Characterization of the Drosophila melanogaster alkali-metal/proton exchanger (NHE) gene family

The presence of cation/proton exchangers in eukaryotic cells was first proposed in 1961 by Mitchell as part of his chemiosmotic hypothesis, and these exchangers are now known to be important in pH homeostasis, cell volume regulation and transepithelial Na⁺ transport (Wakabayashi et al., 1997). NHEs are electroneutral and exchange 1 Na⁺ for 1 H⁺. The exchange is reversible and driven by the electrochemical gradients for the two cations. Under physiological conditions, the inwardly directed Na⁺ current produced by the Na⁺/K⁺-ATPase provides a constant force that extrudes H⁺ from the cell (Wakabayashi et al., 1997). There are at least six human NHE genes, NHE1–NHE6; five are plasma membrane NHEs (Klanke et al., 1995; Malakooti et al., 1999; Noel and Pouyssegur, 1995; Orlowski and Grinstein, 1997; Wakabayashi et al., 1997) and one (NHE6) is a mitochondrial exchanger (Numata et al., 1998). The NHE1 gene (Noel and Pouyssegur, 1995) appears to be the ubiquitously expressed, ‘house-keeping’ type of the exchanger that plays a major role in controlling the intracellular pH of nearly all animal cells. Similar genes have been identified in various other organisms, such as the rat, mouse, rabbit, pig, trout, Caenorhabditis elegans, Amphiuma tridactylum, yeast and bacteria (Attaphitaya et al., 1999; McLean et al., 1999; Noel and Pouyssegur, 1995; Padan and Schuldiner, 1994; Wakabayashi et al., 1997), and an NHE family member has been implicated in salt tolerance in the plant Arabidopsis thaliana (Shi et al., 2000).

In insects, NHEs may play a significant additional role. Some animal plasma membranes, including most insect epithelia, are energised by proton-motive forces instead of the basolateral Na⁺/K⁺-ATPase that energises most animal epithelia (Harvey and Wieczorek, 1997; Klein et al., 1991; Wieczorek et al., 1991). In insect epithelia, apical plasma membrane H⁺ V-ATPases generate transmembrane electrochemical gradients, which in turn drive other processes such as acidification, fluid secretion and sensory signalling. According to the Wiezcorek model, the electrogenic V-ATPase drives one or more alkali-metal/proton exchangers, resulting in a net transepithelial transport of Na⁺ or K⁺. It has been established that the two transport functions are pharmacologically distinct because the V-ATPase is bafilomycin-sensitive (Wieczorek et al., 1991) and the antiport is sensitive to amiloride (Wieczorek, 1992). Although there is no reason a priori to assign such V-ATPase-partner antiporters to the NHE family [indeed, in Manduca sexta midgut, the exchanger may be electrogenic (Azuma et al., 1995)], both Na⁺ and K⁺ transport in insect epithelia are amiloride-sensitive (Hegarty et al., 1992; Wieczorek, 1992). It is therefore particularly interesting to characterise the insect NHE exchangers, both as possible candidates for the Wieczorek exchanger and as potential components of animal cell ionic regulation. Surprisingly, although there are some preliminary reports, no paper describing a sequence for insect cation/proton exchangers has been published.

The fruit fly Drosophila melanogaster is a useful genetic model with a completed genome sequence (Adams et al., 2000), powerful transgenic technology (Rubin, 1988; Spradling and Rubin, 1982; Spradling et al., 1995). It also serves as a good experimental model, permitting the use of biochemical, cell biological and physiological techniques in disciplines such as developmental biology, neurobiology (Rubin, 1988) and integrative physiology (Dow et al., 1998). The Malpighian tubule of D. melanogaster is known to be sensitive to both bafilomycin and amiloride (Dow et al., 1994), consistent with the V-ATPase/antiporter in that it has been shown to be energised by an apical V-ATPase confined to the principal cells (Davies et al., 1996). However, amiloride is a relatively non-specific probe for NHE function because it also inhibits a range of Na⁺ channels (Kleyman and Cragoe, 1988). In the present paper, we show that fluid secretion in the Malpighian tubules is inhibited by amiloride derivatives that are consistent with inhibition of NHEs rather than Na⁺ channels. Furthermore, no expression of epithelial Na⁺ channels (ENaCs) could be detected by reverse transcriptase/polymerase chain reaction (RT-PCR) in Malpighian tubules. In contrast, the Drosophila NHE family is shown to consist of three genes, called DmNHE1, DmNHE2 and DmNHE3, that encode distant relatives of the NHE exchanger family, all of which are expressed in Malpighian tubules.

Oregon R (wild-type) Drosophila melanogaster were maintained on a 12 h:12 h light:dark cycle on standard corn meal/yeast/agar medium at 25°C. All manipulations, unless stated otherwise, were carried out at room temperature (22–25°C).

Malpighian tubules were dissected from adult female and male flies, and fluid secretion assays were performed as described previously (Dow et al., 1994). The bathing medium was a mixture of Schneider’s insect culture medium and Drosophila saline (1:1 v/v). Drosophila saline (pH 6.7) consisted of (in mmol l^–1): NaCl, 117.5; KCl, 20; CaCl₂, 2; MgCl₂, 8.5; NaHCO₃, 10.2; NaH₂PO₄, 4.3; Hepes, 15; glucose, 20. Volumes of secreted fluid were determined at 10 min intervals. The data were analysed using an Apple Macintosh computer and Excel 4.0. All data are reported as means ± s.e.m. Statistical significance of differences between treatments was assessed using Student’s t-test for unpaired samples, taking the critical value of P to be 0.05 (two-tailed).

Cardioacceleratory peptide 2b (CAP_2b) and Drosophila leucokinin were custom-synthesised by Research Genetics, Inc. and added to tubules at 10^–7 mol l^–1. This combined treatment powerfully stimulates diuresis, acting both on active cation transport and on the Cl^– shunt conductance (Dow and Davies, 2001; Dow et al., 1998), and so was expected to unmask any inhibition by amiloride. Amiloride (Sigma-Aldrich A7410) and 5-N,N-dimethyl amiloride (DMA) (Sigma-Aldrich A4562) were dissolved to 10–500 mmol l^–1 in dimethylsulphoxide (DMSO), then 1:100 in Schneider’s/saline, and used at a range of concentrations together with 1:100 (final dilution) DMSO in Schneider’s/saline as the vehicle. Benzamil (Sigma-Aldrich B-2417), 2′,4′-dichlorobenzamil (DCB; Molecular Probes D-6898) and 5-N-ethyl-N-isopropyl amiloride (EIPA; Sigma-Aldrich A3085) were dissolved in methanol to 50–100 mmol l^–1, then diluted 1:100 in Schneider’s/saline and used at a range of concentrations, together with 1:100 (final concentration) methanol in Schneider’s/saline as the vehicle. Neither DMSO nor methanol vehicles had any effect on Malpighian tubules at these final concentrations (data not shown). The amiloride analogue was added to half the tubules after 30 min, and all the tubules were then treated with Drosophila leucokinin and CAP_2b at 60 min. Secretion assays were performed at a range of concentrations from 10^–4 mol l^–1 to 10^–8 mol l^–1, and dose/response curves were plotted. For each experimental set of at least 10 tubules, the response to amiloride was defined as the mean maximum secretion rate (controls) minus the mean maximum secretion rate (amiloride-treated). This value was expressed as a percentage of the control maximum secretion rate.

Cyberscreening was performed using Netscape Communicator 4.5 on an Apple Macintosh computer and searching NCBI (http://www.ncbi.nlm.nih.gov/) and BDGP (http://www.fruitfly.org) databases with BLASTN, BLASTP, BLASTX or TBLASTN searches, as appropriate. Sequence alignments were performed and displayed using MacVector 6.5.1 or 7.0, AssemblyLIGN, SeqVu 1.0.1, ClustalW PPC and TreeView PPC.

mRNA was prepared using the Dynabeads Oligo (dT)₂₅ kit according to the manufacturer’s protocol (Dynex Technologies) and reverse-transcribed with SUPERSCRIPT II RNase H^– reverse transcriptase (Life Technologies) to produce a solid-state cDNA library. For each PCR reaction, 1 μl of beads, corresponding to cDNAs derived from one Malpighian tubule or approximately 0.2 head, was used. RT-PCR was performed on cDNA from whole male flies, whole female flies, heads, bodies, tubules, larvae and pupae using primers designed to bracket introns as a guard against genomic DNA contamination.

PCRs were performed as follows: 94°C for 1 min; followed by 30 cycles of 94°C for 30 s, 55°C for 30 s and 72°C for 2 min (depending on the length of the DNA template), and finally one cycle of 72°C for 5 min. PCR reactions were analysed on 1 % agarose gels, stained with ethidium bromide and photographed according to standard protocols (Sambrook and Russell, 2001).

Low-stringency searches of the Drosophila genome allowed three candidate genetic loci to be identified. cDNA clones were identified by BLAST searching against expressed sequence tags (ESTs), using genomic sequence for each gene as a probe. All available clones were obtained from Research Genetics. No EST hits were obtained for DmNHE2, implying that it was not widely expressed. EST clones HL05853, AT11019 and LP03712, identified as the longest available 5′ clones for DmNHE1, DmNHE2 and DmNHE3, respectively, on the basis of available EST information, were sequenced in full on both strands. To survey for possible alternative 3′ splicing, the 3′ ends of the other EST clones were also sequenced.

Primers used were as follows:

DmNHE1-2306R, CCCCACAACAGCCATTTAAT; Dm-NHE1-F1, AGCGACCACGTCACGTTTTGTC; DmNHE3-1943F, TACGAATGGCAGTTTGGG; DmNHE3-2700R, CATTTTCGATTTCAGTTGAGACC; DmNHE2-F2, TC-TACATGCTTCCACCGATTATCC; DmNHE2-R2, AGT-GAGGCAAATAGAAACACGTCC; DmNHE2-1684F, TT-GGCGTGGTGCTCTATTTC; DmNHE2-3509F, CCT- GCGGAAAGATGGGAATTTAC; DmNHE2-3803F, TGT- GATGTACCACATGATGGAG; DmNHE2-3849R, TGT-CCAAGCCAATCTCATTGTAGG; DmNHE2-4088R, AAA- TGGGTTCTATGACACGCAC; DmNHE2-4859F, TCAC-TTGATGGCTGGAATTGAG; DmNHE2-5324F, GAGCTG-AGCCGAAGATCATC; DmNHE2-5367R, CATCGTGAG-TTTGGAGTACGTC; DmNHE2-59554R, TCAGAGATC- AGAGAGACAGAGAGAG; PM001, CGTTAGAACGCGG-CTACAAT; M13 Forward, CTGGCCGTCGTTTTAC; M13 Reverse, CAGGAAACAGCTATGAC.

It has been shown previously that Drosophila Malpighian tubules are sensitive to amiloride (Dow et al., 1994). Amiloride inhibits Na⁺ channels, Na⁺/H⁺ exchangers and Na⁺/Ca²⁺ exchangers (Kleyman and Cragoe, 1988; Orlowski and Grinstein, 1997). However, work carried out on mammalian NHEs and Na⁺ channels has shown that different amiloride analogues specifically inhibit channels or exchangers. Channels are inhibited more effectively by amiloride or by 2-carbonylguanido substituents, such as benzamil, whereas NHEs are much more sensitive to R5 group substituents, such as 5-N-ethyl-N-isopropyl amiloride (Kleyman and Cragoe, 1988; Orlowski and Grinstein, 1997). In Malpighian tubules, it was therefore imperative to study amiloride pharmacology with fluid secretion assays using a number of amiloride analogues that had previously been shown to be mammalian NHE- or Na⁺-channel-specific. Dose/response curves for the five analogues are shown in Fig. 1. All analogues inhibited fluid secretion, but at different concentrations. By comparison with results obtained in vertebrate systems (Table 1), it is clear that the relatively low sensitivity to amiloride and 2-carbonylguanidino substituents compared with R5 substituents (particularly EIPA) is diagnostic for an NHE, rather than a channel, target for amilorides in Malpighian tubules.

Although the pharmacology of the response to amiloride of Malpighian tubules was consistent with an effect on an NHE, rather than a channel, a genomic/RT-PCR strategy was employed to establish whether Na⁺ channels are present in Malpighian tubules that could explain this inhibition of fluid secretion by amiloride.

Amiloride-sensitive Na⁺ channels previously described in Drosophila include two Drosophila degenerin/ENaC family genes, pickpocket (ppk) and ripped pocket (rpk) (Adams et al., 1998). Pickpocket appears to be abundantly transcribed in early-stage embryos and is possibly involved in early development, whereas ripped pocket is found in a subset of neurons of the peripheral nervous system and is amiloride-sensitive (Adams et al., 1998). Ripped pocket is identical to dGNaC1 (Darboux et al., 1998a) and pickpocket is identical to dmdNaC1 (Darboux et al., 1998b), genes found in the peripheral nervous system and the gonads, respectively. dGNaC1 has been shown to be amiloride-sensitive by expression in Xenopus laevis oocyte, whereas the amiloride sensitivity of dmdNaC1 is inferred from sequence similarity.

Our search of the Berkeley Drosophila genome project for Na⁺ channels led to the identification of nine genes, rpk, ppk, CG10972, CG14398, CG4805, CG8546, CG9499 and Nach. These genes are divergent in sequence, being at least as related to the human search sequence as to each other (Fig. 2A). By RT-PCR, none of these genes was detected in Malpighian tubules: some appeared to be expressed in heads or whole flies only, and for others, no expression could be detected (Fig. 2B). This is consistent with recent data implicating Drosophila ENaCs in very specialised roles in thermoreception (Zinkevich et al., 2001) and salt taste perception (Liu et al., 2001).

A low-stringency BLAST search of the Berkeley Drosophila genome database using the human NHE1 protein sequence revealed three putative Drosophila members of the NHE family of exchangers, termed DmNHE1, DmNHE2 and DmNHE3 in the order in which they were identified. Three EST hits have been described for DmNHE1: HL05853, GH04225 and GH04168, with HL05853 being the longest. Nine EST hits were identified for DmNHE3: LP03712, GH16168, GH025044, GH27182, LD37666, LD07057, LD20719, LP02917 and SD07542, the last being the longest. DmNHE2 was identified in genomic clone DS02919 from cytological position 39A3–39B1, but there were no EST clones available for this gene. HL05853, the EST clone for DmNHE1, and LP03712, the EST clone for DmNHE3, were sequenced and the cDNA and inferred protein sequences determined (Fig. 3, Fig. 4). DmNHE1 (GenBank accession number AF142676) or CG12178, localised to cytological position 21B1, encodes a 649-amino-acid protein (AAD32689) with a predicted relative molecular mass of 71 277, whereas DmNHE3 (AF199463) or CG11328, localised to 27A1, encodes a 687-amino-acid protein (AAF13702) with a predicted relative molecular mass of 71 276.

For DmNHE2, the story is more involved. Partial cDNA sequencing of DmNHE2 (which was not available as an EST) was achieved with an RT-PCR-based strategy, designing primers against putative exons flanking introns and sequencing the cDNA fragments obtained from Malpighian tubule cDNA. This partial sequence was deposited in GenBank (accession AF239763). However, another cDNA sequence corresponding to DmNHE2 has been deposited in GenBank (X. Lin, D. C. Huang, W. Yan and D. L. Barber, unpublished: accession number AF235935). Although longer than our sequence, it is clearly incomplete at both the 5′ and 3′ ends, lacking a 5′ untranslated region (UTR) and finishing on an exon boundary without a credible polyadenylation site. Since then, the Berkeley Drosophila Genome project (BDGP) has produced an automated annotation of the locus, which erroneously splits the AF235935 transcript into two genes, named NHE2 and CG9255. Very recently, four new ESTs from an adult testes library have been deposited in GenBank (clones AT08048, AT04839, AT11019 and AT12693). They all start within 15 bases of each other and extend the available sequence by nearly 400 bases 5′ to the AF23595 sequence. There was, therefore, no single sequence that describes the very large DmNHE2 transcript. Accordingly, we sequenced the longest available testes clone (AT11019) to resolve the issue. Our sequence (Fig. 5) is clearly a novel splice variant of DmNHE2. Neither the AF23595 nor Gadfly sequences had a credible signal peptide (PSORT II), which is essential for this integral membrane protein. Our longer cDNA had several novel potential methionine initiator sites, only one of which read into a plausible signal peptide. Accordingly, we take this to be the likeliest start site (Fig. 5).

Using the PSORT II (Nakai and Horton, 1999) and PROSITE (Bairoch et al., 1997) protein prediction programmes, it appears that all three Drosophila NHEs are plasma membrane integral proteins (61 %, 57 % and 70 % predicted plasma membrane targeting, respectively), with 11 putative transmembrane domains. Although some of the Drosophila NHEs appear to sit close to mitochondrial NHEs in the similarity tree (see Fig. 7), they lack mitochondrial targeting sequences (PSORT II). The transmembrane probability plots (von Heijne, 1992) for the three proteins can be seen in Fig. 6. Clearly, the Drosophila NHEs share the same organisation (short N terminus, compact 10–12-pass transmembrane domain of approximately 400 residues, long hydrophilic C terminus) with human NHE1, the archetype of the NHE family, and so can be assigned with confidence to the family. DmNHE1 and DmNHE3 contain a signal peptide sequence with a putative cleavage site at residue 18/19. For DmNHE2, there are three contending sequences. The Gadfly CG9255 DmNHE2 sequence is clearly incomplete (Fig. 6), so we reject this automated annotation in this case. The translation of the AF23595 sequence for DmNHE2 lacks an N-terminal signal peptide sequence; however, if translation were to start at the initiator methionine corresponding to bases 628–630 of our sequence, then there would be a clear 21-residue signal peptide (PSORT II prediction), and the disposition of the N terminus of the peptide would be almost identical to that of human NHE1 (Fig. 6). At the C terminus, DmNHE2 appears to encode an extremely long cytoplasmic C-terminal tail, with a very short 3′UTR that lacks a polyadenylation site. However, our confidence in this structure is increased by the conceptual translation of an Aedes aegypti sequence which, though lacking an N-terminal signal peptide, has a very similar C-terminal cytoplasmic domain. It it therefore likely to be a partial, but authentic, cDNA. However, the testes cDNA we describe here appears to represent a complete, authentic cDNA that has a 5′UTR, a signal peptide and a polyadenylation site. DmNHE2 must therefore be considered to have alternative transcripts, both identical at the 5′UTR and the N-terminal and membrane-spanning coding regions, but having very different C termini through facultative read-through of the last exon of our sequence (Fig. 5). As the cytoplasmic C terminus is considered to have control properties, this difference is likely to be functionally very significant.

All three isoforms have multiple putative N-glycosylation sites, putative phosphorylation sites for cAMP- and cGMP-dependent protein kinases, protein kinase C and casein kinase type II sites (Prosite predictions). DmNHE3 also appears to have two leucine zipper motifs in transmembrane region (TM) 7, which is quite unusual because only human NHE5 and the Arapidopsis thaliana NHEs have leucine zipper motifs (three in human NHE5 in TMs 1 and 2; one in Arabidopsis thaliana NHE in the intracellular domain between TM6 and TM7).

The Drosophila NHE sequences were used in further BLAST searches to identify other members of the NHE family. Amongst proteins identified were those from a variety of species, mammalian (human, rat, bovine) and other vertebrate (trout, Cyprinus carpio NHE) and invertebrate (crab, Caenorhabditis elegans) sequences. Amongst these sequences were a number of novel human NHE protein sequences, a protein previously identified as KIAA0939, which is the closest homologue to DmNHE1 (55 % identity at the amino acid level) and 13 sequences newly emerged from the human genome project, partial protein sequences from working drafts of the human genome. KIAA0939 was identified in IMAGE clone 3134373 (GenBank accession number BF222481), from a Homo sapiens kidney cDNA library.

The Drosophila NHEs were aligned to the human and other members of the NHE family of exchangers (Fig. 7), and a phylogenetic tree was constructed to examine similarities between the different family members (Fig. 8). The alignment shows that the Drosophila NHEs are definitely members of the family, with different Drosophila NHEs more similar to different branches of the family. DmNHE1 is most closely related to the novel human NHE KIAA0939 and to some novel sequences from working drafts. DmNHE3 is more closely related to some new human NHE draft sequences and the human NHE6, which is the mitochondrial NHE isoform (Numata et al., 1998). DmNHE2, in contrast, is more closely related to the crab NHE from Carcinus maenas and to the exchanger from the yellow fever mosquito Aedes aegypti.

The amiloride-binding region, which is present in all known NHE family members (Wakabayashi et al., 1997), is present in all the DmNHEs, although it is not perfectly conserved, which could explain the slightly lower amiloride sensitivity we observed (Table 1). The highly conserved region between NHE isoforms, which is thought to be involved in ion exchange or at least essential for ion translocation (Noel and Pouyssegur, 1995), is also well conserved in the Drosophila NHEs, which are, however, more divergent than the other family members. In that respect, DmNHE2 is the most similar to the other family members. Another domain present in many NHEs is a calmodulin-binding region found in the cytoplasmic C terminus of the exchangers. This region, termed calmodulin-binding region A in NHE1, is also found in the NHE1 isoform of other organisms, such as the mouse, Xenopus laevis, bovine NHE1, rat, pig, Amphiuma tridactylum and rabbit, in the NHE2 isoform of human and rabbit and the rat NHE4 isoform. It is not, however, found in any of the Drosophila NHEs.

Expression patterns of DmNHE1, DmNHE2 and DmNHE3 were mapped by RT-PCR using cDNA derived from a variety of tissues using primers designed to bracket introns and using genomic DNA as a positive control for the reaction (Fig. 9). This experiment showed that all three genes are expressed in the head, body and Malpighian tubules and at all developmental stages, indicating that they are widely expressed.

The Drosophila Malpighian tubule is exquisitely sensitive to agents hypothesised to affect the components of the Wieczorek model for insect epithelia, namely the apical V-ATPase and associated exchanger. However, while bafilomycin is agreed to be a selective inhibitor of V-ATPase, amiloride could target a range of molecules on apical or basal surfaces. Here, we show that the Malpighian tubules are blocked by a characteristic range of amiloride derivatives characteristic of exchangers, rather than channels. The order of inhibition of fluid secretion in Malpighian tubules is EIPA≫2,4-dichloro-benzamil>DMA>amiloride≈benzamil (Fig. 1; Table 1). These results are consistent with those recently obtained in Aedes aegypti (Petzel, 2000), increasing our confidence that amiloride targets NHEs in insect Malpighian tubules. This is also consistent with our RT-PCR data showing that Drosophila NHE genes, but not ENaC genes (Fig. 9, see also Fig. 2), are expressed in Malpighian tubules. Our failure to identify ENaCs in Malpighian tubules is consistent with electrophysiological analysis, which showed that amiloride inhibited transepithelial Na⁺ secretion in Aedes aegypti Malpighian tubules without any effect on transepithelial and fractional membrane resistance (Hegarty et al., 1992).

This paper describes three genes that appear to encode the Drosophila members of the NHE gene family. Their protein sequences are quite different from the protein sequences of the other members of the family, but there is sufficient similarity to the other NHEs to assign them unambiguously to this group of proteins (Fig. 6). More specifically, DmNHE1 appears to be very similar to a novel human NHE, KIAA0939, which has been found in kidney (IMAGE 3134373) and brain (GenBank accession number AB023156) (Nagase et al., 1999). DmNHE2 is most similar to two invertebrate NHEs, the NHE found in Carcinus maenas and the newly described NHE3 in Aedes aegypti (GenBank accession number AF80554; S. S. Gill, H. Wediak and L. S. Ross, unpublished). DmNHE3 sits near human mitochondrial NHE6 (although DmNHE3 encodes no mitochondrial targeting sequences) and also close to Arabidopsis thaliana and yeast genes.

The three DmNHEs described above are predicted to be plasma membrane integral proteins with 10–12 transmembrane domains just like the other members of the family (Fig. 6). All the Drosophila NHEs have a putative signal peptide and a possible cleavage site. This is similar to the position in mammalian NHEs, although it is not certain whether the signal peptide is ever cleaved (Zizak et al., 2000); see Shrode et al. (Shrode et al., 1998) and Wakabayashi et al. (Wakabayashi et al., 2000). The presence of distinct messages for DmNHE2, encoding peptides with differing C-terminal domains, has interesting implications for control of the exchanger.

In principle, the elucidation of genes in Drosophila would allow the reverse genetic analysis of their function in mutants. However, there are no candidate P-element insertions documented at any of the three loci. The nearest mutation is an insertion, 2 kb beyond the 3′ end of DmNHE3, that generates a lethal recessive phenotype. However, this insertion is at the 5′ end of a novel gene (CG11329), and so the lethality is probably attributable to the latter locus.

Are any of these genes candidates for the Wieczorek exchanger? Their relative dissimilarity to cardinal vertebrate NHEs (Fig. 8) would allow them to be ascribed different functional properties. For example, DmNHE2 sits in a branch of the similarity tree with only invertebrate representatives and so would be a strong candidate. Our data show that, in Drosophila, all three exchangers are widely expressed (Fig. 9) and are certainly present in a relevant epithelium (the Malpighian tubule). However, the same general expression pattern would argue against a specialised role in transporting epithelia only, and our pharmacological analysis (Fig. 1) does not distinguish between an apical or basolateral localisation. Recent electrophysiological evidence suggests that amiloride may be acting at the basolateral membrane of Aedes aegypti Malpighian tubules (Petzel, 2000), and our results cannot be taken to contradict this view. In insects, it may have been naïve to assume that sensitivity to bafilomycin and amiloride is sufficient proof that an epithelium conforms to the Wieczorek model. However, whether DmNHE1, DmNHE2 or DmNHE3 transpires to be the elusive apical exchanger, or a vital part of the cell’s ion-regulatory machinery, the description of this gene family in a genetic model organism should be useful.

We are grateful to the Wellcome Trust for a studentship to M.E.G. and to the BBSRC for grant support to J.A.T.D. We thank Dr Shireen A. Davies for critical reading of the manuscript.