PAT-related amino acid transporters regulate growth via a novel mechanism that does not require bulk transport of amino acids

Cells and tissues in multicellular organisms typically grow in a coordinated fashion at a rate that can be modified by nutritional status. Recent studies in Drosophila have highlighted a crucial role for the insulin receptor (InR) pathway in regulating the endocrine response to nutrient levels (reviewed by Stocker and Hafen, 2000; Goberdhan and Wilson, 2003a). This evolutionarily conserved system not only affects cell, tissue and organismal growth, but also modulates multiple metabolic functions, developmental timing, fertility and longevity. Elevated InR signalling also plays a central role in tumourigenesis; mutations in PTEN, a negative regulator of the pathway, occur at remarkably high frequency in several different human cancers (reviewed by Goberdhan and Wilson,2003b).

Eukaryotes possess a second and more ancient nutrition-regulated system,involving the kinase TOR (target of rapamycin)(Jacinto and Hall, 2003),which is responsive to local nutrient levels, particularly amino acids. In higher eukaryotes TOR controls downstream translational regulators, including p70-S6 kinase (S6K) and 4E-binding protein (4EBP) and is essential for normal cell-autonomous growth (Zhang et al.,2000; Oldham et al.,2000). Studies in Drosophila have recently linked the endocrine InR signalling system with TOR (reviewed by Goberdhan and Wilson, 2003a; Aspuria and Tamanoi, 2004). It is proposed that Akt, a key InR-regulated kinase, phosphorylates the tumour suppressor protein, Tuberous sclerosis complex 2 (Tsc2), dissociating the Tsc1/Tsc2 complex. This activates the G protein Rheb, which positively regulates TOR.

TOR activity can be affected by changes in intracellular amino acid levels(Christie et al., 2002; Beugnet et al., 2003). Experiments in cell culture indicate that this response is modulated by amino acid transporters, but no amino acid transporter has yet been shown to regulate growth of non-endocrine cells in vivo. We have therefore genetically screened for such transporters. Although a number of different classes of mammalian amino acid transporter have been identified (see Hediger, 2004), many still have poorly characterised in vivo functions. In Drosophila, members of two transporter classes have been implicated in growth: minidiscsencodes a component of the heterodimeric family of transporters(Martin et al., 2000); and slimfast (slif)(Colombani et al., 2003)encodes a cationic amino acid transporter. These genes are required for normal growth, but they both primarily act in the fat body, an amino acid-sensitive,growth-regulatory endocrine organ with functional similarities to the mammalian liver and white adipose tissue (Britton et al., 1998).

Here, we show that two genes encoding proteins related to a third class of transporters of previously unknown biological function, the proton-assisted transporter (PAT or SLC36) family(Bermingham and Pennington,2004), specifically modulate tissue-autonomous growth in multiple non-endocrine tissues in vivo. They also genetically interact with TOR and other InR signalling components, indicating a direct or indirect regulatory link with this signalling system. One of these transporter genes, pathetic (path), is expressed in a wide range of tissues and is essential for both endocrine and local growth regulation. PATH has unusual functional properties, establishing it as the founder member of a new class of low capacity, very high-affinity PAT-related transporters, which control growth via a novel amino acid sensing mechanism. As the functions of path can be substituted by another Drosophila PAT-related transporter whose properties are much more similar to mammalian PATs, we propose that this gene family has a conserved and unique role in regulating TOR-dependent growth and the response to insulin-like molecules in development and disease.

The following fly strains were used: Rheb^AV4(Patel et al., 2003), UAS-Dp110^D954A, UAS-Dp110(Leevers et al., 1996), UAS-Dakt1 (Staveley et al.,1998), UAS-Tor, UAS-Tor^TED(Hennig and Neufeld, 2002), Tsc2¹⁹², Tsc1²⁹(Gao and Pan, 2001), UAS-S6K (Zhang et al.,2000), UAS-slif(Colombani et al., 2003) and GS lines (Toba et al., 1999). GS inserts near path (13857, 11111) and CG1139 (10666, 7120)produce similar phenotypes (see Fig. 2). GS inserts near CG1139 lie on the other side of a gene (CG11815), which is transcribed in the opposite direction;overexpression of a UAS-CG1139 construct confirmed that growth effects were specifically mediated by CG1139. Other GS lines tested included 9983, 10651, 11951, 13265 and 13758. Flies were reared on standard cornmeal agar food.

Ommatidial sizes were calculated from stereomicroscope images (Leica Wild M3Z microscope and Axiovision software)(Radimerski et al., 2002b);the area of a 3 ×3 set of ommatidia in the centre of each eye was measured. Wing area data were collected using the same equipment, while wing cells were imaged using a Zeiss Axioplan 2 microscope. The number of wing cells in a rectangular 47250 μm² box next to the posterior cross vein was counted, and scaled up to give total cell number in the posterior compartment. Data are presented as mean±s.e.m. for at least six measurements per genotype and using female flies, unless otherwise stated.

Putative full-length path-RA (RH24992; almost all 80 available ESTs represent the RA form) and CG1139 (LP06969) cDNAs were subcloned into transformation vector pUAST (Brand and Perrimon, 1993) using the restriction sites flanking the cDNA inserts (EcoRI/XbaI and EcoRI respectively). Transgenics were produced using standard procedures. At least four independent lines were established and tested. In situ hybridisation was based on work by Tautz and Pfeifle (Tautz and Pfeifle,1989), using a PCR-amplified path template cDNA. cRNA was synthesised by in vitro transcription (mMessage mMachine, Ambion).

Healthy looking Xenopus laevis oocytes (stage V and VI) were obtained as described (Meredith et al.,1998), and maintained at 18°C in modified Barth's medium. Transport measurements were performed at least 72 hours after microinjection of oocytes with 27.6 nl cRNA (1 μg/μl) with daily medium changes. Uptake assays with [³H]-alanine (59 Ci/mmole, Amersham Biosciences) were performed essentially as described, with five oocytes per data point(Meredith et al., 1998). Membrane potential was measured using a Dagan CA-1B amplifier. Oocytes impaled with a AgCl/3M KCl-filled glass microelectrode (resistance approximately 0.5 MΩ) were superfused with uptake medium containing 2 mM amino acid or uptake medium alone. Intracellular amino acid concentrations were calculated based on the work of Christie et al.(Christie et al., 2002).

Western blotting followed standard procedures. Oocytes were lysed in 3μl/oocyte lysis buffer and loaded with 1:1 (v:v) loading buffer. PVDF membranes were probed with 1:1000 (v:v) anti-phospho-p70 S6K (Thr389) antibody(Signaling Technologies, NEB) or 1:200 (v:v) anti-total S6K (Santa Cruz),followed by 1:2000 goat anti-rabbit HRP-conjugated secondary antibody (Santa Cruz) and the membrane developed using the ECL system (Amersham).

We screened for amino acid transporters that play a crucial role in TOR regulation in growing tissues. Our hypothesis was that these transporters might, like other characterised components of the InR/TOR signalling network(e.g. Leevers et al., 1996; Goberdhan et al., 1999; Goberdhan and Wilson, 2003a),affect growth when overexpressed in the eye or wing. Using the GAL4/UAS system(Brand and Perrimon, 1993), we attempted to overexpress genes predicted to encode transporters, by employing GS insertions near these genes (Toba et al., 1999) and UAS constructs.

Heterodimeric and cationic transporter genes produced little or no effect on growth when expressed in differentiating ommatidia using GMR-GAL4(e.g. slif; Fig. 1B). By contrast, overexpression of GS elements upstream of CG3424(subsequently renamed path) and CG1139, which encode proteins similar to the mammalian PATs (>30% identity; see Fig. S1A in the supplementary material), increased both ommatidial and overall eye size(Fig. 1C,D, respectively). When two GS insertions in path (Fig. 1E,L) or in CG1139(Fig. 1F,M) were co-expressed,the resulting eyes bulged more from the head capsule and the ommatidial array was mildly disorganised, indicating a more extreme dose-dependent overgrowth phenotype. Similar growth effects were also produced by UAS-path(e.g. Fig. 2K-M) and UAS-CG1139 (Fig. 1H)transgenes, confirming that these genes caused the GS-associated phenotypes.

As previously reported (Hennig and Neufeld, 2002), expression of UAS-Tor under GMR-GAL4 control produced no obvious eye phenotype(Fig. 1G). However, when co-overexpressed with either path(Fig. 1I) or CG1139(Fig. 1J), UAS-Torsignificantly enhanced the overgrowth phenotype induced by these transporters alone (Fig. 1C,Drespectively).

CG1139 and path also modulated growth in the developing wing, but their effects differed. Although slif had no significant effect with the MS1096-GAL4 driver(Fig. 2B), which is expressed throughout the developing adult wing, but preferentially on the dorsal surface, in animals overexpressing CG1139 GS insertions, overall wing size was increased (Fig. 2D)and excess growth on the dorsal surface caused the wing to turn down. However,the effects of CG1139 on wing growth were not simple. Surprisingly,overexpression of the UAS-CG1139 transgene, which has a more potent growth-stimulating effect in the eye than CG1139 GS inserts(Fig. 1H), reduced wing growth and produced upturned wings (Fig. 2F), suggesting that this gene might inhibit growth in some developmental contexts, if sufficiently overexpressed.

This effect was reminiscent of UAS-Tor, which, despite the well-documented growth-promoting functions of Tor, reduces growth when overexpressed in the wing (Fig. 2E) (Hennig and Neufeld,2002). It has been proposed that TOR overexpression interferes with proper formation of a TOR-containing multiprotein complex in the wing to inhibit growth. To test whether a similar dominant-negative mechanism might explain the growth inhibitory effects of CG1139, we overexpressed two CG1139 GS insertions. Although both these insertions promoted growth when expressed independently, in combination they significantly reduced growth(Fig. 2I). In addition,co-expression of a CG1139 GS insertion with UAS-CG1139enhanced the growth inhibitory effect of the latter construct(Fig. 2J).

Co-overexpression of CG1139 GS inserts with Tor greatly exaggerated the reduced growth phenotype produced by Tor alone(Fig. 2H), suggesting that these molecules might exert their dominant-negative effects on interacting complexes. Overexpression of path GS insertions or UAS-pathreduced wing size in the MS1096-GAL4 assay (e.g. Fig. 2C), suggesting that path also has dominant-negative effects in the wing. In addition,like CG1139, path overexpression strongly enhanced the dominant-negative effects of Tor(Fig. 2G).

To quantify growth changes in the wing, the transporter genes were overexpressed in the posterior wing compartment with en-GAL4. Effects on overall wing growth mirrored those observed with the MS1096-GAL4driver (Fig. 2K) and in most cases total cell number was significantly reduced(Fig. 2M). Unexpectedly,however, path and CG1139 overexpression produced an increase in cell size, even when overall wing area was reduced(Fig. 2L). For each transgenic line tested, the increase was approximately inversely related to the reduction in cell number. Again, we tested whether these effects might be comparable with those of TOR. It has been reported that TOR overexpression in wing imaginal discs reduces the size of proliferating cells, as does a dominant-negative form of TOR, TOR-TED(Hennig and Neufeld, 2002). However, they did not analyse the size of cells in the adult wing. In fact,posterior wing cells were enlarged in en-GAL4 UAS-Tor adult animals,despite the reduction in overall compartment size, mimicking the effects of PAT-related transporters (Fig. 2L). The effects of overexpressing PATH, CG1139 and TOR on cell growth and proliferation therefore vary in parallel, depending on developmental context, with their growth inhibitory actions in the wing most likely being caused by dominant-negative effects during the proliferative stage of development.

To test whether PAT-related transporters normally promote growth in flies,we searched for specific mutations in either CG1139 or path. We identified a transposable element insertion, path^KG06640, located within the first intron of the larger path transcription unit, RA (see Fig. 3A). The KG transposon contains gene silencing elements, which increase the likelihood that insertions will affect gene expression(Roseman et al., 1995). Flies homozygous for path^KG06640 were markedly smaller than wild-type animals (Fig. 3B) and∼40% lighter in weight (Fig. 3C). Wing size was decreased by nearly 20%(Fig. 3D), with significant reductions in both cell number and size (P<0.001). Mutant eyes contained small ommatidia (compare Fig. 3E with 3F). Like many viable mutants in which InR signalling is disrupted (e.g. Böhni et al.,1999), path^KG06640 homozygotes were developmentally delayed, eclosing roughly 3 days later than normal. Homozygous females also laid no eggs.

Animals transheterozygous for the path^KG06640chromosome and a large deficiency uncovering the path gene, Df(3L)AC1, were even smaller than path^KG06640homozygotes (Fig. 3B-D,G),suggesting path^KG06640 is not a null allele. Transheterozygotes were also developmentally delayed and female sterile. Transposase-mediated excision of the path^KG06640 element reverted all recessive phenotypes in 90% of cases, linking these defects to the KG insertion. Expression of a UAS-path transgene under the control of a constitutive GAL4 driver, arm-GAL4, also significantly rescued the growth, developmental delay and female sterility mutant phenotypes(see Fig. 3C,D,H). Rescued flies were slightly smaller than controls, but this was probably caused by constitutive overexpression of path, as overexpressed pathslightly reduced growth in otherwise normal flies(Fig. 3C). CG1139^GS10666 also rescued path^KG06640mutant phenotypes (Fig. 3C,D). Finally, reducing the copy number of both the Tsc1 and Tsc2genes, which normally antagonise TOR signalling, also partially rescued the path mutant phenotypes. Rescued flies were not developmentally delayed, the path growth phenotype was strongly suppressed(Fig. 3C,D) and about 10% of females laid a small number of eggs, some of which developed into adult flies. path^KG06640 homozygous larvae and path^KG06640/Df(3L)AC1 transheterozygotes were small and took roughly 3 days longer than wild-type larvae to pupate(Fig. 3I). However, in contrast to several other developmentally delayed mutants with reduced InR signalling,imaginal discs from path mutant wandering third instar larvae were not reduced in size (Fig. 3J,K; Fig. 4E-H). pathmutant discs were small when compared with similarly aged normal discs, but the extended mutant larval period allowed discs to grow to roughly normal size by pupariation. By contrast, the length of pupal development was not noticeably affected in path mutants, which probably explains why normal-sized adult structures are not formed. Unlike the larval endoreplicative tissues (ERTs) of viable Rheb, Tor and slifmutants, which are greatly reduced in size(Oldham et al., 2000; Stocker et al., 2003; Colombani et al., 2003), ERTs such as the salivary gland and the fat body were only modestly reduced in path mutants and fat body cells had non-starved morphology (data not shown).

Overall, we conclude that the recessive phenotypes associated with the path^KG06640 mutant chromosome are caused by insertional mutation in path and are modified by at least one InR signalling complex: Tsc1/Tsc2. Based on the mutant growth phenotypes, we named the gene pathetic (path).

path is highly represented in the EST database. RNA in situ hybridisation performed on embryonic and late larval stages revealed that it is expressed in most tissues. However, transcript levels were highly dynamic in embryogenesis with surges of expression in many structures, including muscle primordia, salivary glands, proventriculus (and other parts of the gut), trachea and gonads (see Fig. 4A-C). Expression was strongly reduced in homozygous path^KG06640 mutants(Fig. 4D). Larval imaginal discs also expressed path in all or most cells. Expression was particularly strong in the pouch and hinge regions of the wing disc(Fig. 4E) and in the morphogenetic furrow of the eye disc (Fig. 4G). Transcript levels were reduced in path^KG06640 homozygotes(Fig. 4F,H), but the reduction was less severe than in embryos, particularly in the eye.

The relatively modest reduction in path expression in mutant imaginal discs suggested that the significant growth defects associated with this mutant might not be explained by a cell-autonomous reduction in PATH activity. Indeed, when the eyFLP/FRT approach(Newsome et al., 2000) was used to generate eyes in which many ommatidia were derived from path^KG06640 homozygous cells, no significant growth disadvantage was observed in these cells. Even when almost all the eye was mutant for path, because the other clonal material produced was homozygous for a cell lethal chromosome, the resulting eyes were indistinguishable in size from normal eyes (compare Fig. 4J with 4I and 4K).

However, path^KG06640 could produce tissue-autonomous effects in the eye in specific genetic backgrounds. For example, in our hands,it was difficult to produce viable animals with eyes that mainly consisted of homozygous Tsc2 clones, using eyFLP-induced recombination with a cell lethal chromosome. Less than 1% of females and no males of the appropriate genotype survived through pupal stages to adulthood. When clonal material in the eye was mutant for both Tsc2 and path^KG06640 in an otherwise heterozygous animal, the lethal effects of Tsc2 were suppressed, and about 10% of mutant animals, both males and females, eclosed. Their eyes were smaller and bulged less than Tsc2 mutant eyes (compare Fig. 4M,N), revealing a tissue-autonomous effect of path on the Tsc2 phenotype.

Our genetic data showed a role for PAT-related transporters in growth regulation and indicated that, like other amino acid transporters, this effect might involve TOR. However, they did not reveal whether these transporters directly modulate TOR signalling or have an indirect effect via a parallel growth regulatory pathway. To investigate this issue further, we developed an in vivo genetic assay to detect changes in TOR/S6K signalling activity in a tissue where PAT-related transporters stimulate growth. The forkhead transcription factor FOXO is a target of InR signalling in higher eukaryotes,including flies (Jünger et al.,2003; Puig et al.,2003). FOXO normally inhibits cell proliferation and stimulates apoptosis, but is inactivated when phosphorylated by InR-regulated Akt. Overexpression of a GS insertion in the first intron of foxo,foxo^GS9928, with GMR-GAL4 produced distinctive ommatidial loss particularly in the posteroventral part of the eye(Fig. 5B)(Jünger et al.,2003).

This phenotype was strongly enhanced by co-expressing inhibitory molecules in the InR pathway, such as a dominant-negative form of Dp110(Fig. 5E)(Leevers et al., 1996) and completely suppressed by activating components, such as the Drosophila PI3-kinase, Dp110 and Akt1(Staveley et al., 1998)(Fig. 5C,F). By contrast, other positive growth regulators, which are believed to act independently of the InR pathway, did not produce similar suppression effects (see Fig. S2 in the supplementary material).

Radimerski et al. (Radimerski et al.,2002a; Radimerski et al.,2002b) have provided biochemical and genetic evidence that TOR signalling can negatively feed back to the InR signalling cascade, thus inhibiting Akt1 activity. We therefore reasoned that co-overexpression of TOR signalling components with GMR-GAL4 foxo^GS9928 might enhance the growth inhibitory effects of FOXO. Indeed, S6K, which normally promotes modest overgrowth with GMR-GAL4(Fig. 3M)(Zhang et al., 2000), strongly enhanced the GMR-GAL4 foxo^GS9928 reduced eye phenotype(Fig. 3N), mirroring the effects of dominant-negative Dp110 (Fig. 5E). A UAS-containing insertion in Rheb,Rheb^AV4 (Patel et al.,2003), which by itself produces a greatly enlarged eye(Fig. 5G,J)(Patel et al., 2003), enhanced FOXO activity even more strongly in the eye, leading to loss of virtually all ommatidia in females (Fig. 5H)and pupal lethality in males with massive tissue degeneration in the head of late pupae. As expected, these phenotypes were suppressed by Akt1 overexpression (Fig. 5I,L), but not by co-expression of other UAS constructs (e.g. UAS-GFP,UAS-lacZ). Thus, growth-stimulating TOR signalling components act entirely differently to all other positive growth regulators in this assay,reducing eye size when co-expressed with FOXO.

Co-expression of foxo with either CG1139 or pathalso enhanced the foxo reduced eye phenotype(Fig. 5P,S), an effect completely suppressed by Akt1 (Fig. 5Q,T), consistent with a role for these transporters in modulating TOR signalling. Interestingly, although TOR alone did not affect the GMR-GAL4 foxo^GS9928 phenotype(Fig. 5X), it did enhance the effects of a weakly expressed UAS-path transgene (compare Fig. 5V with 5Y). Thus, TOR and the PAT-related transporters interact genetically in multiple developmental assays, and produce similar phenotypes when overexpressed in multiple tissues and in an in vivo assay that is specific for the TOR signalling pathway.

To examine the transport properties of PATH and CG1139, we overexpressed each in Xenopus oocytes. This heterologous system has been extensively employed in the characterisation of many mammalian amino acid transporters, including PATs (e.g. Boll et al., 2002; Chen et al.,2003b). CG1139 behaved very similarly to mammalian PATs in radiolabelled amino acid influx (Fig. 6A,B) and inhibition (Fig. 6C,E) assays, and in electrophysiological studies(Fig. 6F,G). Like human PAT1 and PAT2, CG1139 specifically transported alanine, glycine and proline in an electrogenic and proton-stimulated fashion. Not only was influx of radiolabelled amino acids enhanced by extracellular acidification(Fig. 6A,B), but when compared with results obtained at pH 7.4, uptake of alanine, glycine and proline produced a larger plasma membrane depolarisation at pH 5.5, with reversal of membrane polarity (Fig. 6F,G). CG1139 and PAT1 have similar affinity for alanine [K_m of 1.2±0.2 mM (R²=0.99; Fig. 6E) versus an IC₅₀ of 1.7±0.23 mM for PAT1(Chen et al., 2003a)], glycine(estimated IC₅₀=3 mM, 2.3 mM for PAT1; Fig. 6C) and proline (estimated IC₅₀=8 mM, 2.0 mM for PAT1).

By contrast, PATH behaved very differently in all these assays. Initial influx experiments using radioactive alanine, under the same conditions as for CG1139, revealed little if any change in influx at different amino acid concentrations. However, when non-radiolabelled competitor was greatly reduced in concentration, significant amino acid influx was observed(Fig. 6B). The affinity of PATH for alanine was found to be 100- to 3000-fold higher than mammalian PATs (Km of 2.7±1.0 μM, R²=0.95; Fig. 6E). In addition, the transport activity of PATH was higher at pH 7.4 than at pH 5.5(Fig. 6B), unlike all other characterised PAT-related transporters. Inhibition studies revealed that although PATH also had a high affinity for glycine, it did not for proline(Fig. 6D).

In vivo genetic studies described above show that PATH and CG1139 have similar effects on growth and in path rescue experiments. However, in Xenopus oocytes, PATH has a transport capacity about 400 times lower than CG1139. It would thus be expected to increase overall intracellular levels of alanine by less than 0.2% during a 30 minute incubation. We tested whether despite this low transport capacity, extracellular alanine could activate the downstream target of TOR, S6K, in PATH-expressing Xenopus oocytes. S6K activation was detected in oocytes microinjected with leucine or alanine using an antibody that specifically crossreacts with the phosphorylated activated form of this molecule(Fig. 6H; data not shown)(Christie et al., 2002). Extracellular alanine had no effect on control oocytes, but did stimulate S6K signalling in oocytes expressing PATH (Fig. 6H). Thus, PATH can activate growth and a TOR target, S6K, in the absence of significant overall change in intracellular amino acid concentration.

Using in vivo genetic analysis combined with transport studies in the heterologous Xenopus oocyte system, we have shown a novel and unexpected function for the PAT-related transporters in growth control and multiple genetic interactions with the TOR and InR signalling cascades. CG1139, a gene encoding a fly transporter with similar properties to mammalian PATs, modulates growth in multiple tissues. PATH, which has very different transport properties, produces surprisingly similar effects in vivo,suggesting that despite their differences, members of this family of sequence-related transporters have a conserved role to play in growth regulation (Fig. 7 and later discussion).

Four mammalian PAT family members have been cloned(Boll et al., 2002; Chen et al., 2003a; Chen et al., 2003b). The first identified member was christened LYAAT1, owing to its lysosomal localisation in rat brain (Sagne et al.,2001), but it is now clear that these transporters are also found at the cell surface. The human version of LYAAT1, PAT1, resides at the apical brush border membrane of intestinal cells(Chen et al., 2003a) and mouse PAT2 is predominantly localised to the endoplasmic reticulum and recycling endosomes in neurons (Rubio-Aliaga et al.,2004). The finding that PAT-related transporters modulate growth reveals an important new biological role for this class of molecules.

path and CG1139 phenocopy the growth effects of Tor in the eye and wing, and strongly interact with Tor in multiple genetic assays. In addition, like positive growth regulatory components of the TOR signalling pathway, the transporters reduce eye size in the FOXO interaction assay. Members of other classes of transporter, including the cationic transporter Slif, did not produce this range of effects. Our data are therefore consistent with a model in which both TOR(Fang et al., 2001; Wedaman et al., 2003) and the PAT-related transporters act within one or more membrane-associated molecular complexes required for TOR activation. We cannot eliminate the possibility that PATs act in a parallel growth regulatory pathway, but our data indicate that one function of this pathway would be to alter the sensitivity of the cell to InR activation (Fig. 7).

Under normal conditions, the net result of TOR activation is overgrowth,but when FOXO, a direct target of Akt1 that is not part of the Akt1/Tsc/Rheb/TOR link (Fig. 7), is overexpressed in the differentiating eye, the TOR-induced reduction in Akt1 signalling increases cell death. Currently, the feedback mechanism involved has not been fully elucidated, although studies in mammals,where a similar phenomenon occurs, suggest that the phosphorylation state of IRS-1, an adaptor molecule for the InR, is involved(Jaeschke et al., 2002; Um et al., 2004; Harrington et al., 2005). One consequence of Akt repression by TOR signalling is to make cells more insulin resistant, a defining cellular and genetic defect in Type II diabetes. As PAT-related transporters enhance the effects of FOXO, partial cell surface inhibition of PAT activity might prove beneficial in individuals with Type II diabetes.

Consistent with our proposal that path functions by modulating the response to InR/TOR signalling, the effects of reduced path on growth, developmental timing and female fertility broadly mirror those seen in InR signalling mutants. path is still transcribed, albeit at reduced levels, in path^KG06640 mutant imaginal discs, and clonal analysis reveals that at least in the eye, an altered external growth signal is primarily responsible for the path mutant phenotype. As this phenotype is strongly suppressed by a combination of heterozygous Tsc1 and Tsc2 mutations, this global function of path probably involves modulation of InR/TOR signalling. Toralso affects endocrine as well as cell-autonomous growth regulation, at least in part by modulating fat body signalling(Colombani et al., 2003). However, the endocrine effects of path are unlikely to involve the fat body, as this organ is relatively unaffected in path mutants and path is weakly expressed in this tissue. These differences in tissue expression may explain why, unlike Tor mutants, ERTs and primordia of adult structures ultimately grow at a slower rate to roughly normal size in path larvae.

The partial suppression of a Tsc2 clonal overgrowth phenotype in the eye by the path^KG06640 mutation indicates that this allele does have tissue- and presumably cell-autonomous effects on growth in non-endocrine tissues in sensitised InR signalling backgrounds. As we have recently identified another broadly expressed PAT-related transporter with growth regulatory activity, the relatively modest effects of the pathmutation in imaginal discs may also be partly explained by functional redundancy in this family.

Analysis of the transport characteristics of CG1139 in the heterologous Xenopus oocyte system revealed remarkable similarities to mammalian PATs. By contrast, PATH displayed many novel properties, despite sharing similar levels of sequence identity with mammalian PATs (see Fig. S1A,B in the supplementary material). These include a mechanism inhibited by extracellular acidification, altered substrate specificity, remarkably high substrate affinity and low transport capacity. Although these experiments were not performed in vivo, to our knowledge there is no reported example where the functional properties of an amino acid transporter in the Xenopusoocyte system are not very similar to its properties in vivo, and it seems particularly unlikely that the substrate affinity of PATH would specifically increase by ∼500-fold when synthesised in the oocyte.

Taken together with our primary finding that PATH and CG1139 are important regulators of growth in vivo, these data have at least three major implications. First, amino acid transporters have previously been shown to modulate TOR by changing intracellular amino acid concentrations(Christie et al., 2002; Beugnet et al., 2003), although observations in whole tissues suggest sensing of extracellular amino acid levels may be more important in regulating protein metabolism in vivo(Mortimore et al., 1992; Bohé et al., 2003). PATH can promote growth in vivo and in the heterologous oocyte system it can modestly activate S6K, despite its extremely limited transport capacity,which, during the time course of our Xenopus assays, is predicted to change bulk intracellular amino acid concentrations by less than 0.2%. PATH must therefore control growth via an alternative and novel mechanism, either by modulating local amino acid concentrations in a restricted compartment near the transporter, or through transport-independent amino acid sensing(Fig. 7). Based on genetic and biochemical experiments, our current hypothesis is that this mechanism involves TOR activation and is likely to be conserved for other PAT family members, given the ability of CG1139 to largely substitute for path function in vivo. Interestingly, if PATs regulate growth via a transport-independent mechanism, the recent discovery that short chain fatty acids are also substrates for human PAT1 and PAT2(Foltz et al., 2004) opens up the possibility that these molecules could act as novel growth regulators.

Second, the fact that PATH and CG1139, a transporter with 500-fold lower affinity, can have similar effects on growth in vivo, strongly suggests that PATH (K_m ∼3 μM) is saturated in developing tissues. Indeed,extracellular amino acid concentrations are estimated to be in the millimolar range in flies (e.g. Pierce et al.,1999). Thus, the growth-promoting ability of PATH is effectively determined by its pattern and levels of expression and not by the extracellular amino acid concentration. Of course, it is possible that in some physiological conditions or within certain tissues, the high affinity of PATH will be crucial, but to date, rescue experiments of path^KG06640 mutant animals with both path and CG1139 have revealed no obvious differences between these molecules.

If PATH couples to TOR signalling, its properties ensure that all tissues that express this transporter will normally activate TOR at least at a basal level. Under these conditions, endocrine signalling by insulin-like molecules will therefore partially dominate over local nutrient signals in determining cell growth rates, consistent with previous observations(Britton et al., 2002)(Fig. 7). Interestingly, other insect genomes encode proteins with high sequence homology to PATH (XM 308237 in the mosquito Anopheles gambiae; XP 396451 in the Western honeybee Apis mellifera; see Fig. S1A,B in the supplementary material), and presumably employ similar mechanisms. Although there is no evidence at present for vertebrate sequence homologues, it is noteworthy that standard transport assays, which initially suggested PATH was an orphan transporter, have also failed to identify substrates for PAT3 and the ubiquitously expressed PAT4.

Finally, PAT-related transporters have been identified at the cell surface and in lysosomes. Subcellular shuttling of transporters has a precedent in yeast, where TOR modulates movement of at least one amino acid permease between the cell surface and the vacuole(Chen and Kaiser, 2003). Interestingly, yeast Tsc1, Tsc2 and Rheb have also been implicated in the regulation of amino acid transporters(Urano et al., 2000; van Slegtenhorst et al.,2004). However, contrary to expectations from multicellular organisms, where the Tsc complex inhibits TOR signalling, yeast Tsc upregulates and Rheb downregulates one or more amino acid permeases. Perhaps in higher eukaryotes Rheb positively regulates some transporters and negatively regulates others (Saucedo et al., 2003). Alternatively, endosomal and lysosomal targeting of PAT-related transporters upon Rheb activation could increase activity,although in this model, activation of PATH might only occur if it acts via a transport-independent mechanism, as its transport activity appears to be inhibited by acidification. Resolution of this issue will require a biochemical and cell biological analysis of the interactions between upstream InR signalling components, TOR complexes and PAT-related transporters.

We thank J. Lengyel, S. Leevers, E. Hafen, B. Edgar, T. Neufeld, D. Pan, A. Brand, A. Manoukian, H. Bellen and P. Léopold for fly stocks; S. North and Z. Deussen for technical assistance; and M. van den Heuvel for helpful comments. Other fly stocks were provided by the Bloomington Stock Center and the Kyoto Drosophila Genetic Resource Center. We are particularly grateful to T. Aigaki and H. Bellen (P-element Gene Disruption Project) for sending flies prior to distribution to stock centres. We also received cDNAs through the Drosophila Genomics Resource Center. D.C.I.G. and C.W. were funded by the Biotechnology and Biological Sciences Research Council;C.W. was also supported by Diabetes UK (ref. BDA:RD02/0002540). D.M. and C.A.R.B. were funded by a Wellcome Programme Grant.