determine the presence or absence of transcript in each segment(Fig. 3). As a control for consistent starting cDNA template across all segments we used primers for the gene rp49, and as a control for a gene with enriched expression in the head and thorax we used primers for the gene for the histamine synthesizing enzyme, histidine decarboxylase (HDC)(Burg et al., 1993; Pollack and Hofbauer,1991).

Six candidates were not expressed evenly across all body segments(Fig. 3). Four candidates, SerT, DAT, CG1732 and CG10804, showed enriched expression in the head and thorax of both males and females. The cell bodies of the CNS are found in the head and thorax, and cells that immunolabel for serotonin(Valles and White, 1988),dopamine (Budnik and White,1988; Nassel and Elekes,1992) and GABA (Buchner et al., 1988) reside within the CNS. Expression of the orphan transporter, CG10804, was only found in the head and thorax,suggesting that CG10804 functions in the CNS. Two of the candidates, CG7075 and CG4476, were expressed more abundantly in the abdomen of the male and female fly, respectively.

The remaining fifteen SLC6 transporters showed approximately equivalent expression across the three body segments in both genders. This result suggests that these transcripts are expressed either in organs in each of the segments or in an organ present in all three segments. Although twelve of these transcripts displayed robust expression, three transcripts (CG8850,CG13793 and CG33296) revealed only weak expression.

Previous immunocytochemistry has revealed the location of serotonergic and dopaminergic cells in the adult fly CNS; these neurons are grouped in named clusters (Budnik and White,1988; Valles and White,1988). The riboprobes against SerT(Fig. 4A,B) and DAT(Fig. 4C,D) labelled CNS cells of roughly the same number and in the same general location as the immunolabelled cells. Due to limited resolution of the in situhybridization technique and use of alternating sections between sense and anti-sense probes, we were only able to estimate the number of neurons that were labelled by each riboprobe. For the SerT probe, about 75 neurons labelled in cell clusters that approximated the position of the named clusters from previous serotonin immunolocalization studies in both the head and thoracic ganglion. Labelled cells in the head(Fig. 4A) were found dorsally on either side of the midline in the position of the cluster called SP1(arrows), ventrally in the position of the SE2 cluster (barbed arrow), and proximal to the optic lobe in the position of the LP2 cluster (notched arrow). The photoreceptor cells were never labelled. Cells in the thorax(Fig. 4B) are equivalent in position to Budnick and White's `A1-7' neurons (arrows)(Budnik and White, 1988). Other sections showed neurons in positions of other named clusters and no cells were labelled in positions that substantially differed from the immunolabel. It is likely, therefore, that our SerT riboprobe is labelling serotonergic neurons.

The DAT riboprobe likewise labelled discrete cells in the pattern described for dopaminergic neurons using immunocytochemistry(Budnik and White, 1988). We identified up to 30 large cell bodies in a single preparation with alternating slides, approximately the number of large cell bodies labelled by the dopamine antibody. In the section shown in Fig. 3C, the riboprobe labelled cells in the positions of three named clusters: two dorsal clusters on either side of the midline corresponding to the DM cluster of neurons (arrows); neurons lateral to the dorsomedial (DM)cluster in the more dorsal DL1 cluster (white arrowheads); and cells more ventral in the DL2 cluster (carrot). We also detected DAT positive cells corresponding to the AbU (arrows) and ThL clusters (barbed arrow) in the thoracic ganglion (Fig. 4D). In other sections, we detected cells in other named clusters. Curiously, the riboprobe did not label a group of cells surrounding the medulla, the MC cell group that clearly labels for catecholamines[fig. 7A in Budnik and White(Budnik and White, 1988)], nor did it label a cluster of small cells along the midline referred to as the anteromedial (AM) neurons. Thus, as with the SerT riboprobe, the pattern of in situ label overlapped with that from dopamine immunohistochemistry, suggesting that, with the exception of the AM and MC cells, the DAT probe labels dopaminergic cells. It does not appear that the DAT probe labels octopaminergic in addition to dopaminergic cells, as we do not see DAT positive cells in the easily recognizable antennal lobe (AL) cluster of octopaminergic cells localized along the alimentary canal (Monastirioti et al.,1995). In addition, the DAT probe does not label octopaminergic cells in the larva and does not have a high affinity for octopamine in uptake experiments in Xenopus oocytes(Porzgen et al., 2001).

Although we did not find a Drosophila SLC6 homolog for either the plasmalemmal octopamine or histamine transporter, in the course of this study we localized an ortholog of a mammalian vesicular monoamine transporter 2(VMAT2). In mammals, VMAT2 packages four biogenic amines (dopamine, serotonin,norepinephrine and histamine), and it is present in neurons using each of these four neurotransmitters (Peter et al., 1995). A Drosophila ortholog of the mammalian VMAT2 would be expected to label all monoaminergic cells, revealing cells that contain dopamine, serotonin, octopamine and histamine. The closest homologous sequence to VMAT2 in Drosophila is CG33528, which is 56% identical (Greer et al.,2005).

A riboprobe designed against CG33528 labelled on the order of 125 cells in the central brain (Fig. 4E,G) and thorax (Fig. 4F), which are in the anatomical position of named clusters of monoaminergic neurons. The riboprobes for SerT and DAT each labelled cells in the same location as the CG33528 label (compare Fig. 4A,C with 4E), but there were consistently more cells labelled by the CG33528 probe than for the SerT and DAT probes combined. This pattern may represent labelling of octopaminergic neurons since antibody labelling revealed that CG33528 was found in serotonergic, dopaminergic and octopaminergic(Greer et al., 2005) neurons,but was not found in the histaminergic photoreceptors(Chang et al., 2006). We were able to detect CG33528 label in both presumptive MC cells (open arrows in Fig. 4E),dopaminergic neurons present in the medullary cortex, and in the presumptive AL cells, octopaminergic cells that reside just lateral to the oesophagus(data not shown).

Curiously, we were unable to detect label in photoreceptors, which in the fly are histaminergic (Pollack and Hofbauer, 1991). To confirm that our in situ approach would work in the photoreceptors, we made a riboprobe to HDC, which is expressed in photoreceptors (Burg et al., 1993). The HDC riboprobe labelled the photoreceptor layer of adult head sections and in addition, about 20 discrete, bilateral cells in the central brain (Fig. 4J) as well as cells in the thoracic ganglion (data not shown) all in the anatomical positions consistent with immunolabelled histaminergic neurons (Pollack and Hofbauer,1991).

Unexpectedly, the riboprobe for CG33528 also labelled cells in a punctuated arc at the fenestrated layer at the base of the photoreceptors(Fig. 4G-I), a distinct region where the photoreceptor cell bodies form the axons that ultimately synapse in the optic lobes. This band of beaded label is less than 5 μm in width, the centers of the beads spaced 8-10 μm apart, and the label is intermixed with the fenestrations at the base of the photoreceptors that are present at the distal-most edge of the lamina cortex. Fingers of label occasionally penetrate into the photoreceptor layer (block arrows in Fig. 4H).

The cell types of this layer have been described for Drosophila melanogaster (Eule et al.,1995) and ultrastructurally in Musca domestica(Saint Marie and Carlson,1983). There are two glial subtypes that reside at this distal margin of the lamina cortex, the fenestrated and pseudocartridge glia. In both species, these cells have their somata spaced approximately 10 μm apart; in Musca, both glial subtypes have been shown to wrap photoreceptor axons traversing this layer on their way to forming synapses in the optic lobes (Saint Marie and Carlson,1983). The fenestrated glia, but not the pseudocartridge glia,have processes that invade the photoreceptor layers, similar to the CG33528 labelling. When alternating sections were labelled with CG33528 riboprobe (Fig. 4H) and an antibody to Neurexin IV (NrxIV; Fig. 4I), a component of septate junctions found on the plasma membrane of glia and not neurons(Baumgartner et al., 1996), the labelling pattern was similar. Like CG33528, the NrxIV antibody labelled the base of the photoreceptors in a beaded fashion spaced approximately 10 μm apart (Fig. 3I). This labelling pattern suggests that CG33528 is present in the distal-most glia of the lamina cell body layer, in particular the spacing and the fingers penetrating into the retina suggest that that these are the fenestrated glia.

CG1732 (the putative GABA transporter) and CG15088 (a member of the IAAT subfamily) are expressed in the CNS in a glial-like pattern(Fig. 5). Glia of the fly CNS form the barrier between hemolymph and the brain, wrap axons and isolate synaptic terminals in areas such as the lamina(Saint Marie and Carlson,1983), among other roles. We have chosen to use the Drosophila glial terminology(Eule et al., 1995). In short,a layer of glia, called perineural glia(Fig. 5B, arrows), forms the outermost layer in the adult brain. Interior to the perineural glia, the somata of the subperineural glia (Fig. 4B, carrots) are interspersed with neuronal somata to form the cortex. Neuronal processes form the neuropil and neuropilar glia wrap the neuronal processes within the neuropil. As is true in general of invertebrate nervous systems, there are no neuronal somata within the neuropil.

The riboprobe CG1732 labels a subset of regularly spaced cells that surround neuropil structures in the central brain(Fig. 5A,C) and ventral ganglion. Neuropil, including the subesophageal ganglion, the medulla and lobula (Fig. 5A), the antennal lobes and ventral nerve cord (data not shown), are surrounded by somata postive for CG1732. Between the medulla and lobula, CG1732labels a subset of giant glia at the inner chiasm (barbed arrow in Fig. 4A), a region that contains only glial and not neuronal cell bodies(Tix et al., 1997). This pattern is substantively different than that of GABA immunolabelled cells,which have cell bodies throughout the cortex surrounding the medulla and in a large cluster next to the lobula (Buchner et al., 1988).

CG1732 label (Fig. 5A,C) was compared in sequential sections to immunolabel for the glial nuclear marker, Repo (Fig. 5B,D) (Xiong et al.,1994). Repo labelled nuclei in positions of the perineural glia(arrows, Fig. 5B), the subperineural glia (carrots, Fig. 5B) and in regularly spaced intervals that form the outlines of several neuropilar structures. CG1732 labelled only cells located between the cortex and neuropil (compare label surrounding the medulla and lobula in Fig. 5A with Fig. 5B). CG1732 did not label perineural (arrows, Fig. 5A) or subperineural glia (carrots, Fig. 5A), which were labelled by Repo. Thus, the riboprobe for CG1732 only labels the subset of glia that surround neuropil.

The label for CG1732 had two additional, notable characteristics. First, there was a purple hue within the neuropil and not present throughout the cortex, which suggests that the mRNA for CG1732 is possibly present within glial processes. Second, the purple hue is particularly dense in the distal onethird of the medulla (Fig. 5A,C), where the photoreceptors R7 and R8 make their first synapse. The dense label shows a banded pattern that runs parallel with the distal margin of the medulla and is visible in the left medulla(Fig. 5A). Repo-positive and CG1732-positive cells (notched open arrows in Fig. 5C,D) are located in similar positions and with similar spacing surrounding the neuropil. It appears that labelled processes emanate from regularly spaced cell bodies at the distal border of the medulla (Fig. 5C) and penetrate into the neuropil. The dense CG1732-labelling pattern ends abruptly at a consistent layer in the neuropil. Medulla neuropil glia are morphologically similar and localized similarly to these CG1732 labelled cells[fig. 3A-C in Richardt et al.(Richardt et al., 2002)] and also send processes into the distal portion of the medulla, where the CG1732 label is heaviest. CG1732-positive cells were not found in the cortex that surrounded the lamina or medulla.

Since CG1732 is likely to be a GABA transporter and GABA uptake has been described in the lamina of the optic lobe(Campos-Ortega, 1974), we looked more closely at this region of the CNS. The optic lobes at higher magnification (boxed area of Fig. 5A,C) show that the lamina neuropil is lightly labelled with heavier label over a cluster of large cell bodies at the proximal border of the neuropil (arrowheads), where the outer chiasmatic and marginal glia both reside (Eule et al., 1995). The lamina neuropil has a weak purple tint that ends abruptly at the distal margin suggesting that cells that have processes that penetrate the lamina express this transporter. Epithelial glia would be the obvious candidate, as their processes surround the photoreceptor terminals in the laminar, and indeed in overdeveloped preparations, the cell bodies of the epithelial glia at the distal margin of the lamina appear to have faint label (data not shown). Nevertheless we were unable to definitively determine if the epithelial glia express CG1732.

The Repo antibody labels cells that both surround the neuropil and are intermixed with neuronal somata in the cellular cortex of both the lamina and the medulla. Prominent among the Repo-labelled nuclei are cells at the proximal border of the lamina (arrowheads, Fig. 5D) that are in the same position as CG1732-positive cells (arrowheads, Fig. 5C). Outside the neuropil borders, there exist Repo-labelled glia at the base of the photoreceptors(Fig. 5D, arrows) that are not labelled by CG1732, again indicating that that CG1732 labels only a subset of glia.

CG15088 labels a different set of glia from CG1732(Fig. 5E). This putative SLC6 transporter is found within the IAAT subfamily rather than in a neurotransmitter transporter subfamily. CG15088 is present in both a punctate and diffuse pattern throughout the cortex; the labelled cells line and define the neuropils at the inner border of the cortex and also are found throughout the cortex (Fig. 5E). This pattern is consistent with both that of subperineural and perineural glia. In contrast to CG1732, which labelled structures within the neuropil but not the cortex, the riboprobe for CG15088labels the cortex and does not label neuropil. CG15088 label has a similar pattern to that of Repo-positive cells(Fig. 5F). Label from both probes surround neuropil structures such as the antennal lobes, ventral bodies and the ellipsoid body shown in Fig. 5E,F (asterisks), and both reagents label cells throughout the cortex. Also, we have detected CG15088 label in the inner chiasm giant glia where only glial cell bodies are found (data not shown). These results suggest that CG15088 may be expressed in all glia of the Drosophila cortex where it may play a role in transporting nutrient amino acids within the CNS.

CG5549, a candidate in the amino acid subfamily, and CG5226 and CG10804, both segregating with the orphan neurotransmitter transporter subfamily, are expressed broadly throughout the cellular cortex in the head (Fig. 6A,C,E) and the thorax (Fig. 6B,D,F). The cortex comprises primarily neurons with glial cell bodies intermixed, but dense label throughout the head prevented us from distinguishing whether both types of cells were labelled. In the region of the inner optic chiasm between the medulla and lobula neuropil, only glial cell bodies reside (Tix et al.,1997), and we used this region to determine if at least this subset of glia was labelled by the probes. The riboprobes for CG5549and CG10804 (but not CG5226) labelled large cell bodies at the inner optic chiasm (Fig. 6E, filled arrow). Of these three candidates, only CG5549labelled the photoreceptors. CG5549 also labelled a cell layer in the cardia (asterisk in Fig. 6D),an organ at the transition between the foregut and midgut. The cardia is responsible for the creation of the peritrophic membrane, a four-layered extracellular barrier between the contents of the gut and the gut epithelium(King, 1988).

SLC6 members have been shown to transport compounds other than neurotransmitters, including amino acids and osmolytes, such as betaine and taurine. For Drosophila SLC6 transporters that had no detectable CNS expression, we extended our analysis to verify that the riboprobe would bind to transcripts expressed in other organs. We determined if any of the candidate genes were expressed in other recognizable organs such as the alimentary canal, Malpighian tubules and/or reproductive organs.

Two candidates from the IAAT subfamily (CG3252 and CG15279) were found in the cardia of the gut(Fig. 7A,B). The Drosophila cardia has been sub-divided longitudinally into 6 zones based on the morphological and intracellular characteristics of the constituent cells (King,1988). CG3252 is present in the long columnar cells that make up zone 5 of the cardia (Fig. 7A) and the label persists through zone 6 and into the posterior ventricular cells of the midgut (data not shown) where nutrient uptake is thought to occur. CG15279 (Fig. 7B) is expressed in cells near the junction of the cardia and the ventricular cells but these cells are much shorter than the zone 5 cells. These cells are likely to be the cells of zone 6. CG15279 label does not persist beyond the transition of the cardia to the intestine.

Three other candidates are also detected in cells along the alimentary canal. Two candidates from the IAAT subfamily, CG8850 and CG1698, are present in the digestive portion of the midgut. CG8850 (Fig. 7C) is found in cells that line the gut lumen in a restricted portion; adjacent cells do not express CG8850. Midgut epithelial cells are responsible for the uptake of nutrients from consumed food(Miller, 1950). The CG1698 (Fig. 7D)riboprobe labels a broader domain of the midgut that stretches from the thoracic into abdominal segments. CG8291, a candidate that does not segregate into a distinct subfamily, is expressed in the rectal bulb of the alimentary canal (Fig. 7G,arrows).

Fig. 7E-J shows tissues labelled by four transcripts found in the reproductive organs. CG7075is found exclusively in the testis of the male fly(Fig. 7E). The testes are recognizable by the elongated, striated appearance of the developing sperm within the organ (Miller,1950). The riboprobes for CG1698(Fig. 7F,I) and CG8291(Fig. 7G,J) both label the reproductive organs in both genders. CG1698 labels the testis in the male (Fig. 7F) and a subset of cells in the female reproductive tract and potentially in the epidermis of the calyx, the region of the oviduct that meets the ovary(Fig. 7I). CG8291 was detected in the testis (Fig. 7G, open arrows) and the nurse cells and oocyte in the female reproductive organ (Fig. 7J). Nurse cells furnish the developing oocyte with nutrients and maternally derived mRNAs (Mahajan-Miklos and Cooley,1994). CG4476 is expressed exclusively in the nurse cells and oocytes of the female reproductive system(Fig. 7H).

Only one candidate, CG15279, was found in the Malpighian tubules(Fig. 7K), the organ that carries out kidney-like function in the fly. The figure shows two closely apposed tubes that fit the morphological description of the cells and anatomical description of the Malpighian tubules(Miller, 1950). This apposition, and the cell morphology of the labelled structures, suggests that these structures are the Malpighian tubules somewhere after the stalk splits from one to two tubes. We were able to detect positive cells with these same characteristics in multiple regions of the abdomen in frontal sections (data not shown). Since the Malpighian tubules comprise a total of four branches that wind through the abdomen, it seems likely that the tubular cross-sections we observe belong to this system.

We detected CG13795, an outlier candidate sequence, in the embryo but not in the adult (Fig. 7I). The probe for CG13795 labels the midgut precursors and the Garland cells at stage 14 and 15; as the gut matures to stage 16 it loses the expression of CG13795, but labelling of the Garland cell cluster,with its characteristic U-shape (Miller,1950), is retained. Garland cells encircle the proventriculus, are known for their high endocytosis activity, and are proposed to carry out a liver-like function in cleaning the hemolymph of the fly(Kosaka and Ikeda, 1983).

In this study, we set out to identify the full complement of Drosophila SLC6 transporters present in the Drosophilagenome and determine their expression pattern in the adult CNS. Our search identified 21 sequences with sequence similarity to the SLC6 family, Table 1 lists these sequences and summarizes the overall findings. A majority of these sequences either segregated into four previously recognized subfamilies or helped to define a novel subfamily, the IAAT subfamily. Using in situ hybridization, we localized seven candidates to the CNS. Of these seven genes, two known genes(SerT and DAT) were expressed in a discrete pattern in the brain and thoracic ganglion predicted by known immunolabelling, two candidates(CG1732 and CG15088) were expressed in a pattern consistent with glia, and three candidates (CG5549, CG5226 and CG10804)were broadly expressed in the CNS. Neither phylogenetic analysis nor the in situ hybridization results suggested a candidate for an octopamine or histamine transporter despite the known presence of neurons containing these amines. Eight candidates labelled cells in the alimentary canal,reproductive organs and Malpighian tubules but were not detected in the CNS. We did not attempt or were unable to determine a pattern for the six remaining candidates for reasons discussed below.

The crystal structure of a bacterial Na⁺-dependent leucine transporter homologous with mammalian SLC6 was recently solved. Of particular interest, the structure reveals residues important for substrate binding,sodium binding and gating of the transporter. One major difference between our multiple sequence alignments and those of Yamashita et al.(Yamashita et al., 2005) is the inclusion of a larger number of sequences (see Fig. 1B and Fig. S1 in supplementary material). There are no absolutely invariant residues identified in our complete alignment of 84 putative transporters, and the only invariant residue in the Drosophila transporter alignment is a glycine residue at the end of TM2. We also find several residues that seem to be more conserved among metazoan transporters than would be suggested by the Yamashita alignment. The densest concentration of amino acid residues with direct contacts with the substrate and sodium were found in TM1, TM6 and TM8 with other critical residues scattered throughout the protein. Many of the known and candidate Drosophila SLC6 sequences contain these precise residues or have conservative amino acid substitutions, although there are clear differences among certain divergent transporter subgroups that may underlie differences in substrate specificity, ion binding and gating (see Fig. 1 and Fig. S1 in supplementary material).

Blot is perhaps one of the more divergent transporter sequences, yet it contains a conserved arginine responsible for gating and a conserved asparagine involved in sodium binding present in TM1, as well as many other conserved residues. In addition, Blot has been localized and the absence of Blot has a phenotype (Johnson et al.,1999), therefore Blot is a functional protein despite the disparity in sequence homology.

For CG33296 and CG13793-CG13796, the situation is not as clearcut, but our data suggests that they are expressed, as we have detected at least low expression by RT-PCR of each of these genes. Thus far, the predicted proteins for CG13793 and CG33296 are missing the critical TM1 and part of TM2. This could be a result of protein prediction errors or a common first exon that is shared between these proteins (see below). Another possibility is that these proteins have adopted a slightly different method of functioning from the ancestral SLC6 transporter. CG13794, CG13795 and CG13796 contain divergent TM1 sequences (though they still retain several highly conserved residues)throughout this and other segments of the alignment. For CG13795, we have demonstrated discrete expression in the Garland cells in late stage embryos,suggesting a specific role in these cells.

Our results confirm and expand a sixth, novel subfamily of SLC6 transporters, referred to as IAAT (Boudko et al., 2005), which we show includes six putative transporters from Drosophila and seven putative transporters from Anopheles. With the exception of minor differences in branching geometry and branching order at nodes with low bootstrap values, the clusters of transporters in the two trees are largely the same(Boudko et al., 2005). The sequence differences between the IAAT and other SLC6 subfamilies may underlie physiological differences in the function of these transporters. There are three cloned members of this subfamily, the Manduca sexta proteins known as potassium-coupled amino acid transporter-1 (KAAT1)(Castagna et al., 1998), the cation-anion-activated amino acid transporter/channel-1 (CAATCH1)(Feldman et al., 2000) and the Aedes aegypti amino acid transporter (AeAAT1)(Boudko et al., 2005). All three are competent to transport amino acids using either K⁺ or Na⁺ as the driving ion, especially at highly negative membrane potentials. In contrast, other SLC6 transporters use only Na⁺ as their driving ion. The molecular basis of this ion specificity may relate to residues in the transporter responsible for coordinating these ions. In this regard, it is intriguing that the IAAT transporters all contain an alanine or serine substitution at the site of the glycine residue in TM1a reported to be involved in coordinating sodium ion Na1(Yamashita et al., 2005). It should also be noted that other sodium coordinating residues or residues whose side chains reportedly interact with sodium ions in TM6a and TM8 are somewhat less conserved across the entire family of transporters.

The physiological difference in ion selectivity is thought to reflect the environment of the insect gut in which these transporters operate. In Manduca, goblet cells secrete high concentrations of potassium into the lumen of the gut; in addition, the columnar cells that take up the amino acids have an unusually high membrane potential across their luminal (apical)membrane (Harvey and Wieczorek,1997). Nutrient uptake into the columnar cells is driven by both the high concentration of K⁺ and the large negative membrane potential. All three of the cloned transporters are associated with the insect midgut cells: KAAT1 was localized to columnar cells, CAATCH1was cloned from a library created from midgut epithelium, and AeAAT1was cloned from a posterior midgut library and was localized to various structures in the gut, including the cardia, posterior midgut and Malpighian tubules (Boudko et al.,2005).

Our in situ data, however, demonstrate that DrosophilaIAAT subfamily members are expressed in more varied tissues than the gut. Confirming this, RT-PCR data suggests that five out of six Drosophilamembers of the IAAT subfamily are expressed in all three segments of the fly,the exception being CG4476, which is only weakly expressed in tissues outside of the female abdomen. In situ hybridization revealed that IAAT members were expressed in such diverse tissues as CNS (CG15088),midgut (CG1698 and CG8850), cardia (CG3252 and CG15279) and the Malpighian tubules (CG15279). As the brain is not expected to have a high potassium environment or unusually high membrane potentials as in Manduca gut, it is possible that the Drosophila transporters from the IAAT subfamily could use either Na⁺ or K⁺ as their driving ion, depending on ion availability in a given environment.

We can compare our IAAT localization results with two independent sources of information: microarray expression studies being carried out on isolated tissues and the embryonic in situ hybridization project being carried out by the Berkeley Drosophila Genome Project (BDGP)(Tomancak et al., 2002). Microarray data from the adult Malpighian tubules demonstrated that only the SLC6 transcript, CG15279, was enriched (30-fold) compared against the rest of the fly (Wang et al.,2004). This agrees with our in situ hybridization results(Fig. 7K).

BDGP has partially completed a systematic determination of embryonic expression for each annotated Drosophila gene using in situhybridization. Patterns for four of the IAAT subfamily members have been released. CG3252 was found in the embryonic hindgut and the Malpighian tubules, whereas we detected it in the cardia of the adult alimentary canal. CG4476 was detected in early embryos, probably due to maternal contribution, and in the embryonic stomatogastric nervous system,germ cells and endocrine system. In the adult, we detected CG4476 in nurse cells and developing oocytes of the female reproductive system, which confirms that mRNA for CG4476 is maternally contributed to the embryo. BDGP reported expression of CG1698 in the embryonic proventriculus, frontal ganglion and stomatogastric nervous system, but we detected CG1698 in the reproductive organs and midgut of the adult. CG8850 was identified in the embryonic Malpighian tubules by BDGP,but we found it expressed in cells lining the adult ventriculus in the midgut. These expression differences between the adult and the embryo are likely due to the specific requirements of the cells at the adult and embryonic stage.

Discrepancies exist between our RT-PCR and in situ localization observations. Indeed such discrepancies might be expected. The areas of detection differed between in situ hybridization and RT-PCR. Our in situ hybridization focused on the CNS and on other tissues in which we could clearly distinguish a signal, while for RT-PCR an entire body segment was used to derive the cDNA. Therefore, the RT-PCR may have detected gene expression in a tissue that we did not examine using in situhybridization. Also, RT-PCR is the more sensitive technique; a transcript expressed at low levels but broadly throughout a tissue would be easier to pick up by RT-PCR.

We have localized SerT and DAT to neurons in the adult CNS that label in a pattern similar to the label for the respective neurotransmitter. Neurotransmitter transporters have been localized to presynaptic neurons, postsynaptic neurons and surrounding glia; SerTin the embryo (Demchyshyn et al.,1994) and DAT in the larva(Porzgen et al., 2001) have been localized to cells in the pattern of cells that immunolabelled for their respective neurotransmitter. Not every cell that labels for dopamine was positive for DAT labelling, indicating that all cells that release dopamine may not re-capture this compound. The localization of monoaminergic neurotransmitter transporters agrees with localization data from other species, which are found in the presynaptic neurons(Hoffman et al., 1998).

There is only a single Drosophila SLC6 transporter(CG5549) that segregated into the amino acid subfamily. CG5549 is expressed in each segment of the fly, as determined by RT-PCR, and is expressed broadly throughout the adult CNS, including the inner chiasm giant glia and the photoreceptors. Cloned members of this subfamily have substrates which include amino acids that double as neurotransmitters,such as proline or glycine (Malandro and Kilberg, 1996), or a broader substrate profile of cationic and neutral amino acids for the transporter B^O+(Sloan and Mager, 1999). The broad expression of CG5549 in both neurons and glia of the CNS suggests a basic role for this candidate, such as nutrient uptake.

The two candidates that segregate with the orphan transporters, CG5226 and CG10804, both displayed a broad localization pattern throughout the CNS. Within the CNS, both CG5226 and CG10804 were detected in neurons. CG10804 expression was also detected in the inner chiasmatic glia but we were unable to detect CG5226 in this subset of glia. RT-PCR revealed expression differences. CG10804 was enriched in the head and thorax, segments containing cell bodies of the CNS, whereas CG5226 was evenly expressed across all segments. These expression differences suggest that these two orphan transporters may be carrying out broad but different functions in the fly, such as importing a necessary compound for cellular function or clearing a compound with a wide extracellular distribution. The embryonic localization by BDGP of CG5226 (CG10804 has not been released) shows that CG5226 is expressed in cells throughout the more mature CNS of embryonic stages 13-16, but not in earlier CNS stages when the CNS is still developing. The similar localization pattern for CG5226in the embryo and adult suggests a similar role in more mature CNS cells. It has been predicted that substrates for members of the orphan neurotransmitter transporter family will include amino acids(Boudko et al., 2005), based on phylogenetic analysis, and our localization data is consistent with this prediction. Interestingly, these transporters possess highly divergent extracellular linker sequences 2, 4a and 6, consistent with a role for these sequences in substrate selectivity.

In Drosophila, uptake of the neurotransmitter GABA has been demonstrated in the lamina (Campos-Ortega,1974), but the gene responsible for this uptake has not been identified. There is compelling evidence that CG1732 is a GABA transporter (GAT) in Drosophila. It is the only candidate that segregated into the GABA subfamily, and it has significant identity with other known GABA transporters: Manduca sexta GAT (80%), Trichplusia ni GAT (81%), human GAT1 (59%) and human GAT3 (53%). By RT-PCR, CG1732 is enriched in the head and thoracic segments but it is also present in the abdomen segment.

Evidence from other species suggests that GABA transporters can be present in both neurons and glia (Borden,1996), but our data demonstrate that CG1732 is expressed in a subset of glia. In Drosophila, the pattern of CG1732label only partially overlaps with the pattern of GABA immunolabelling in the optic lobes (Buchner et al.,1988). The GABA antibody labels approximately 1500 somata in the cortex of the medulla and, in the neuropil, there is faint label in the lamina, and denser label in the medulla. CG1732 label is found in the neuropil but not in cell bodies of the cortex, suggesting that GABAergic neurons are not the primary source of CG1732 expression in the medulla. Instead, the labelling pattern for CG1732 is better represented by the immunolabelling pattern for the glial marker, Repo. In addition, glial expression of a Drosophila GABA transporter is supported by GABA uptake studies in the lamina of housefly and Drosophila, where ³H-GABA was accumulated in the glia rather than neurons of the optic lobes(Campos-Ortega, 1974). These authors described uptake into glia of the distal border of the lamina(epithelial glia) and at the proximal border (marginal glia). We were not able to conclusively show expression of CG1732 in these cell types, but the lamina does show weak labelling in the lamina neuropil. Finally, in the embryo BDGP localized CG1732 expression to a subset of cells in the ventral nerve cord that are in a position consistent with the channel glia. Thus, our evidence, combined with published findings, indicates that expression of CG1732 is in a subset of glia in Drosophila.

In another insect, Manduca sexta(Umesh and Gill, 2002), MasGAT immunoreactivity coincided with GABA immunoreactivity(Homberg et al., 1987) in the neuropil but did not appear to be in coincident locations in the cortex. In adult Manduca, somata immunoreactive for GABA appear throughout the cortex of the optic lobe (Homberg et al.,1987), as is the case for Drosophila. Somata in this same area of cortex are not labelled by the antibody against MasGAT(Umesh and Gill, 2002) but somata near the neuropil are clearly labelled and often send a process into the nearest neuropil. In fact, some of these author's photographs reveal a band of dark staining surrounding the optic lobe neuropil, similar to CG1732 labelling that we detect. These results suggest various possibilities. First, MasGAT is expressed in GABAergic neurons, but the final position of the MasGAT protein is at the terminals and not the cell bodies of GABAergic neurons. Therefore the antibody labelled only the terminals, leaving GABAergic cell bodies unlabelled. This explanation is possible though other somata were clearly labelled by the MasGAT antibody. It is also possible that only a small subset of neurons very near the neuropil express MasGAT. Another possibility is that MasGAT is expressed in glia similar to our conclusion in Drosophila.

In our phylogenetic analysis, five candidates did not fall into any distinct subfamily, not even into the orphan subfamily. We have called these`unclassified sequences' to distinguish these proteins from the previously defined orphan subfamily. We were able to localize these candidates outside the CNS.

Two unclassified candidates (CG7075 and CG8291) are associated with the reproductive tissue. By in situ hybridization, CG7075 is found in the testes of the male, and our RT-PCR results show that CG70705 is expressed almost exclusively in the male abdomen, which supports our in situ localization results. In the adult, CG7075 is likely to be a testesspecific gene because we did not detect it elsewhere in the fly. Also, CG7075 was found as an EST(BF486171) from adult testes; testis expression was noted in a Flybase communication (Bazinet,2000.7.10). CG7075 was also found to be a testisspecific gene as determined by microarray analysis(Parisi et al., 2004).

CG8291 is localized in the reproductive tissue in both genders, in nurse cells and oocytes in the female and in the testes in the male. In a microarray study, Parisi et al also found that CG8291 had enriched expression in the testis (Parisi et al.,2004). The CG8291 riboprobe also labelled a structure along the alimentary canal, the rectal bulb, of both males and females. The rectal bulb is thought to be responsible for the breakdown of the peritrophic membrane, which forms a barrier between the contents of the gut and the gut epithelium and is generated by the cardia(Miller, 1950). Unfortunately,there is no clear common feature among the tissues in which CG8291 is expressed which could lead to a hypothesis regarding CG8291function.

Interestingly, several of our unclassified candidate genes are organized in a cluster at the genetic region of 28C. The open reading frames for CG13793-CG13796, CG33296 and CG7075 all reside in what appears to be a region of genetic duplications, with up to 98% identity in portions of the repeated regions. CG13795 is 92% identical to CG13793 and 80-89% identical to CG33296, so it was difficult to confidently distinguish one sequence from another in this area. The organization of this region may indicate a unique relationship between these genes. One possibility is that a common exon could be spliced to any one of the genes to create differentially functioning proteins. Alternatively, the subtle sequence differences could be significant for the timing or localization of candidate expression. This genetic region could form some sort of locus that possibly functions similarly to the cholinergic locus in Drosophila, where two sequences with different functions share the first exons (Kitamoto et al.,1998). This region could also represent an area of gene duplication, in which transporters with different substrate profiles are generated as postulated by Boudko et al.(Boudko et al., 2005).

Five identified candidates (ine, blot, CG13793, CG13794 and CG33296) were eliminated from the in situ hybridization study. Two of the candidates, blot and ine, have already been cloned and localized (Burg et al., 1996; Huang et al., 2002; Johnson et al., 1999) and were not pursued further. Three candidates (CG13793, CG13794 and CG33296) were not pursued further because they did not have a predicted N terminus that agreed with other SLC6 homologs, and these candidates are located in a region likely with a genetic duplication. We approached CG13795 and CG13796 as representative sequences from this area. Neither CG13795 nor CG13796 were found in the adult, and therefore we did not pursue CG13793, CG13794 or CG33296 further. The fact that we were able to detect CG13795 in the embryo shows that the entire region cannot be attributed to pseudogenes and our RT-PCR results show that there is weak expression of each of these transcripts in the adult. Perhaps this genetic region represents an area where the genome has expanded and will generate an expansion of transporter population as described(Boudko et al., 2005).

Finally, one potential candidate with weak homology to SLC6 transporters(CG31904) was excluded from our study because homology extended only over a small fraction of the predicted protein. This small region is predicted to be fused with the adult cuticular protein 1 (which has no sequence similarity to SLC6 transporters), raising strong suspicions about an error in gene prediction.

One surprising aspect of this study is that we did not find candidate sequences for the histamine and octopamine transporters in the monoamine subfamily. The absence of a Drosophila octopamine transporter (OAT)is perplexing since an OAT gene does exist in another insect, Trichplusia ni (Malutan et al.,2002), and, like T. ni, there are cells in Drosophila that immunolabel for octopamine(Monastirioti et al., 1995). Furthermore, octopamine influences behaviors such as grooming and locomotion(Yellman et al., 1997). The question remains whether Drosophila uses an octopamine transporter and if so how divergent the Drosophila sequence is from that of T. ni.

The absence of a histamine transporter sequence is similarly puzzling. Histamine is the neurotransmitter of Drosophila photoreceptors, and genetic evidence indicates that a mechanism to accumulate histamine into photoreceptors exists in Drosophila(Melzig et al., 1998), as it does in other species (Battelle et al.,1999; Stuart et al.,2002; Stuart et al.,1996). Flies lacking the enzyme that synthesizes histamine,histidine decarboxylase (HDC), are blind and have lost histamine immunolabelling in the optic lobe. These flies can then be fed histamine,which restores vision and histamine immunolabelling, suggesting that in the fly there is a process by which histamine can be accumulated in the photoreceptors and restore function to those cells. Based on data from another species (Stuart et al., 2002; Stuart et al., 1996), one would predict that the histamine uptake is dependent on Na⁺ and Cl^-, as are the other SLC6 transporters.

Indeed, none of the SLC6 transporters were expressed in the pattern of histamine uptake, that is, in the photoreceptors and in cells located in the antennal lobe (Melzig et al.,1998), as would be expected in a direct uptake model. At the photoreceptor terminals, histamine may recycle in a more circuitous route through the surrounding glia (Borycz et al., 2002). Recent work on mutant flies has revealed that histamine may be conjugated to β-alanine into an intermediate compound,carcinine, by the enzyme Ebony. Ebony was localized by antibody to both epithelial and medulla neuropil glia that surround the photoreceptor terminals(Richardt et al., 2002). In this scheme, histamine released from the photoreceptors would be taken up into the glia and converted to carcinine. The carcinine then must be shipped to the photoreceptors where it would be catabolized into histamine and β-alanine by the enzyme tan. In this scheme the histamine transporter would reside in the surrounding glia and a separate transporter would exist in the photoreceptors to take up carcinine. However, we did not find an SLC6 homolog associated solely with either the photoreceptors or with their surrounding glia.

A sequence that is potentially involved with monoaminergic neurotransmission is the VMAT, CG33528(Greer et al., 2005). It is 44% identical to the mammalian VMAT2, which packages histamine into vesicles(Erickson et al., 1995). Our anti-sense riboprobe decorated a small number of cells in the central brain that are likely to be monoaminergic neurons(Monastirioti, 1999) and,unexpectedly, glial cells in a thin layer at the base of the photoreceptors. But this riboprobe did not label the photoreceptors, which are histaminergic and would be expected to express such a transporter. Immunocytochemistry also fails to find CG33528 in the photoreceptors or other histaminergic neurons although it co-localized with dopamine, 5-HT and octopamine neurons as expected (Chang et al.,2006).

Curiously, the CG33528 label at the base of the photoreceptors is in glia, not neurons. Two types of glia (fenestrated and pseudocartridge) have been described in this region based on anatomy and enhancer traps(Eule et al., 1995; Saint Marie and Carlson,1983). In the house fly, glia surround the photoreceptor axons as they traverse the lamina cortex. Fenestrated glia contain large vesicles(∼80-100 nm) and psuedocartridge glia contain even larger vacuoles(∼0.2-2.0 μm); since these glia are thought to help form a type of blood-brain barrier (Auld et al.,1995), the vesicles and vacuoles may relate to this function. The monoamines histamine and serotonin are both present in the lamina, histamine in the photoreceptors and serotonin in nerve terminals that course amongst the monopolar cell bodies in the lamina cortex(Buchner et al., 1988; Pollack and Hofbauer, 1991). A vesicular transporter in glia is not without precedent. Mammalian astrocytes have been documented to express vesicular glutamate transporters and release glutamate from vesicles onto neurons in a regulated manner(Bezzi et al., 2004).

Is CG33528 involved in the recycling of histamine at the photoreceptor terminals? We think the anatomy argues against this possibility. First,neither the processes of the fenestrated nor pseudocartridge glia extend into the lamina neuropil where the photoreceptors synapse onto the monopolar cells(Saint Marie and Carlson,1983). Epithelial glia reside in this region. Second, CG33528 has a much higher affinity for monoamines other than histamine(Greer et al., 2005). Given the juxtaposition of serotonin nerve endings with the glia expressing CG33528,perhaps this transporter is primarily concerned in some way with serotonin. Serotonergic terminals have been shown to penetrate into the lamina cortex(Buchner et al., 1988).

List of abbreviations
 
• AA

  amino acid

 
• Ag

  Anopheles gambiae

 
• AL

  antennal lobe

 
• AM

  midline cell

 
• BDGP

  Berkeley Drosophila genome project

 
• blot

  bloated tubules

 
• CNS

  central nervous system

 
• DAT

  dopamine transporter

 
• Dm

  Drosophila melanogaster

 
• EST

  expressed sequence tag

 
• GABA

  γ-aminobutyric acid

 
• GAT

  GABA transporter

 
• HDC

  histidine decarboxylase

 
• IAAT

  insect amino acid transporter

 
• ine

  inebriated

 
• LaN

  lamina neuropil

 
• MC

  medulla cell

 
• MeN

  medulla neuropil

 
• Ms

  Manduca sexta

 
• NA

  numerical aperture

 
• NGS

  normal goat serum

 
• NrxIV

  Neurexin IV

 
• NSS

  neurotransmitter:sodium symporter family

 
• OAT

  octopamine transporter

 
• PBS

  phosphate buffered saline

 
• PCR

  polymerase chain reaction

 
• PFA

  paraformaldehyde

 
• RT

  room temperature

 
• RT-PCR

  reverse transcription-polymerase chain reaction

 
• SerT

  serotonin transporter

 
• SNF

  sodium neurotransmitter transporter family

 
• TBS

  Tris-buffered saline

 
• TM

  transmembrane

 
• Tn

  Trichoplusia ni

 
• VMAT

  vesicular monoamine transporter

 
• VNC

  ventral nerve cord

We would like to acknowledge the contributions of many people that made this work possible. Steve Crews provided support for this foray into fly biology. We thank David Krantz and Rapheal Romero for valuable discussions and sharing unpublished results. Antibodies were provided by Manzoor Bhat(Neurexin IV) and the Developmental Hybridoma Studies Bank (Repo; Iowa City,IA, USA). We appreciate the histology expertise of Kirk McNaughton and Barb Taylor. Alan Fanning and Jim Anderson allowed us use of their Nikon microscope and Husok Lee generously provided us with an HDC plasmid. This work was funded by the University Research Council and Medical Alumni Association of the University of North Carolina at Chapel Hill and NIH grant MH65501, both to A.E.S.