Identification of two partners from the bacterial Kef exchanger family for the apical plasma membrane V-ATPase of Metazoa

Multicellular organisms rely on specialised transporting epithelia for the vectorial transport of ions, solutes and waste. It is clear that many epithelia rely not just on the basolateral Na⁺-K⁺ ATPase, but on an apical plasma membrane, proton-motive V-ATPase (Wieczorek et al., 1999). The plasma membrane V-ATPase can be deployed in several configurations (Fig. 1). When expressed alone, its intrinsic electrogenicity acts to electrically polarise an apical domain. This is particularly seen in sensory systems, such as the trichogen sensilla of insects (Stengl et al., 1992) or the human ear (Alper, 2002; Karet et al., 1999); the principle is that, by hyperpolarising the plasma membrane of adjacent sensory neurons, a small change in membrane resistance induces a larger change in receptor current, thereby increasing sensitivity. When colocated with an apical chloride channel, such as a member of the ClC family, the V-ATPase is capable of acidifying the apical space, as seen in the resorptive lacuna formed by osteoclasts (Alper, 2002; Kornak et al., 2001). A third use of the apical plasma membrane V-ATPase is conspicuous in nearly all insect epithelia; a specialised plasma membrane isoform of the V-ATPase (Allan et al., 2005) drives a colocated apical exchanger to achieve a net flux of alkali-metal cations (Na⁺, or more usually K⁺) (Azuma et al., 1995; Lepier et al., 1994). This close coupling of a primary ATPase and an exchanger (the `Wieczorek model') thus underlies what was classically described as an electrogenic K⁺ ATPase (Harvey et al., 1983).

Although the role of the plasma membrane V-ATPase is now established, the control by phosphorylation has been demonstrated (Voss et al., 2007) and the identities of the genes encoding the proteins making up the characteristic plasma membrane isoform are known (Allan et al., 2005), the exchanger has defied attempts at assignment. The obvious candidates are the members of the alkali-cation/H⁺ exchanger (CPA) family (Brett et al., 2005), particularly the Na⁺/H⁺ exchangers (NHEs), because insect epithelia are known to be amiloride sensitive (Giannakou and Dow, 2001), and amiloride sensitivity has been demonstrated for insect NHEs (Kang'ethe et al., 2007). However, a role in the apical plasma membrane for the classical NHEs has not yet been established unambiguously in insects.

In addition to the three classical NHEs (nhe1, nhe2 and nhe3) (Giannakou and Dow, 2001), two Drosophila genes (CG10806 and CG31052) have recently been annotated as showing distant similarity to the NHEs. However, it is clear that they are much more closely related to the prokaryote Kef (K⁺ efflux) genes kefB and kefC (which were the first genes of the widespread CPA2 exchanger family to be discovered), and they have been named NHA1 and NHA2, respectively (Brett et al., 2005). Intriguingly, representatives of this family are present in yeast (Maresova and Sychrova, 2005; Ramirez et al., 1998), plants (Maresova and Sychrova, 2006; Quintero et al., 2000; Sze et al., 2004), insects (Pullikuth et al., 2006) and mammals; the family is designated CPA2, by contrast with the better-known CPA1 (the NHEs) (Brett et al., 2005). In yeast, ablation of K⁺/H⁺ antiporter 1 (KHA1) causes salt sensitivity, and this can be rescued by homologous or heterologous expression of KHA candidates from yeast or plants (Maresova and Sychrova, 2006). A mosquito CPA2 was recently described (Rheault et al., 2007) and was named nha1, although it is not clear whether the protein transports Na⁺, K⁺ or both.

This paper seeks to establish whether any of these five exchanger genes are candidates for the Wieczorek exchanger, by virtue of their localisation. Such a gene should be expressed abundantly in tissues in which the plasma membrane V-ATPase is known to be important, notably most epithelia; it should localise to the apical surface; because most insect epithelia generate K⁺-rich secretions (Phillips, 1981), it might demonstrate the ability to carry K⁺ in preference to Na⁺; and its disruption might impact epithelial function.

The cloning and organisation of the three CPA1 genes (nhe1, nhe2 and nhe3) has been described previously. The two CPA2 genes (CG10806 and CG31052) are described here. CG10806 resides at 27C1 on chromosome 2L. It is annotated as having two very similar transcripts: isoform B (CG10806-RB) has five exons, whereas isoform A (CG10806-RA) has an additional exon immediately 5′ to the presumed start codon, potentially encoding a nearly identical protein, but with eight additional N-terminal residues (MFSSHKK). Neither protein has a signal peptide (as determined by SignalP 3.0) (Emanuelsson et al., 2007).

CG31052 is located at 94D6-7 on chromosome 3R. Based on computer annotation and EST evidence, it is thought to encode a single transcript. Again, although it is clearly an integral membrane protein, it does not appear to have a classical signal-peptide sequence.

Recently, FlyAtlas, a powerful online resource, has allowed the mapping of gene expression across multiple Drosophila tissues, based on Affymetrix microarray analysis (Chintapalli et al., 2007). The resource was mined for expression data for the three Drosophila CPA1 genes and two CPA2 genes (Table 1). For comparison, a plasma membrane V-ATPase-subunit gene, vha68-2 (Allan et al., 2005; Guo et al., 1996), is shown: this partner gene to the putative apical exchanger [similar to the other V-ATPase subunits (Allan et al., 2005)] clearly shows extremely high levels of expression specifically in epithelia (midgut, hindgut, larval and adult tubule). None of the NHE genes show this expression pattern; nhe1 is ubiquitously expressed, nhe2 is expressed only at low levels and nhe3 is enriched in nervous tissue. By contrast, the CPA2 genes show expression patterns similar to those for V-ATPase, with massive expression of both genes in the hindgut, and high levels of expression of CG31052 in the Malpighian tubules. The kefB and kefC homologues are thus highly plausible candidates for Wieczorek exchangers.

The expression patterns of the CPA2 genes were validated by in situ hybridisation (Fig. 2). As predicted by FlyAtlas, massive expression of both genes was seen in epithelial cells of the hindgut (although not in the rectal papillae), and significant expression was seen in the main segment of the tubules. The exclusion from rectal papillae, although unexpected in terms of classical models for insect excretion (Hanrahan and Phillips, 1983; Phillips, 1977), is consistent with expression patterns reported for V-ATPase genes in Drosophila (Allan et al., 2005). Both independent gene-expression measures thus show a high level of co-expression of V-ATPase with the CPA2 genes, but not the CPA1 genes, in epithelia.

Another requirement for a Wieczorek exchanger is that it is expressed on the same apical plasma membrane as the plasma membrane V-ATPase. For this, it is necessary to study the protein, rather than the mRNA. Accordingly, antibodies were raised against all five genes, and expression studied in multiple tissues (Figs 3, 4). Consistent with the FlyAtlas data, nhe1 was widely detected in the testes, accessory gland, midgut, hindgut and tubules (Fig. 3). However, in the midgut and hindgut, expression was confined to visceral muscle, and in tubule, expression was not localised to a plasma membrane. Antibodies to both long and short isoforms of Nhe2 performed similarly; they labelled the gonads and major epithelia. Although there was some evidence of apical localisation in the ejaculatory duct and the V-ATPase-containing cuprophilic cells of the midgut were also labelled, both antibodies labelled the stellate cells of the Malpighian tubule, rather than the V-ATPase-expressing principal cells. At present, the stellate cells in Drosophila are thought to provide a route for chloride and water shunt conductance (Kaufmann et al., 2005; O'Donnell et al., 1996), so Nhe2 might help provide a highly specific homeostatic mechanism for this cell type. Antibodies to Nhe3 labelled epithelia only weakly and with an intracellular perinuclear pattern; consistent with the array data, expression was much stronger in the CNS, with a cell-type-specific (rather than ubiquitous) staining pattern. None of the CPA1 genes thus localises consistently to the apical membrane of cardinal V-ATPase-energised epithelia, such as the Malpighian tubule.

By contrast, antibodies to both CPA2 genes conspicuously label the apical brush border of Malpighian-tubule principal cells throughout the length of the tubule (Fig. 4). Stellate cells were not labelled (data not shown). These are thus the first exchangers shown to localise to the same membrane as a plasma membrane V-ATPase. To confirm this, Drosophila were generated that were transgenic for the full-length open reading frames (ORFs) of CG10806 and CG31052, which were tagged with enhanced yellow fluorescent protein (eYFP) and under control of UAS. When these constructs were driven with c42, a driver specific for the tubule principal cell (Rosay et al., 1997; Sözen et al., 1997), the chimeric protein was targeted highly specifically to the apical brush border (Fig. 5). In epithelial cells, therefore, both CPA2 proteins are targeted unambiguously to the apical plasma membrane, as required for a Wieczorek exchanger.

Yeast cells deleted for their endogenous exchangers are sensitive to external salt, and have proved useful for studies of heterologous CPA1 (Kang'ethe et al., 2007) and CPA2 genes (Maresova and Sychrova, 2006) by complementation. Accordingly, the ORFs encoding Drosophila CG10806::eYFP and CG31052::eYFP were expressed in nha1 nhx1 yeast. Both heterologous proteins were expressed successfully at the correct size (Fig. 6). The two proteins localised differently in yeast; although CG31052 appeared to localise to the vacuole, as described previously for heterologous KHA proteins in yeast, CG10806 appeared to localise both to the vacuole and to the plasma membrane (Fig. 6).

The two proteins were also functionally distinct in their ability to protect against external salt loading. When exchanger-deficient yeast were transfected with the Drosophila CPA2 homologues and serially diluted onto high NaCl or high KCl medium, it was possible to assess the degree to which the Drosophila genes can rescue the salt-sensitivity phenotype of the mutant strains. CG31052 provided modest rescue of survival in high NaCl, whereas CG10806 was inimical to survival under this condition; the effect was slightly stronger for isoform A. By contrast, CG10806 (notably isoform B) rescued survival in high KCl. The reason why a slight difference in the N-terminal of CG10806 could have a large effect on the extent of the rescue is not clear, but could reflect differential targeting within the yeast cell. Nonetheless, if the distinct ionic specificities of CG10806 and CG31052 identified here in yeast were replicated in Drosophila, it would imply that the two genes are co-expressed on the apical surface of epithelia in order to provide apical exchange facilities for both Na⁺ and K⁺.

The data point clearly to a role for CPA2 members as Wieczorek exchangers. If this were the case, then manipulation of CPA2 gene expression would be expected to impact the transport status of V-ATPase-energised epithelia. Accordingly, the fluid-secretion phenotype of CPA2-overexpressing Malpighian tubules was compared with that of wild-type flies. Normal tubules secrete at a rate of around 0.5 nl/minute, and this can be doubled by the action of the diuretic neuropeptide Capa-1 (Kean et al., 2002) (Fig. 7). Although overexpression of CG31052 did not have an impact on resting or stimulated secretion, overexpression of CG10806 had a most striking effect. Basal levels of secretion were significantly higher, whereas Capa-1 stimulation caused the fluid secretion rate to fall away to zero (Fig. 7). The addition of cyclic GMP (cGMP; the second messenger for Capa-1 action) gave a similar response (Fig. 7C), implying that the Capa-1 effects can be attributed to its action through cGMP. Although CG10806 is implicated in K⁺ transport in yeast (Fig. 6), overexpression is clearly detrimental to tubule function in flies.

Does manipulation of gene expression levels of CPA2 family members impact the ionic composition of secreted fluid? Na⁺ and K⁺ concentrations were measured in pooled secretate from both overexpressing and RNAi flies, under control of the principal-cell-specific driver c42 (Fig. 8). Overexpression of CG31052 increased both Na⁺ and K⁺ levels in secreted fluid; there was a small but significant increase in Na⁺ concentration in secreted fluid of CG10806-overexpressing tubules. By contrast, RNAi alleles were without effect. We interpret these data as showing that the low levels of either CG31052 or CG10806, persisting after RNAi treatment, are sufficient to maintain secretion rates, but their normal expression levels limit alkali-metal transport rates. Thus, overexpression allows both Na⁺ and K⁺ to be secreted more effectively. The role of CG31052 appears dominant, consistent with its higher level of expression in tubule and other epithelia (Table 1), and the data suggest that it is capable of handling both Na⁺ and K⁺.

The CPA2 genes, originally identified in bacteria and shown to be functionally significant in yeast, are also found widely in sequenced animal genomes. In Drosophila, the two CPA2 genes are clearly family members, not just from sequence similarity, but because they rescue the corresponding yeast mutant. They provide the strongest candidates yet for the apical `Wieczorek exchanger' – which is posited to act as a partner for the apical V-ATPase – because they are apically located, are abundant in tissues with high levels of the plasma membrane V-ATPase, and their expression level affects epithelial transport of both Na⁺ and K⁺.

Why are two exchangers co-expressed in the same tissue? They appear to have similar but distinct properties when assayed both in yeast and in the intact fly. We speculate that the two CPA2 gene products might have different ionic specificities, one preferring Na⁺ and the other K⁺: from our yeast work, these would be CG31052 and CG10806, respectively. This two-exchanger system would provide great homeostatic flexibility by differential regulation of expression. Most insects secrete K⁺ preferentially, because this ion is considered to be present to excess in the diet (Harvey et al., 1983); however, accurate and separate homeostasis of Na⁺ and K⁺ in response to dietary or environmental load is nonetheless essential in all insects. Indeed, there is evidence for neuroendocrine regulation of Na⁺ excretion, akin to the action of atrial natriuretic peptide in vertebrates (Coast et al., 2005). In more-extreme cases, the blood-feeding insects, such as mosquitoes or Rhodnius prolixus (the `kissing bug'), have systems tuned to excrete Na⁺ in bulk, because it is present in the blood meal in huge abundance. Although differential regulation could be achieved by modulating basolateral permeability to Na⁺ or K⁺ (Maddrell and O'Donnell, 1993), the ability to selectively express a Na⁺- or K⁺-preferring exchanger at different stages in the life cycle would then be highly adaptive.

However, other questions remain. In particular, the very high pH – in excess of 12 – in the midgut of some insects (Dow, 1984) is thought to be caused by an intrinsically electrophoretic exchanger that exchanges two protons for one K⁺ ion (Azuma et al., 1995). Are insect CPA2 proteins electrogenic, or is the lepidopteran exchanger relatively unique? The imminent sequencing of lepidopteran genomes should make this relatively straightforward to study. In addition, it remains unclear whether the CG31052 and CG10806 gene products each transport Na⁺, K⁺ or both. The impacts on Malpighian tubules of manipulating either CG10806 or CG31052 in vivo are complex and subtle, probably because of partial redundancy between these co-expressed genes with their colocated gene products. In order to unambiguously resolve their function, it will be necessary to express them heterologously in Xenopus oocytes and cell lines, both separately and in combination. Potentially, yeast might provide a useful heterologous system for the study of animal CPA2 genes, because we were able to demonstrate a modest, but significant, rescue of exchanger-deficient mutants. Although it has been asserted that these genes are NHA homologues, and thus transport Na⁺, this is based on in silico analysis; such analysis is unlikely to be informative, because the ionic specificity of CPA2 family members is far from certain (Banuelos et al., 2002). It will also be interesting to explore the roles of CPA2 homologues (NHEDC1 and NHEDC2) in humans, in which this class of transporter remains surprisingly unexplored.

In addition to the three classical NHE family members (Giannakou and Dow, 2001), two further genes (CG10806 and CG31052) were annotated as potential alkali-metal/proton exchangers by the genome project in 2006. These were also recently described as candidate exchangers (Pullikuth et al., 2006). The CG31052 gene was independently identified in a high-throughput screen for protein-processing mutants in cell lines, and is also known as tango12 (transport and Golgi organization 12) (Bard et al., 2006); the other gene, CG10806, has been incorrectly annotated as a serine protease (www.flybase.org). Related sequences identified by BLASTP searches with the Drosophila-deduced peptide sequences for CG10806 and CG31052 were aligned to other species to validate their membership of the CPA2 family, and the new online FlyAtlas.org dataset (Chintapalli et al., 2007) was searched to identify expression patterns of all five genes in multiple adult Drosophila tissues by comparison with known plasma membrane V-ATPase-subunit genes (Allan et al., 2005).

Drosophila were reared on standard medium at 25°C and 55% relative humidity, and with a 12:12-hour photoperiod, as described previously (Giannakou and Dow, 2001). For overexpression studies, the ORF of CG31052 and both ORFs of CG10806 were translationally fused at their C-termini to the ORF for eYFP and were cloned into the UAS transformation vector pUAST. Constructs were microinjected commercially into w¹¹¹⁸ flies. Multiple transformants for each of these UAS-linked constructs were screened for efficiency of overexpression or knockdown by quantitative PCR after crossing to the appropriate GAL4 drivers.

To drive these lines, the following GAL4 drivers were crossed to UAS lines as required (induced expression described in brackets): actin-GAL4 (ubiquitous); heatshock-GAL4 (ubiquitous after exposure to 37°C for 15 minutes); c42 (tubule principal cells of main region) (Rosay et al., 1997).

Antigenic regions (identified in MacVector) of 15 aa were selected for all five Drosophila CPA genes (supplementary material Table S1), and rabbit antibodies generated commercially (Genosys, Paris). Antibodies were validated either by western blot (supplementary material Fig. S1) or by demonstrating stronger staining in overexpression lines of transgenic flies (see below). Expression of each gene was surveyed in multiple adult tissues using whole mounts as previously described (Broderick et al., 2004).

Whole-mount in situ analysis for CG10806 and CG31052 was performed as described previously for NHE1-NHE3 (Giannakou and Dow, 2001). DIG-labelled in situ hybridisation probes complementary to the full-length ORFs of CG31052 and CG10806 were synthesised. Alkaline hydrolysis was performed to shorten the probes to approximately 200 bp. Antisense (experimental) and sense (control) probes were synthesised for each gene and in situ hybridisation was carried out according to previously published protocols (Allan et al., 2005). Probes were used at a concentration of 2 μg/ml during the hybridisation step. After development, tubules, midgut and hindgut were mounted in 80% glycerol and images were captured using a Zeiss inverted microscope fitted with a Zeiss Axiocam camera.

Overexpression or RNAi constructs were crossed to the appropriate GAL4 drivers, and the progeny screened for lethality or obvious morphological defects.

Adult flies were briefly anaesthetised on ice and tubules dissected from week-old adults in Schneider's medium (Invitrogen). Fluid-secretion rates for each tubule were measured every 10 minutes for 1-2 hours as described previously (Dow et al., 1994), both under resting conditions and after stimulation by the diuretic neuropeptides Capa-1(Kean et al., 2002) or leucokinin (Terhzaz et al., 1999). Significant changes in secretion compared with parental controls were assessed by Student's t-test, taking P<0.05 as the critical level (two-tailed), and with at least ten tubules in each sample.

Na⁺ and K⁺ levels in secreted fluid were assessed by flame photometry. To obtain sufficient secretate for analysis (2 μl), the secreted fluid from 20 tubules was aggregated over the course of 1 hour.

Yeast strains (Saccharomyces cerevisiae) deleted for alkali-metal/proton exchangers (Δena1-4, Δnha1, Δnhx1) fail to grow in medium containing high levels of NaCl or KCl, and this can be rescued by expression of either yeast Nha1 or nha1 homologues from more-distantly related species, such as Arabidopsis (Maresova and Sychrova, 2006). Yeast strains G19 (MATα, ade2, his3, leu2, trp1, ura3, ena1-4::HIS3) (Madrid et al., 1998) and AXT3 (MATα, ade2, his3, leu2, trp1, ura3, ena1-4::HIS3, nha1::LEU2, nhx1::TRP1) (Quintero et al., 2000) were a kind gift of Imelda Mendoza (CICA, Spain). The ORFs encoding CG31052 and both isoforms of CG10806 (translationally fused to eYFP) were cloned into the yeast expression vector pYES2.1 (Invitrogen), allowing expression of the recombinant protein under control of the inducible GAL1-10 promoter upon galactose induction. The vector also contains the URA3 selectable marker. Yeast cells were handled as described previously (Gray et al., 1997). Briefly, they were transformed by the lithium acetate procedure, and transformants were selected on synthetic media containing glucose (2%) and lacking uracil. To confirm the expression and identify the sub-cellular localisation of the transformed constructs, cells were viewed by confocal microscopy after 6 hours of induction of protein expression, using a Zeiss 510 Meta confocal microscope. Rescue by the transgenes was assessed by plating onto medium supplemented with different concentrations of NaCl or KCl as described elsewhere (Maresova and Sychrova, 2006). The experiments were performed in synthetic minimal media (S) lacking uracil (–ura) and containing either the non-inducing sugar raffinose (Raff, 2%) or the inducing sugar galactose (Gal, 2%), designated SRaff-ura and SGAL-ura, respectively. Transformed yeast were grown overnight in SRaff-ura and protein expression was induced by addition of 2% galactose. Six hours later, cells were adjusted to OD₆₀₀=1 and serial tenfold dilutions made. A total of 10 μl of each dilution was spotted onto SGAL-ura agar plates, containing 2% Bacto-agar, and supplemented with either NaCl or KCl. Pilot experiments showed that wild-type and mutant strains could be discriminated at 200 mM NaCl or 1 M KCl, so subsequent experiments were performed at these concentrations. Plates were then incubated for 3 days at 30°C.

This work was supported by grants from the BBSRC and by funding from Pfizer Veterinary Medicine Discovery Research. We are most grateful to José Pardo and Imelda Mendoza (CICA, Madrid) for the provision of yeast strains; to Peter McCartney and Sue-Anne Krause for their help; and to Alan Taylor (University of Glasgow) for use of flame photometers.