The Na⁺/K⁺ ATPase is required for septate junction function and epithelial tube-size control in the Drosophila tracheal system

Epithelial and endothelial tubes are the essential functional units of many organs including the vascular system, kidney and lung. All of these organs contain tubes of characteristic but widely ranging sizes because efficient flow through any tubular network requires the size of a tube be matched to the flow it supports. A tube that is too small in diameter will have insufficient flow (e.g. obstructive heart diseases), whereas a tube that is too large in diameter can impinge on surrounding tissue (e.g. enlarged kidney tubules in polycystic kidney disease) or have an undesirably slow rate of flow (e.g. blood clotting in cavernous hemangiomas). Thus, control of tube architecture presents fundamental cell biological and developmental problems with medical implications. However, the molecular mechanisms by which epithelial cells measure, change, and then maintain tube diameter and length are not known.

To identify and study the functions of genes required for epithelial tube-size control we are using molecular genetic approaches with the Drosophila tracheal system. The tracheal system is the gas exchange organ of the fly, but also resembles and performs some of the same functions as the vertebrate vascular system by directly delivering oxygen through a ramifying network of tubes (Manning and Krasnow, 1993). The tracheal system is created from clusters of invaginated epithelial cells that organize into branches which extend and interconnect to create a complex tubular network(Affolter and Shilo, 2000; Samakovlis et al., 1996). The genetic tools available in Drosophila, coupled with the reproducible and rapid development of the tracheal system, make it a powerful system for studying the genetic and molecular basis of tubesize control(Beitel and Krasnow, 2000; Lubarsky and Krasnow,2003).

Epithelial tube morphogenesis has been shown to require coordinated cell shape changes and dynamic adjustments of cell junctions(Lubarsky and Krasnow, 2003; O'Brien et al., 2002; Hogan and Kolodziej, 2002). Adherens junctions have been shown to play an important role in tracheal morphogenesis (Tanaka-Matakatsu et al.,1996; Uemura et al.,1996; Oda and Tsukita,1999; Chihara et al.,2003; Lee and Kolodziej,2002) and in vertebrate epithelial tube formation(Pollack et al., 1997; Pollack et al., 1998). However, to date, the involvement of Drosophila septate or vertebrate tight junctions in epithelial tube morphogenesis has not been examined.

Insect septate junctions form the trans-epithelial diffusion barrier that regulates passage of solutes through the spaces between adjacent cells in an epithelium (Tepass et al.,2001). In vertebrates, the paracellular diffusion barrier in epithelial cells is provided by tight junctions(Tsukita et al., 2001; Anderson, 2001). Despite their common barrier function, septate and tight junctions are generally referred to as analogous structures because they have very different morphologies when examined using electron microscopy and because septate junctions are located basolateral to the adherens junction while tight junctions lie apically. Nonetheless, Drosophila septate and vertebrate tight junctions contain some similar proteins that are crucial for their function(Tepass et al., 2001). For example, Drosophila discs large encodes a PDZ domain-containing MAGUK scaffold protein similar to vertebrate ZO-1; coracle is similar to band 4.1; and scribble to human scribble(Willott et al., 1993; Takahisa et al., 1996; Fehon et al., 1994; Bilder and Perrimon, 2000; Nakagawa and Huibregtse,2000). These observations raise the possibility that the functional similarities of septate and tight junctions may reflect similarities in molecular architecture.

In previous work, we showed that the tracheal system undergoes highly regulated tube-size increases during development and identified eight `tube expansion' genes that are specifically required for remodeling the size of the tracheal tubes once the initial network has formed(Beitel and Krasnow, 2000). We show that one of those genes is the nrv2 locus, which encodes two isoforms of a β subunit of the Na⁺/K⁺ ATPase, and that mutations in the ATPα α subunit (Atpalpha– FlyBase) locus also cause similar tracheal tube-size defects. Further,we show nrv2 and ATPα mutations disrupt septate junction function.

The Na⁺/K⁺ ATPase is an α/β heteromultimer that creates the essential electrochemical gradient across the plasma membrane by transporting two K⁺ into and three Na⁺ out of the cell for each ATP hydrolyzed (Blanco and Mercer, 1998; Chow and Forte,1995). The α subunit is a ten transmembrane-domain protein containing the Na⁺ and K⁺ channels and the phosphorylation and nucleotide binding sites(Chow and Forte, 1995). The single transmembrane domain β subunit is required for transport of theα subunit out of the ER (Geering et al., 1996; Hasler et al.,1998) and is thought to modulate the affinity of the αsubunit for Na⁺ and K⁺(Blanco and Mercer, 1998; Chow and Forte, 1995; Hasler et al., 1998). In flies there are three β subunit loci, nrv1 and nrv3 (CG8663),each of which produce one isoform, and nrv2 which encodes two isoforms, Nrv2.1 and Nrv2.2. There are two α subunit loci, ATPα that produces at least 12 α subunit isoforms(Palladino et al., 2003), and CG17923, which appears to be minimally expressed because its transcripts are poorly represented in existing cDNA libraries. Of these loci, only nrv1,nrv2 and ATPα have been characterized in detail, primarily for their roles in the nervous system(Lebovitz et al., 1989; Palladino et al., 2003; Sun and Salvaterra, 1995b; Sun et al., 1998). A tracheal phenotype was recently described for ATPα mutants, althoughβ-subunit mutants were not analyzed(Hemphala et al., 2003).

In vertebrates, different Na⁺/K⁺ ATPase isoforms have been proposed to have unique functions because they are expressed with tissue and temporal specificity, and because different isoforms have distinct biochemical and pharmacological properties in vitro(Blanco and Mercer, 1998). However, in vivo experiments have so far failed to provide support for this hypothesis (Weber et al.,1998). Our results show that septate junctions and specific Na⁺/K⁺ ATPase isoforms have previously unidentified roles in tracheal tube-size control and that the tube-size control and trans-epithelial barrier functions of septate junctions are distinct.

Flies were obtained from the Bloomington Stock Center or published sources,except for the following: nrv2^23B and nrv2^l(2)k13315e22c-GAL4 from J. Genova and R. Fehon(personal communication) and btl-GAL4 UAS-GFP from M. Metzstein and M. Krasnow (unpublished). nrv2^nwu3 was generated by imprecise excision of the l(2)k04223 P-element insertion and deletes the first two and most of the third common exons. The genomic sequence across the deletion breakpoints is AAGGCCCTCGCTATAACCAAGGAGAATAAC. This allele is likely to be a molecular null because the deleted region encodes transmembrane domain and extracellular regions, which are required for assembly with the αsubunit and for export of the Na⁺/K⁺ ATPase from the endoplasmic reticulum (Geering et al.,1996; Hasler et al.,1998). Germline clones were produced with the nrv2^nwu3, FRT 40A chromosome using the FLP-ovoD method(Chou and Perrimon, 1996). EMS mutagenesis generated two nrv2 nonsense mutations, one in the first common exon (nwu5: CTGCATGTGG→CTGCATGTGA)and one in the second common exon (nwu6:CAAGCACTGG→CAAGCACTAG) that were identified by non-complementation with nrv2 ^l(2)k4223. Third chromosome lines were balanced by either currently available balancers or the TM6B Hu Sb¹ e Tb¹ ca¹ chromosome onto which we mobilized the dfd Hz2.7lacZ element(McGinnis et al., 1990).

Antibodies 2A12 and TL1, embryo fixation, staining and staging procedures are described elsewhere (Samakovlis et al., 1996). Other antibodies used were AS55(Reichman-Fried et al., 1994),H5F7 (Sun and Salvaterra,1995a), anti-α5(Lebovitz et al., 1989),anti-Armadillo (Riggleman et al.,1990), anti-Coracle (Fehon et al., 1994), anti-DLG (Woods et al., 1996), anti-DLT (Bhat et al., 1999), anti-Neurexin(Baumgartner et al., 1996) and anti-β-Galactosidase (Capel). Embryos stained with α5 anti-αwere fixed in 4% paraformaldehyde and hand devitellinized. For H5F7 anti-β, standard formaldehyde/heptane fixation and methanol devitellinization produced basolateral staining that overlapped with Coracle and Neurexin, while paraformaldehyde fixation and hand devitellinization yielded cytoplasmic/perinuclear staining. Using either method, ectodermal staining was reduced to background levels in nrv2 null mutants. H5F7 staining of embryos expressing UAS-nrvX transgenes revealed that for the btl-Gal4 and e22c-Gal4 drivers, nrv2.1, nrv2.2 and nrv3 had equivalent staining levels and subcellular localizations,but that although detectable, nrv1 staining was significantly weaker.

Dorsal trunk and transverse connective lengths were measured as described by Beitel and Krasnow (Beitel and Krasnow,2000), except Metamorph software (Universal Imaging) was used for the morphometric analysis. Confocal images were captured using a Leica TCS SP2 maintained and supported by the Northwestern Biological Imaging Facility. To assess protein levels, heterozygotes and homozygous mutants were imaged at the same settings on the same slide in the same session. Adjustments performed in Photoshop were applied equally to all images. All embryos were stage 16 unless otherwise noted.

Texas Red-conjugated 10 kDa dextran was injected into embryos as described by Lamb et al. (Lamb et al.,1998). We believe that abnormally rapid leakage across tracheal epithelia indicates septate junction defects and not other tracheal defects as the disconnected tracheal segments of hnt mutant or homozygous balancers embryos do not leak dye (data not shown).

Double-stranded RNAs were generated by PCR amplification of templates with primers containing the T7 promoter sequence, in vitro transcription, and were injected into early embryos following standard protocols(Kennerdell and Carthew,1998). btl-Gal4 UAS-GFP homozygote embryos were used to visualize the tracheal system. RNAi-injected embryos were allowed to develop until stage 16-17, injected with 10 kDa dye and viewed with a Zeiss Axioplan2 microscope. Common exon regions were chosen for dsRNAs templates for all genes tested, except for nrv2.1 and nrv2.2 where unique 5′ exons were targeted. Specificity of isoform-targeted dsRNAs was demonstrated by injecting nrv2.1 or nrv2.2 dsRNA into btl-GAL4 UAS-nrv2.2 embryos. nrv2.2 dsRNA injections caused tracheal phenotypes and reduced levels of the overexpressed Nrv2.2 (5/5 embryos), while nrv2.1 dsRNA injections caused tracheal morphology defects but did not affect Nrv2.2 levels (9/9 embryos).

The sequence of nrv3, the predicted Na⁺/K⁺ATPase β subunit CG8633, was determined from cDNA GH12088 (GenBank Accession Number AY314744). UAS-nrv1, UAS-nrv2.1,UAS-nrv2.2 and UAS-nrv3 were constructed by inserting nrv1, nrv2.1, nrv2.2 (Sun and Salvaterra, 1995b) and nrv3 cDNAs into the pUAST vector(Brand and Perrimon, 1993). nrv2 alleles were sequenced by PCR amplification of genomic DNA from heterozygous flies, followed by cycle dye termination sequencing.

In previous work, we showed that embryos homozygous for the l(2)k04223 strain had tracheal tube-size regulation defects(Beitel and Krasnow, 2000). Using inverse PCR, we determined that the transposable element in this strain was inserted in an intron in the nervana2 (nrv2) locus,which encodes two alternatively spliced isoforms of a Na⁺/K⁺ ATPase β subunit(Sun and Salvaterra, 1995a; Sun and Salvaterra, 1995b). Two lines of evidence demonstrate that the tracheal phenotype of l(2)k04223 homozygotes results from disruption of the nrv2 locus. First, an independent transposable element insertion in the nrv2 locus,l(2)k13315, causes the same tracheal phenotypes as l(2)k04223 and fails to complement l(2)k04223 for tracheal phenotypes and viability (data not shown). Second, in an EMS non-complementation screen we isolated two additional mutations that fail to complement l(2)k04223 (see Materials and methods). Both of these mutants contain a single nonsense base change in the exons common to both nrv2 isoforms, and both have the same tracheal phenotypes as l(2)k04223. Together, these results show that a Na⁺/K⁺ATPase β subunit is required for tracheal tube-size control.

To better define the role of nrv2 in epithelial morphogenesis, we generated a putative nrv2 null allele, nrv2^nwu3,that deletes the first three common nrv2 exons which encode transmembrane and extracellular domains (see Materials and methods). In addition, Genova and Fehon (Genova and Fehon, 2003) provided a second putative null allele, nrv2^23B, that removes all nrv2 common exons. The phenotypes caused by nrv2^nwu3 or nrv2^23B do not become more severe in trans to a chromosomal deficiency known to delete nrv2, providing genetic evidence that these are null alleles.

In nrv2-null embryos, beginning at the time of tracheal tube expansion, multicellular tracheal tubes become increasingly abnormal so that most tube lengths are significantly increased and all tube diameters are irregular with expansions and constrictions along their lengths(Fig. 1B,F-I; Table 1). Defects are also present in regions of single cell tubes formed by autocellular junctions,particularly near the ends of the ganglionic branches where there are lumenal staining discontinuities (Fig. 1G).

Although the process of tracheal tube expansion is drastically disrupted in nrv2-null mutants, the earlier of phases of tracheal tube morphogenesis, including early tube-size regulation, are normal(Table 1). Furthermore, overall embryonic morphogenesis of nrv2-null mutant embryos also appears to be grossly normal as evidenced by the correct outgrowth of all tracheal branches to their target tissues and the absence of major patterning or morphogeneic defects as assessed using DIC optics(Fig. 1B,G). One possible explanation for the apparent specific morphogenic requirement of nrv2in tracheal tube expansion is that there is a maternal contribution of nrv2 that provides enough activity to support early morphogenic processes, but not enough to support tracheal tube expansion, which occurs late in embryonic development. However, in situ hybridization and microarray analysis did not reveal any significant maternal nrv2 transcript(data not shown and Berkeley Drosophila Genome Project) and embryos lacking both maternal and zygotic nrv2 are indistinguishable from nrv2 zygotic null embryos (Fig. 1H). Thus, nrv2 does not play a role in early epithelial formation or general embryonic morphogenesis, but instead appears to be specifically required for remodeling the length and diameter of tracheal tubes.

To test whether nrv2 functions as part of the Na⁺/K⁺ ATPase to control tracheal tube size, we examined the embryonic phenotypes of mutations in the major Na⁺/K⁺ ATPase α subunit locus, ATPα(Lebovitz et al., 1989; Sun et al., 1998; Palladino et al., 2003). The transposable element insertions ATPα^l(3)1278,ATPα^l(3)04694 and ATPα^l(3)07008 caused tracheal defects similar or identical to nrv2-null mutations, including tube length increases, diameter expansions and ganglionic branch discontinuities(Fig. 1M,N). The ATPα–null mutations ATPα^DTS1R1 and ATPα^DTS1R2(Palladino et al., 2003) also caused nrv2-like length and ganglionic branch defects, but caused only mild diameter defects (Fig. 1O). Although one would normally expect the ATPα-null mutations to cause more severe phenotypes than partial loss-of-function mutations, the hypomorphic ATPαmutations might cause strong nrv2-like phenotypes by producing inactive α subunits that could unproductively interact with Nrv2 and deplete the pool of Nrv2 available for productive interactions with other binding partners, such as α subunits expressed from the secondary Na⁺/K⁺ ATPase α subunit locus CG17923. However,despite some differences between the phenotypic effects of different ATPα mutations, the observation that both null and partial loss-of-function mutations cause nrv2-like tracheal tube-size defects demonstrates that the ATPα locus is required for tracheal tube-size control and suggests that the nrv2 β and ATPα α subunits function together in this process.

During our investigations of the role of the Na⁺/K⁺ATPase in tube-size control, J. Genova and R. Fehon reported that Na⁺/K⁺ ATPase mutants had salivary gland septate junction defects (personal communication). We therefore tested whether tracheal septate junction barrier function was defective in Na⁺/K⁺ ATPase mutants using the dye exclusion assay of Lamb et al. (Lamb et al.,1998), which tests the ability of an epithelium to exclude a 10 kDa dye. In wild-type embryos, tracheal septate junction barriers become functional and excluded dye starting at late stage 15 (e.g. Fig. 2a). However, the tracheal and salivary gland epithelia in nrv2 and ATPα mutants do not acquire the ability to regulate paracellular transport and cannot prevent the dye from inappropriately diffusing into the tracheal and salivary gland lumens (Fig. 2b,c).

To understand the nature of the septate junction defects in Na⁺/K⁺ ATPase mutants, we determined the subcellular distribution of septate junction and cell polarity components in stage 16 nrv2 and ATPα null mutants. We examined three ectodermally derived epithelia – trachea, epidermis and salivary glands– and found similar defects in all three tissues. The effects were most clearly seen in the large columnar cells of the salivary gland in which the septate junction occupies only a small region of the lateral surface of the cell (Fig. 3)(Tepass and Hartenstein,1994). By contrast, septate junctions occupy most of the lateral surface of tracheal cells, making visualization of mislocalized septate junction components more difficult. In both nrv2 and ATPα mutants, the septate junction components Coracle, Neurexin and Discs Large mislocalize along the lateral and sometimes the basal cell surfaces, rather than being tightly localized to the apicolateral septate junction (Fig. 3A-D and data not shown). In some cases, the levels of these proteins appeared to also be reduced (e.g. Fig. 3D). The septate junction component Fasciclin III was undetectable in nrv2 and ATPα mutant salivary glands and strongly reduced in trachea(Fig. 3F and data not shown).

Although every septate junction marker tested is severely affected by the Na⁺/K⁺ ATPase mutations, the effects appear to be specific for septate junctions since the localizations and levels of the adherens junction components E-cadherin (Shotgun) and β-catenin(Armadillo), and the apical determinant Discs Lost were unaffected in nrv2 and ATPα null mutants(Fig. 3F,H). Together, these data demonstrate that Na⁺/K⁺ ATPase mutations specifically disrupt septate junctions.

The function of the Na⁺/K⁺ ATPase in the septate junction could be direct as structural component, indirect through its generation of the electrochemical gradient, or both direct and indirect. To assess a possible direct role, we investigated the subcellular localization of the ATPα and Nrv2 proteins. In salivary glands, epidermis and trachea,ATPα staining predominantly colocalizes with Coracle and Neurexin at the apicolateral septate junction region, although some variable, low intensity ATPα staining of the basolateral cell surfaces that did not correlate with genetic background was also observed(Fig. 4A-C and data not shown). In nrv2 null mutants, ATPα levels are significantly reduced in the trachea and essentially absent in salivary glands(Fig. 4D). Where ATPαstaining is visible, it is no longer localized to the apicolateral septate junction region, but instead along the entire lateral surface. Thus, the ATPα localizes to the septate junction, and this localization is Nrv2 dependent.

Staining nrv2-null heterozygotes and homozygotes with an anti-β subunit antibody (Sun and Salvaterra, 1995a) revealed that Nrv2 accounts for the essentially all of anti-β subunit staining in ectodermally derived tissues, but the subcellular localization of Nrv2 could not effectively be assessed becauseβ subunit localization depended on fixation conditions and ranged from basolateral to perinuclear (see Materials and methods). As localization of the ATPα subunit is dependent on Nrv2(Fig. 4D-D″), and the ATPα and nrv2 mutants have the same phenotype, it seems likely that the functional Nrv2 protein will be localized to the septate junction.

Significantly, although Coracle and Neurexin are mislocalized and their levels may be reduced in ATPα and nrv2 mutants(Fig. 3B and data not shown),the localizations and levels of ATPα are unaltered in coraclemutants (Fig. 4E). Together,the above results suggest that the Na⁺/K⁺ ATPase is a structural component of the septate junction, and that it is required for stable formation of a complex containing Coracle and Neurexin.

To investigate the relationship between septate junctions and tube-size control, we determined the effects of mutations in other septate junction components on tracheal morphology. Null mutations in coracle and neurexin cause tracheal morphology defects that are essentially identical to those of nrv2 and the tube expansion mutants varicose, sinuous and convoluted(Fig. 5C-E,I,K; Table 1)(Beitel and Krasnow, 2000). Furthermore, mutations in gliotactin and neuroglian, two genes required for the blood-brain barrier formed by the septate junction(Auld et al., 1995; Dubreuil et al., 1996; Genova and Fehon, 2003; Schulte et al., 2003), also cause tracheal morphology defects similar to those caused by mutations in the tube-expansion genes (Fig. 5B,G; Table 1). Thus, the septate junction complex appears to be required for tracheal tube-size control.

To determine whether the paracellular barrier function of the septate junction is important for tracheal tube-size control, we examined the correlation between the tracheal and paracellular barrier phenotypes caused by tube expansion and septate junction mutants. In mutants with paracellular barrier defects, there is no relationship between paracellular defect and the type or severity of tracheal morphology defect. For example, although megatrachea and cystic mutant epithelia are both permeable to a 10 kDa dye (Fig. 2f,g),they have strikingly different types of tracheal phenotypes – megatrachea mutants have extremely long tubes with almost no diameter defects, whereas cystic mutants have dramatic diameter expansions and constrictions without any increase in length(Fig. 2F,G)(Beitel and Krasnow, 2000). Furthermore, a strain containing a weak coracle allele(Ward et al., 2001) has minimal tracheal morphological defects despite having epithelia permeable to a 10 kDa dye (Fig. 2E). If septate junction permeability defects caused the tracheal defects, then mutants with similar paracellular barrier defects should cause similar tracheal phenotypes. Because we find no correlation between the type or severity of tube-size and barrier defects, it appears that tracheal tube-size control by the septate junction does not depend on paracellular diffusion barriers.

Although the above results demonstrate that the nrv2 locus plays a crucial role in tracheal tube-size control and septate junction function, they do not address possible functional differences between the two Nrv2 protein isoforms, Nrv2.1 and Nrv2.2, because the available nrv2 mutations affect exons common to both (Fig. 6A; see Materials and methods). Nrv2.1 and Nrv2.2 share the same predicted transmembrane and extracellular domains, but differ dramatically in their ∼49 amino acid intracellular domains in which they are only 29%identical and 48% similar. To investigate possible functional differences between the isoforms, we used dsRNAs corresponding to either the nrv2common exons or to nrv2.1- or nrv2.2-specific exons in RNAi experiments to `knockdown' either both nrv2 transcripts or the nrv2.1 or nrv2.2 transcripts specifically (see Material and methods). We found that injection of any of nrv2 common, nrv2.1 or nrv2.2 dsRNAs caused the same tracheal tube length and diameter defects as nrv2 null mutations, and all three dsRNAs caused defects in septate junction barrier function(Fig. 6D-F). These results suggest that in normal development, both Nrv2 isoforms are required for tracheal tube-size control and septate junction function.

To determine whether the apparent requirement for both Nrv2.1 and Nrv2.2 resulted from inherent functional differences between the two isoforms, we used the two component expression system of Brand and Perrimon(Brand and Perrimon, 1993) to express nrv2.1 and nrv2.2 transgenes using a variety of drivers. Driving one or two copies of either UAS-nrv2.1 or UAS-nrv2.2 or both together specifically in the tracheal system of nrv2 homozygotes using one or two copies of the btl-Gal4 driver (Shiga et al., 1996)resulted in the same partial rescue of tracheal defects(Fig. 1J; Fig. 2, legend). Specifically,the diameter defects and ganglionic branch lumenal gaps were almost completely rescued, but the length and paracellular transport defects were not significantly rescued. Driving UAS-nrv2.1 or UAS-nrv2.2 with the e22c-Gal4 driver, which is expressed throughout the ectoderm and ectodermally derived tissue and begins expressing much earlier in development than btl-Gal4 (stage 5 versus stage 10), completely rescued all tracheal morphological and junctional barrier defects in the large majority of nrv2 homozygotes (Fig. 1J,K; Fig. 2,legend; Fig. 6H). These results suggest that for full rescue, nrv2 is required at a window of time extremely early in tracheal development, possibly before ectodermal cells are specified to be tracheal cells.

The above results also demonstrate that both nrv2.1 and nrv2.2 have tube-size control and paracellular transport activities. However, given that all combinations of Na⁺/K⁺ ATPaseα and β subunits form functional ion pumps(Lemas et al., 1994; Schmalzing et al., 1991; Schmalzing et al., 1997),these results also raise the possibility that neither nrv2 isoform has specific functions and that any β subunit could substitute for Nrv2. We therefore tested whether either of the two other DrosophilaNa⁺/K⁺ ATPase β subunits, nrv1(Sun and Salvaterra, 1995b) or nrv3 (CG8663), had tube-size or barrier activities. Driving UAS-nrv1 or UAS-nrv3 constructs with either btl-Gal4 or the e22c-Gal4 driver did not rescue any of the tracheal tube-size or septate junction barrier defects of nrv2 mutants (e.g. Fig. 6I,J; Fig. 2, legend), despite the fact that expression from the UAS-nrv2.1, UAS-nrv2.2 and UAS-nrv3 transgenes appeared equivalent as assessed by immunohistochemical staining (see Materials and methods). Consistent with these results, injection of dsRNA corresponding to nrv1 or nrv3 into wild-type embryos did not cause either the characteristic tube-expansion defects or septate junction barrier defects(Fig. 6G and data not shown). We therefore conclude that Nrv2.1 and Nrv2.2 both have specific tube-size and septate junction barrier activities not present in other Na⁺/K⁺ ATPase β subunits.

The nearly identical tracheal morphological defects in nrv2, coracle,varicose and convoluted (Fig. 5D,E,I,K; Table 1)suggested that these genes may act in a single linear genetic pathway. If so,then double mutant combinations of nrv2-null alleles and mutations in other tube-expansion genes should have the same tracheal phenotypes as a nrv2-null single mutant. This prediction is true for coracleand gliotactin (Fig. 5F,H), and is consistent with Coracle mislocalization in nrv2 mutants (Fig. 3B). However, the double mutant combination of a nrv2-null mutation and a convoluted mutation that does not cause septate junction barrier defects(Fig. 2h) causes more severe phenotypes than nrv2-null mutations(Fig. 5J). Similarly, the double mutant combination of a nrv2-null mutation and a varicose mutation that causes septate junction barrier defects(Fig. 2, legend) causes more severe phenotypes than nrv2-null mutants(Fig. 5L), indicating that these genes are unlikely to act in a simple linear pathway. The nrv2 cystic double mutant phenotype was indistinguishable from cystic(Fig. 5M,N), suggesting that cystic acts in parallel or downstream of nrv2.

Despite the importance of epithelial tubes to the function of many of organs, little is known about the pathways and proteins responsible for controlling multicellular tube size. Our results show that the Na⁺/K⁺ ATPase and septate junctions have previously unidentified roles in Drosophila tracheal tube-size control. They further provide the first in vivo evidence for functional differences between Na⁺/K⁺ ATPase isoforms as Nrv2 but not other Drosophila β Na⁺/K⁺ ATPase subunits have tube-size and septate junction activity. Below, we discuss models for how tracheal tube size is controlled by septate junctions and implications of our results for vertebrate tube-size control.

A simple explanation for the abnormal sizes of tracheal tubes in mutants having septate junction defects would be that ionic or hydrostatic disequilibria across the tracheal epithelium disrupts tracheal cell morphogenesis. If so, then all mutants with similar paracellular barrier defects should have equivalent tracheal morphologies. However, we found that barrier mutants had tracheal morphologies ranging from near wild type in the case of cor¹⁴* to diameter- or length-specific defects in cystic and megatrachea. These results support the conclusion that septate junction control of tube size is not dependent on regulation of paracellular diffusion.

The mutant phenotypes and genetic interactions among tracheal tube-expansion and septate junction mutants suggest there are at least two genetic pathways by which septate junctions regulate tracheal tube size(Fig. 7A). For example, nrv2 and coracle appear to act in the same genetic pathway as nrv2 and coracle null mutants have the same tracheal phenotypes as each other and as nrv2 coracle double null mutants(Fig. 7A, middle column). This genetic evidence is supported by our observations that the localization of Coracle to septate junctions is disrupted in nrv2 and ATPα mutants. By contrast, although nrv2-null and varicose mutants have the same tracheal phenotypes, nrv2 and varicose are unlikely to act in the same linear genetic pathway because nrv2 varicose double mutants have more severe tracheal phenotypes than nrv2-null mutants(Fig. 7A, left column). This result suggests that either varicose and nrv2 function in separate pathways to control tube size, or there is redundancy between the functions of these genes. We present the separate pathway model as a formal logic diagram in Fig. 7A and as a molecular model in Fig. 7B.

Additional genetic pathways may also be revealed by the differential effects of tube-expansion mutations on septate junction barrier integrity. For example, in contrast to nrv2 and other mutations, the existing convoluted mutations do not affect septate junction barrier integrity and therefore may define a size-control pathway that acts in parallel to septate junction pathways (Fig. 7A, right column, green `convoluted?'). Consistent with this proposal, the double mutant combination of a nrv2 null mutation and a convoluted mutation result in more severe tracheal morphology defects than either nrv2-null or convoluted mutations alone. Alternatively, genes such as convoluted may function in a branch of a septate junction pathway to link septate junctions to tube size control as shown in Fig. 7A (right column,red `conv?'). Although these models are necessarily incomplete, they offer testable predictions about gene interactions and subcellular localizations of uncharacterized gene products that should help define tube-size control and paracellular barrier pathways at the molecular level.

A central issue raised by our findings is the nature of the molecular functions(s) of the Na⁺/K⁺ ATPase and septate junctions in tube-size control. Although the Na⁺/K⁺ ATPase has been studied intensively for more than 40 years for its function as an ion pump (Chow and Forte, 1995; Blanco and Mercer, 1998), our data indicate that the tracheal tube-size function of the Na⁺/K⁺ ATPase is intimately associated with its role in septate junction function. Furthermore, as described above, the paracellular barrier and tube-size control functions of the septate junction are separable.

In one class of model that accounts for these observations, the role of the Na⁺/K⁺ ATPase is to organize septate junctions, which control tube size by an undetermined mechanism. The many functions of vertebrate tight junctions provide possible examples of non-barrier mechanisms by which septate junctions could control tube size. In particular, tight junctions organize polarized apical secretion mediated by the exocyst(O'Brien et al., 2002), bind cytoskeletal components such as ankyrin and fodrin(Fanning et al., 1999),contain potential signaling molecules such as the tyrosine kinases Src, Yes and protein kinase C (Fanning et al.,1999), and have recently been shown to regulate the activity of a Y-box transcription factor (Balda et al.,2003). In addition, both septate and tight junctions complexes contain proteins that organize epithelial cell apical/basal domains(Tepass et al., 2001).

Of the tube-size control models that do not invoke ion-transport functions of the Na⁺/K⁺ ATPase, models involving apicobasal domain organization are particularly attractive. Apical surface regulation is a common theme in tubulogenesis (Lubarsky and Krasnow, 2003; Buechner et al., 1999; O'Brien et al.,2002), and has been shown to play an important role in tube-size control in the Drosophila salivary gland(Myat and Andrew, 2002). Several observations support the possibility that septate junctions control tracheal tube size through the apical cell surface. First, the differential regulation of tracheal apical and basal cell surfaces suggests that tracheal tube size control is mediated at the apical cell surface(Beitel and Krasnow, 2000). Second, the increased tracheal tube lengths and diameters present in tube-expansion mutants necessitate an increased apical cell surface area. Given that the Dlg/Scrib/Lgl complex normally present in septate junctions has an early embryonic function to negatively regulate the extent of the apical membrane domain (Bilder et al.,2003; Tanentzapf and Tepass,2003), this complex could also act later to negatively regulate tracheal apical surface area.

In an alternative class of models that are not exclusive of the above possibilities, the ion-pump activity of the Na⁺/K⁺ATPase may directly or indirectly mediate tube-size control. For example,pharmacologically blocking Na⁺/K⁺ ATPase ion-transport activity leads to increased intracellular Ca²⁺ levels in some cell types (Ravens and Himmel,1999), and Ca²⁺ signaling abnormalities may be the molecular defect that causes the enlarged tubules of polycystic kidney disease(PKD) (Calvet, 2002; Hou et al., 2002; Yoder et al., 2002a; Yoder et al., 2002b). Another example is that the low intracellular Na⁺ level maintained by the Na⁺/K⁺ ATPase is required for formation of tight junctions and stress fibers in Madin-Darby canine kidney (MDCK) cells, an epithelial cell line that can form tubules in response to hepatocyte growth factor (Rajasekaran et al.,2001). Septate junction formation might also require low intracellular Na⁺ levels. Finally, disruption of the cellular Na⁺/K⁺ electrochemical gradient could impact secondary active transport of other solutes that may be important for proper tube-size regulation.

Although the exact biochemical roles of the Na⁺/K⁺ATPase and septate junctions in tube-size control are unclear, identification of these complexes as parts of a tube-size control mechanism is an important step towards further understanding these mechanisms at the molecular level.

The Na⁺/K⁺ ATPase has been implicated in vertebrate tube-size control by the abnormal subcellular localization of the Na⁺/K⁺ ATPase in the inappropriately expanded tubules in individuals with PKD and in several animal models of cystic kidney diseases(Avner et al., 1992; Carone et al., 1995; Ogborn et al., 1995; Wilson et al., 2000; Wilson et al., 1991). However,it has not yet been determined whether this mislocalization contributes to the progression of cystic diseases or whether it is merely a secondary effect of other cellular defects. Our finding that the Na⁺/K⁺ATPase is required for normal tube-size control in the Drosophilatracheal system suggests that the vertebrate Na⁺/K⁺ATPase may play an important role in maintaining the normal size of kidney and other epithelial and endothelial tubes. Ultimately, a molecular understanding of the tube-size control mechanisms should allow development of new strategies for preventing and treating PKD and other diseases resulting from epithelial and endothelial tube defects.

We are grateful to J. Genova and R. Fehon for sharing unpublished data and for helpful discussions, to R. Holmgren for general advice, and to R. Carthew and H. Folsch for comments on the manuscript. Thanks to the following investigators and centers for antibodies and fly stocks: V. Auld, S. Baumgartner, H. Bellen, M. Bhat, P. Bryant, R. Fehon, M. Krasnow, M. Metzstein, H. Oda, M. Palladino, the Developmental Studies Hybridoma Bank, and the Bloomington Stock Center. Additional thanks to V. Wu for the initial identification of nrv2, P. Xu for the UAS-nrv1 and UAS-nrv2 lines, M. Yu, E. Klostermann, M. Kelaita and A. Graff for technical support, and W. Russin of the Northwestern Biological Imaging Facility. This work was supported by American Cancer Society IRG grant#93-037-06 and NSF Career Award IBN-0133411 to G.J.B., and by NIH CMBD training grant #T32 GM008061 support for S.M.P. G.J.B. is a recipient of a Burroughs Welcome Fund Career Award in the Biomedical Sciences.