Scaling the Drosophila Wing: TOR-Dependent Target Gene Access by the Hippo Pathway Transducer Yorkie

Organ growth is controlled by patterning signals that operate locally (e.g., Wingless/Ints [Wnts], Bone Morphogenetic Proteins [BMPs], and Hedgehogs [Hhs]) and scaled by nutrient-dependent signals that act systemically (e.g., Insulin-like peptides [ILPs] transduced by the Target of Rapamycin [TOR] pathway). How cells integrate these distinct inputs to generate organs of the appropriate size and shape is largely unknown. The transcriptional coactivator Yorkie (Yki, a YES-Associated Protein, or YAP) acts downstream of patterning morphogens and other tissue-intrinsic signals to promote organ growth. Yki activity is regulated primarily by the Warts/Hippo (Wts/Hpo) tumour suppressor pathway, which impedes nuclear access of Yki by a cytoplasmic tethering mechanism. Here, we show that the TOR pathway regulates Yki by a separate and novel mechanism in the Drosophila wing. Instead of controlling Yki nuclear access, TOR signaling governs Yki action after it reaches the nucleus by allowing it to gain access to its target genes. When TOR activity is inhibited, Yki accumulates in the nucleus but is sequestered from its normal growth-promoting target genes—a phenomenon we term “nuclear seclusion.” Hence, we posit that in addition to its well-known role in stimulating cellular metabolism in response to nutrients, TOR also promotes wing growth by liberating Yki from nuclear seclusion, a parallel pathway that we propose contributes to the scaling of wing size with nutrient availability.

Wing discs from late third instar larvae-bearing clones of mutant tissue outlined with dashed lines, and marked negatively ("black") by absence of the GFP marker (green): (D) wild type (control) (E) Tor ΔP , (F) ex e1 , (G) ex e1 Tor ΔP . Mutant clones were induced at the end of the first instar, 48±2 hr after egg laying and are associated with sibling "twin-spot" clones marked by two copies of the GFP marker (bright green) that serve as an internal control for the growth of w.t. tissue. Numbers denote mean clone size ratio compared to wt, and asterisks denote significances from t tests (* = p < 0.05, ** = p < 0.01, *** = p < 0.001, n. s. = not significant). In (G), the bottom italicised value is a comparison with the Tor ΔP genotype. Number of clones measured (n) = 32 (wt), 38 (ex), 51 (Tor), 36 (ex Tor). (H-K) Clones of the same genotypes as in (D-G) that coexpress p35 with GFP-NLS (generated using the MARCM technique [36]. Clones are positively labelled by GFP-NLS, and nuclei are counterstained with Hoechst (blue). (n) = 92 (wt+p35), 90 (ex+p35),

TOR Activity Is Required for Yki-Dependent Wing Growth
To determine the relationship between TOR and Yki in the developing Drosophila wing, we first asked whether the capacity of Yki to promote growth depends on TOR activity. To do so, we generated clones of cells mutant for Wts/Hpo pathway components that normally hold Yki activity in check and tested if the resulting increases in tissue mass and cell number depend on TOR activity. All such experiments that we performed indicate an obligate role of TOR activity for Yki-dependent wing growth. Specifically, clones mutant for the FERM (4.1 Ezrin Radixin Moesin) domain protein Expanded (Ex), which is required for normal Hpo activity, grow far larger than control clones (Fig 1F; S1 Data [37]). However, when such clones were also mutant for TOR (ex -Torclones), they were tiny (Fig 1G; S1 Data) and not significantly larger than Torclones (Fig 1D and 1E; S1 Data). The impaired growth of both Torand ex -Torclones is not due to apoptosis, since it is not detectably rescued by expressing the apoptosis inhibitor p35 in the mutant cells (Fig 1H-1K; S1 Data). Equivalent experiments, activating Yki by removing Wts and inhibiting TOR by removing the TOR-activating GTPase Rheb, gave similar results (S1A- S1H Fig). Likewise, overexpressing Yki throughout the wing primordium leads to excessive growth; however, co-overexpression of ΔP60, an inhibitor of InR signaling upstream of TOR [38], blocks this growth, yielding wings similar in size to those in which only ΔP60, alone, was overexpressed (S1I-S1O Fig and discussion in S1 Fig legend; S1 Data). Equivalent results were obtained in the developing head primordium: mutation of wts or ex leads to dramatic overgrowth of head tissue, which is suppressed when the head is also mutant for Tor or the InR pathway kinase Akt (S1P- S1X Fig). Thus, reducing the level of InR and TOR signaling, which is normally set by nutrient status, restricts Yki's capacity to promote growth-suggesting a relationship that could contribute to scaling wing size with nutrient levels.

Yki Does Not Up-Regulate InR/TOR Pathway Activity to Promote Growth
Why is Yki-driven growth limited by the level of TOR activity? TOR might be required upstream to facilitate Yki activity, or in parallel to trigger other, independent growth-related processes. Alternatively, TOR might be required downstream, with Yki promoting growth at least in part by elevating InR/TOR pathway activity. To test this latter possibility, we assayed whether conditions that abnormally increase Yki activity (loss of either Ex or Wts) have a corresponding effect on the levels of phospho-Akt (pAkt-S505), which are normally elevated by enhanced InR signaling (e.g., by removal of the InR pathway inhibitor PTEN; Fig 2A). We find that pAkt-S505 levels are normal in exor wtsclones (Fig 2B and 2C), as well as in protein extracts from the overgrown wing discs of homozygous exor wtslarvae ( Fig 2D). Likewise, the levels of phosphorylated S6-Kinase (pS6K-T398), a readout of TOR activity, were not affected in exor wtsdiscs ( Fig 2E). We therefore infer that Wts/Hpo regulated Yki activity does not stimulate either InR or TOR pathway activity in the wing, consistent with results from Drosophila cell culture [15]. This finding argues against a downstream role of InR/TOR signaling in mediating Yki-driven growth.
Further evidence that TOR activation is not dependent on Yki comes from the observation that ectopically activating the InR/TOR pathway via upstream components can suffice to increase the growth of ykiwing clones (rescued with p35; Fig 2F, 2G and 2J; S1 Data). Specifically, expressing Rheb (to activate TOR) resulted in a 22% increase in clone size (Fig 2I and 2J; S1 Data) and expressing Dp110 (to activate InR) resulting in a 26% increase (Fig 2H and 2J; S1 Data). In both cases, clonal expansion resulted solely from increases in cell size (Fig 2L; S1 , and (C) wts X1 (experimentals, having elevated Yki activity owing to reduced or absent phosphorylation by Wts). Phospho-Akt S505 is not increased in (B) and (C), in contrast to (A). (D) Western blot of protein extracts derived from late third instar wing discs of the genotypes shown, labelled for phospho-Akt S505 (Total Akt and β-actin were used as loading controls; pten 1 /pten dj189 was used as a positive control; the reduction in S505 staining in the Tsc1 Q87X / Tsc1 PA23 lane is due to feedback of TOR activation onto Akt phosphorylation [39]). In contrast to reduced Pten, loss of either Ex or Wts does not cause an increase in pAkt S505. (E) Blot of same genotypes as in (D), labelled for phospho-S6 Kinase T398, an indicator of TOR pathway activity (β-Tubulin was used as a loading control, and runs as two species). As observed for phospho-Akt S505, Phospho-S6 Kinase T398 levels are elevated by loss of Pten and Tsc activity, but not by loss of either Ex or Wts activity. (F-I): Wing discs from late third instar larvae-bearing MARCM clones expressing UAS.p35 (labelled positively with GFP-NLS, green; nuclei are counterstained with Hoechst, blue). The genotypes of clones are (F) UAS.p35, (G) yki b5 +UAS.p35, (H) yki b5 + UAS.p35+UAS.Dp110, and (I) yki b5 + UAS.p35+UAS.Rheb. J-L: Quantification of clones sizes, cell numbers, and cell sizes from genotypes in F-I. Error bars are Standard Error of the Mean and asterisks denote significances from t tests (* = p < 0.05, ** = p < 0.01, *** = p <0.001, n. s. = not significant). n = 42 Data), without effects on cell number (Fig 2K; S1 Data). Hence, TOR can be activated and promote detectable cell growth, albeit not cell proliferation, in the absence of Yki. Notably, control clones expressing Dp110 or Rheb with p35 that are wild type for Yki strongly overgrew (S2A- S2D Fig; S1 Data), and this was in part due to both their increased cell numbers (S2E Fig; S1 Data), as well as strongly increased cell size (S2F Fig; S1 Data). We infer that although InR/ TOR pathway activity appears largely independent of Yki activity, ectopic InR/TOR activation nevertheless requires Yki function as a prerequisite in order to drive cell proliferation.

TOR Inhibition Enhances Yki Nuclear Accumulation despite Inhibiting Yki-Dependent Growth
If Yki depends on, but does not activate, the InR/TOR pathway to promote growth, an alternative explanation is that the InR/TOR pathway acts in parallel to or upstream of Yki. Yki activity depends in large part on phosphorylation by Wts, which causes Yki to be sequestered in the cytosol, reducing its access to the nucleus [15][16][17]. Yki normally appears predominantly cytoplasmic in the developing wing (Fig 3A), consistent with its essential, but tonic, activity in sustaining wing growth. However, upon Wts inactivation, Yki accumulates in the nucleus and inappropriately up-regulates transcription of its growth-promoting target genes, causing pronounced tissue hyperplasia [15][16][17] (Fig 3B). Hence, inhibiting TOR activity might countermand the overgrowth phenotype caused by loss of Wts activity by compromising nuclear access of unphosphorylated Yki. In preparing to test this possibility, we were surprised to discover that reducing TOR activity has the opposite effect on Yki localization: instead of restricting nuclear access, it causes a dramatic increase in nuclear Yki, even as it blocks Yki-dependent growth.
In our initial experiments, we strongly inhibited endogenous TOR activity in the developing wing, using dpp.Gal4 to drive a stripe of expression of UAS transgenes encoding three proteins that have been widely employed to study the consequences of reducing TOR activity: a dominant negative form of TOR (TOR TED ) and the TOR corepressors TSC1 and TSC2 [32,[40][41][42][43][44][45][46]. Co-overexpression of all three proteins causes strong nuclear Yki accumulation in the stripe (Fig 3C and 3D). Expressing either TSC1+TSC2, or TOR TED , alone, in the stripe or in other patterns within the wing, also caused enhanced Yki nuclear localization (S3A- S3C Fig), albeit less strongly, suggesting that the level of nuclear accumulation depends on the strength of TOR inhibition, and furthermore is not region-specific. Note that we prefer the use of transgenes to inhibit TOR signaling in large sectors of tissue, rather than examining tiny Torclones, where interpretation is compromised by protein perdurance. We also observed enhanced Yki nuclear accumulation when InR signaling was inhibited, upstream of TOR (S3D- S3F Fig). Conversely, removal of two downstream metabolic targets of TOR, dS6K, and 4EBP/Thor [10] did not affect Yki nuclear accumulation (S3G- S3J Fig), indicating that the control of Yki localization by TOR involves a mechanism that is not mediated by these metabolic regulators. Thus, reduced InR/ TOR activity results in enhanced nuclear accumulation of Yki-a result that is counterintuitive, because nuclear Yki is usually associated with increased, rather than reduced growth [15][16][17].

TOR Inhibition Reduces Yki Target Gene Transcription
Typically, nuclear Yki up-regulates growth-promoting target genes, one of the best characterized of these being diap1, the gene encoding the Drosophila Inhibitor of Apoptosis protein [18,19]. However, the dramatic increase in nuclear Yki accumulation caused by TOR inhibition (Fig 3C and 3D) is associated with a strong reduction in DIAP protein levels (Fig 4A and 4B), consistent with reduced diap1 gene transcription. Diap1 transcription is normally driven by a Yki-Sd complex that binds~4 kb downstream of the transcription start site of the predominant diap1-RA transcript [19]. A minimal reporter construct, 2B2C-lacZ, containing~300 bp of this enhancer region, is up-regulated by Yki overactivation [19]. Conversely, RNAi knockdown of Yki results in reduced 2B2C-lacZ expression (S4A and S4A ' Fig). Like Yki RNAi knockdown, TOR inhibition caused a marked decrease in 2B2C-lacZ expression (Fig 4C and 4D), indicating a failure of the Yki/Sd-dependent 2B2C enhancer to maintain normal target gene expression.
TOR inhibition similarly compromises the transcription of other Yki target genes. fourjointed (fj) is the second target gene that we have confirmed is reduced by Yki knockdown ( S4B Fig). Like Yki knockdown, TOR inhibition strongly reduced fj-lacZ expression (Fig 4E and 4F). A third reporter, ex e1 -lacZ, was unresponsive to TOR inhibition at 25°C, but raising the temperature to 29°C for several hours prior to dissection (to increase GAL4-dependent TOR TED , TSC1, and TSC2 overexpression) sufficed to repress this reporter (Fig 4G and 4H). Further, in animals homozygous for the ex e1 -lacZ reporter allele, ex e1 -lacZ expression was clearly reduced by TOR inhibition at 25°C (S4C and S4D Fig). We note that such ex e1 -lacZ/ex e1 -lacZ discs have reduced ex gene function and hence abnormally elevated levels of nuclear Yki; nevertheless, a sharp increase in nuclear Yki can still be detected in the TOR-inhibited stripe, coincident with reduced transcription of the ex e1 -lacZ gene (S4E- S4G Fig). Finally, the effect of TOR inhibition appears specific to Yki target genes, since other proteins expressed in the developing wing, including Vestigial (S4I

Inhibiting TOR Impedes Target Gene Access by Yki
The unexpected (and counterintuitive) effects of TOR inhibition on Yki nuclear accumulation and target gene expression are reminiscent of another scenario in which enhanced nuclear access of Yki is associated with the loss of target gene expression. In this case, Yki can be forced to accumulate in the nucleus by overexpression of Sd, the site-specific DNA binding factor that normally mediates target gene activation by Yki in the wing blade ( S6A Fig) [18,19]. Under this condition, the abnormally high levels of Sd sequester the limited levels of Yki available, exerting a classic dominant negative "squelching" effect [47] that titrates Yki away from Sd bound to target gene enhancers. As in the case of TOR inhibition, the level of nuclear Yki increases whilst target gene expression decreases (S6B and S6C Fig). This resemblance suggests that TOR inhibition may exert its effects on Yki target gene expression via a similar mechanism: by directing Yki to the nucleus whilst reducing its access to the relevant target gene enhancers.
To test this possibility directly, we focused on diap1, the locus where the general mechanism of transcriptional control by Yki is best characterized, with a defined enhancer element, 2B2C, that is controlled exclusively by Yki in complex with Sd (henceforth Yki-Sd) [14,18,19,48]. We prepared chromatin from control and TOR-inhibited wing discs (the latter coexpressing TOR TED +TSC1+TSC2 under the control of a Flp-out Act5C>CD2>Gal4 driver) and immuneprecipitated using a Yki antibody [15]. To produce enough tissue despite TOR inhibition, discs were allowed to grow to a large size before a strong heat shock was used to excise the stop cassette within the Act5C>CD2>Gal4 driver and initiate Gal4 expression in nearly all cells (Fig 5A and 5B; loss of magenta CD2 marks the Act5C>Gal4 expressing tissue), with dissection of 200 discs per genotype 8-10 hr later (see Methods).
In control discs, as expected, Yki was enriched~2-3-fold within the 2B2C diap1 element, using primers centered around the CATTCA motif to which Yki-Sd binds (relative to primers defining unrelated DNA sequences, as well as an IgG control treatment; Fig 5C). Knockdown of Yki (using the same Act5C>CD2>Gal4 protocol to drive transient expression of a UAS.yki R-NAi transgene) abolished this peak of enrichment, confirming its dependence on Yki levels and also validating the specificity of the antibody (S8A and S8B Fig). Strikingly, in experimental, TOR-inhibited discs, Yki enrichment at 2B2C was also diminished, by almost half (p < 0.05, n = 4 fully independent experimental replicates; Fig 5C; S1 Data). This reduction occurred despite the clear increase in nuclear Yki resulting from TOR inhibition (Fig 5A and 5B).
These data provide evidence that inhibiting endogenous TOR activity reduces the amount of Yki bound to a canonical Yki-Sd target gene enhancer, even as it causes an increase in Yki nuclear accumulation. Hence, we propose that TOR inhibition renders Yki subject to a phenomenon that we term "nuclear seclusion", whereby Yki is mobilized and/or sequestered in the nucleus but hindered from accessing its target enhancer sequences. We posit that this mechanism accounts at least in part for the reduction in DIAP1 expression observed when TOR is inhibited ( Fig 4B) and suggest that it likewise explains the loss of expression of other TORdependent Yki-Sd target loci for which defined enhancers have not yet been characterized (e.g., fj, ex; Fig 4F and 4H).
It is important to note that in most of our experiments, we have strongly reduced TOR signaling by coexpressing TOR TED +TSC1&2, to approximate, or at least approach, TOR loss-of function. These experiments reveal a basic requirement in wing cells for at least some level of TOR signaling in order to override nuclear seclusion of Yki and permit it access to drive transcription of target genes. However, we additionally propose that nuclear seclusion also operates within the normal physiological range of TOR activity as set by nutrient intake. First, weaker impairment of TOR signaling, by expression of TOR TED alone, produces a wing size comparable to a moderately staved animal (Fig 1C). And yet it also leads to the two hallmarks of nuclear seclusion: increased Yki accumulation (  Tor TED , TSC1, and TSC2 in most cells 8-10 hr after heat shock to excise the >CD2> stop cassette. Yki (red), CD2 (magenta), Sd GFP (from a GFP knockin allele of sd; green) and DNA (blue) are shown at the level of nuclei. Control discs (expressing only GFP-NLS) have no effect on Yki localistion (A) in contrast to experimental discs coexpressing TOR TED , TSC1, and TSC2, which show enhanced nuclear accumulation of Yki as indicated by increased signal at low magnification (B; arrowheads show clusters of Act5C>CD2>GAL4 cells in which the >CD2> stop cassette was not excised, which coincide with reduced Yki nuclear staining). (C, D) ChIP of Yki (C) or Sd GFP (D; using the sd GFP allele and anti-GFP antibody) with mock IP (IgG) at the 2B2C diap1 enhancer and a control locus (PDH: pyruvate dehydrogenase) in control and experimental discs, as in (A,B). Error bars are Standard Error of the Mean, and asterisks denote significances from t tests (* = p < 0.05, ** = p < 0.01, *** = p < 0.001, n. s. = not significant). n = 4 independent experimental replicates. Yki and Sd GFP are strongly enriched at 2B2C in control (w.t.) discs compared to PDH controls. Enrichment of both proteins is reduced in experimental (>TOR TED /TSC1&TSC2) discs; (E) Conserved functional domains of Yki. (F-H) Wing discs uniformly expressing w.t. (F), P88L (G), or WW domain mutant (W292A P295A W361A P364A) (H) forms of GFP tagged Yki that also express TOR TED , TSC1, and TSC2 in a stripe under dpp.Gal4 control (as in Figs 3 and 4; the dashed white lines indicate the A/P compartment boundary, which abuts the right edge of the dpp.Gal4 expressing stripe). The discs are imaged at the level of nuclei and show nuclear accumulation of the w.t. (F) but neither the P88L (G) or WW (H) mutant, forms of Yki, as indicated by signal intensity. (I, J) Confocal sections of wing discs taken at the level of nuclei to the assess the nuclear accumulation of wild-type Yki-GFP (I), as well as a mutant form of Yki-GFP (J), that carries both the P88L substitution (which blocks nuclear accumulation in response to TOR inhibition, G) as well as the S111A, S168A, and S250A (S3->A) substitutions (which obviate phosphorylation of Yki by Wts and cause otherwise wild-type Yki to accumulate in the nucleus). Both proteins are expressed in clones under the direct control of the Tubα1 promoter following Flp-out cassette excision of a Tubα1>DsRed>yki-GFP transgene: the clones are marked black by the absence of DsRed expression (magenta). Wild-type Yki-GFP (I) appears predominantly cytosolic, whereas Yki P88L S3->A -GFP (J) appears much more nuclear, as indicated by the difference in staining patterns imaged at the nuclear plane-largely absent in nuclei for (I) and relatively uniform for (J): (TOR TED +TSC1&2) gives the clearest observable nuclear accumulation, and is the basis for most of our mechanistic insights regarding Yki nuclear seclusion, we posit that this phenomenon operates beyond a simple baseline requirement for TOR signaling and is modulated by fluctuations of TOR signaling within the normal range. Furthermore, we note that because Yki is predominantly cytoplasmic in wing cells (Fig 3A), the small fraction of active, nuclear, Yki might be effectively secluded by quite modest reductions in TOR signaling (without leading to the profound nuclear accumulation witnessed when TOR is strongly impeded).

TOR Inhibition Impedes Target Gene Access by Sd, in Addition to Yki
What mechanism mediates Yki nuclear seclusion in response to TOR inhibition? Given the similar effect of Sd overexpression on Yki localization and target gene transcription, we considered the possibility that the mechanism mediating Yki nuclear seclusion could be a pronounced increase in Sd protein level. However, Sd is already at high level throughout the developing wing, and TOR inhibition does not cause a further increase, as monitored by the expression of a fully functional GFP protein trap allele of the endogenous sd gene (sd GFP ; [49]) (S4J" Fig J; S8G and S8H Fig). Hence, we sought to assess two other possibilities: (i) that Yki, alone, is subject to nuclear seclusion, e.g., by being titrated away from enhancer-bound Sd, or (ii) that Sd, in addition to Yki, and perhaps in complex with it, is targeted for seclusion.
Using sd GFP wing discs and a GFP antibody, we assessed Sd-GFP enrichment at the 2B2C diap1 enhancer by ChIP. Like Yki, Sd-GFP is strongly enriched at 2B2C in wild-type discs ( Fig  5D). However, inhibiting TOR activity caused a reduction in Sd-GFP enrichment at 2B2C to a similar degree as it does Yki enrichment (Fig 5C and 5D). Hence, it appears that TOR inhibition results in nuclear seclusion of Sd as well as Yki, rather than just Yki. One explanation for this observation is that Sd might need to be bound by Yki to bind 2B2C -indeed, such a cooperative interaction appears to be required for enhancer binding by Sd in combination with Vg, another Sd transcriptional coactivator [50]. However, we observed that RNAi knockdown of Yki, which as expected results in reduced Yki binding at 2B2C, has no effect on Sd enrichment ( S8B Fig). That Sd binding to 2B2C apparently does not depend on Yki argues against cooperative target site binding as a possibility. We therefore posit two possibilities: that Sd in complex with Yki is subject to nuclear seclusion, or, alternatively, that TOR inhibition secludes Sd in addition to, but independently of, Yki. If this second scenario were the case however, we may expect Sd in complex with transcriptional activators other than Yki to be influenced by seclusion. Conversely, expression of vg, an autoregulated target of Vg-Sd but not Yki-Sd complexes in wing cells, is not affected when TOR activity is inhibited (S4H and S4I Fig). Hence, we favor the view that nuclear seclusion acts on Sd in complex with Yki, although this additional facet of the mechanism remains to be confirmed.

Nuclear Seclusion Is Mediated by Yki's N-Terminus and WW Domains, But Not by Its Wts Phosphorylation Site
Yki is a modular protein with three regions that are conserved with vertebrate YAP and essential for normal function (Fig 5E): (i) a Wts phosphorylation site which, when phosphorylated, mediates cytoplasmic retention of Yki by 14-3-3 anchor proteins (ii) two WW domains that hence, the P88L mutation, which blocks Yki nuclear accumulation in response to TOR inhibition does not preclude nuclear accumulation in the absence of phosphorylation by Wts. (K) Yki P88L -GFP expressing wing disc carrying clones of wtsclones (outlined with dashed white lines). Nuclear accumulation of the Yki P88L -GFP protein is elevated in the absence of phosphorylation by Wts, as indicated by increased GFP signal intensity imaged at the nuclear plane, corroborating the results in (I,J). doi:10.1371/journal.pbio.1002274.g005 Control of Drosophila Wing Growth by TOR and Yorkie bind both Wts and Ex; and (iii) an N-terminal region, the Sd/TEAD binding domain (TBD), that mediates binding to Sd. Yki normally shuttles between the cytosol and nucleus in response to the state of phosphorylation at the Wts site [15][16][17], with phosphorylated and unphosphorylated forms accumulating, respectively, in the cytosol and nucleus. This shuttling is revealed by the build-up of nuclear Yki in wing cells following treatment with the nuclear export inhibitor Leptomycin B (S6D and S6E Fig; [51]). However, TOR inhibition is unlikely to cause abnormal Yki nuclear accumulation by reducing Wts-dependent phosphorylation of Yki, as this would mimic the consequence of Wts/Hpo inactivation [15][16][17], namely target gene induction rather than the observed repression. We therefore considered an alternative possibility that a different mechanism of Yki regulation, perhaps mediated by one or both of the other conserved Yki motifs, might be at play.
To shed further light on the mechanism of nuclear seclusion, we generated Tubα1. yki GFP transgenes that express GFP-tagged forms of Yki at near-physiological levels (S8C Fig; S1 Data) and analyzed the consequences of mutating each of the three conserved functional domains on the subcellular accumulation of Yki. TOR inhibition caused nuclear accumulation of wild-type Yki GFP as expected ( Fig 5F). However, introducing a mutation, P88L, in the N-terminal TBD that diminishes binding to Sd [19] prevented nuclear accumulation in TOR-inhibited cells (Fig 5G). This block to nuclear accumulation appears to be specific to the loss of TOR pathway activity, as Yki GFP-P88L protein accumulates in the nuclei of wtswing cells that have normal TOR pathway activity (Fig 5K), and the same is true for Yki GFP-P88L protein that is additionally mutated for the Wts phosphorylation sites in otherwise wild type wing cells (Fig 5  and 5J). Likewise, mutating Yki's two WW domains prevents nuclear accumulation following TOR inhibition (Fig 5H), even though the WW domains are known not to interfere with nuclear accumulation following Wts/Hpo pathway inactivation [52]. Thus, it appears that the Wts-phosphorylation site governs whether Yki has access to the nucleus: only Yki in which this domain is not phosphorylated by Wts can enter. In contrast, interactions mediated by the TBD and WW domains appear to determine whether any unphosphorylated Yki that gains access to the nucleus will be secluded there when TOR activity is inhibited. Accordingly, the abnormal nuclear accumulation of Yki resulting from TOR inhibition should represent a build-up of unphosphorylated Yki, even when Wts is present, as any unphosphorylated Yki that enters the nucleus will get trapped there. Consistent with this prediction, we observed a general reduction in phosphorylated Yki-S168 levels following TOR inhibition ( S8D Fig).
(Note that although we interpret this as a consequence of nuclear seclusion, we cannot rule out an additional, independent effect of TOR inhibition in dephosphorylating Yki. However, we note that if such an activity occurs, it has seemingly negligible function consequences for Yki activity; dephosphorylated Yki should increase target gene expression, but our data show that blocking TOR strongly reduces Yki target gene expression).
Based on these data, we infer that low levels of InR/TOR signaling activity result in the induction or function of a presently unidentified Nuclear Secluding Factor (NSF) that acts in wing cells to sequester Yki in the nucleus and to divert Yki-Sd from its target genes. The putative NSF might interact with Yki directly, via Yki's N-terminus and/or WW domains and cause Yki nuclear seclusion by both increasing the binding of Yki to Sd and preventing the resulting NSF-Yki-Sd complex from accessing Yki-Sd target genes. Accordingly, NSF action would result in enhanced Yki nuclear accumulation and reduced Yki-Sd target gene expression. This mode of Yki regulation acts independently of the Hpo-Wts mediated cytosolic tethering of Yki to limit target gene expression. Indeed, even in Hpo pathway mutant tissue (e.g., loss of Ex, or Wts), where Yki is constitutively dephosphorylated, TOR inhibition still leads to increased Yki nuclear accumulation (S4E- S4G Fig)

TOR Regulates Yki Nuclear Seclusion and Cellular Metabolism in Parallel to Control Wing Growth
The requirement for InR/TOR signaling activity to allow Yki to escape nuclear seclusion is revealed by genetic manipulations that compromise InR/TOR pathway activity and cause Yki to accumulate nonproductively in the nucleus. Importantly however, we do not think that TOR signaling promotes Yki-driven wing growth solely by this mechanism. Instead, we posit that this mechanism operates in parallel with other outputs of InR/TOR signaling that regulate a range of basic metabolic processes that are necessary, independently, for cell growth as well as rate-limiting for cell proliferation [5,10,53]. A clear demonstration of this parallel requirement for InR/TOR signaling is the observation that Yki-driven overgrowth of the wing (caused by removal of wts) could be suppressed by removal of S6-Kinase (dS6K), a TOR target that catalyses cap-dependent mRNA translation (S7A- S7D Fig; S1 Data), but which is not itself involved in nuclear seclusion (S3G, S3H and S3J Fig).
Hence, while TOR/InR activity is needed to override Yki nuclear seclusion, it is not sufficient to promote proliferative growth in the absence of concomitant activation of canonical metabolic targets like dS6K. Coupled to our reciprocal finding above, that TOR activity is unable to promote cell proliferation without Yki (Fig 2F-2L), we conclude that neither output of TOR is sufficient. Rather, both outputs are required, first to unleash Yki from nuclear seclusion, and second to provide the metabolic wherewithal for Yki target genes to promote growth. By concomitantly compromising both outputs, nutritional deprivation coordinates Yki target gene expression with decreased metabolic capacity to reduce the growth rate of the wing.

Superphysiological TOR Activity Triggers a Regulatory Feedback That Circumvents Wing Overgrowth
Our results describe how inhibiting TOR signaling reduces Yki function as part of the mechanism that might scale wing growth downwards in response to nutritional deprivation. To further elucidate how TOR and Yki might function together to control wing size, we asked whether superphysiological TOR activity might alter Yki localization and/or activity to scale wing size upwards. However, we were unable to detect any such changes in Yki using either of two well-established approaches to overactivate the InR/TOR signaling in vivo. Specifically, overactivating TOR by removing the negative regulator TSC1 in wing cells did not appear to alter the nuclear-cytoplasmic distribution of Yki as assayed either by immunofluorescence in clones of Tsc1mutant cells (Fig 6A) or by quantitating the nuclear-cytoplasmic ratio of Yki in fractionated whole wing discs from entirely Tsc1mutant larvae (Fig 6B). Similarly, we tested whether clonal removal of TSC1, or alternatively, clonal overexpression of the positive regulator Rheb, affects the expression of any of several Yki target genes, and again, observed no differences (Fig 6C and 6D; S9A and S9B Fig). Hence, superphysiological gains in TOR activity do not appear to influence either Yki localization or activity, suggesting that peak endogenous TOR signaling is sufficient to fully override Yki nuclear seclusion as well as any other InR/ TOR-dependent constraints on Yki coactivator function.
Nevertheless, manipulations that generate superphysiological levels of TOR activity do reveal a further facet of the TOR-Yki relationship in the wing. We noticed that despite strongly elevating TOR signaling in wing discs (Fig 2E), loss of Tsc1 activity failed to alter overall wing disc size (S9C and S9D Fig). Instead, it appears to cause excess growth that is counterbalanced by apoptosis (S9C and S9D Fig). The same was true of discs overexpressing rheb in all cells under the control of the imaginal disc-specific headcase.GAL4 driver (hdc.G4; Fig 6J, 6K and  6N; S1 Data), which produces an equivalent level of TOR activation to loss of Tsc1 (compare phospho-S6K levels in S7I Fig to those of Tsc1discs in Fig 2E). This increase in apoptosis can Fig 6. Superphysiological TOR activity causes excess, Yki-independent growth that is offset by a negative feedback that down-regulates the anti apoptic factors DIAP and bantam. (A) Wing disc carrying a clone of Tsc1 Q87X mutant cells (marked black by absence of GFP, green, and outlined with a be attributed to a repressive effect of superactivation of TOR on the expression of DIAP1 protein and the bantam microRNA (Fig 6E-6G; S9F and S9F ' Fig), both targets of Yki regulation that have antiapoptotic activity [55]. However, the repression of DIAP1 and bantam does not appear to be due to enhanced Wts kinase activity, as Tsc1discs showed a mild reduction rather than an increase in phospho-Yki S168 levels ( Fig 6H). Moreover, the effects on DIAP1 and bantam are seemingly Yki-independent, since the repression of DIAP1 appears to be posttranscriptional (S9A and S9B Fig), and ChIP experiments indicate that Yki binding to its target enhancer in the bantam gene is not reduced in Tsc1wing discs (Fig 6I) [56]. In principle, DIAP repression could be a secondary consequence of bantam repression, since DIAP1 protein is negatively regulated by the Head involution defective protein (Hid) [57], a bantam target [58]. However, we failed to observe any effect of overexpressing or removing bantam on DIAP1 protein levels (S9G and S9H Fig), from which it appears that TOR superactivation does not act via bantam to repress DIAP. Thus, TOR overactivation appears to cause cell death independently of the Wts/Hpo/Yki pathway, possibly mediated by reduced expression of both DIAP1 and bantam.
To gauge if abnormally high levels of TOR activity indeed repress DIAP and bantam to offset tissue overgrowth, these gene products were resupplied to discs overexpressing rheb under hdc.GAL4 control. Overexpression of diap1, alone, under hdc.GAL4 control produced no significant effect on wing disc size (Fig 6N; S1 Data). However, co-overexpression of diap1 together with rheb fully suppressed cell death that would otherwise result from the overexpression of rheb alone (Fig 6L), and disc size increased by 77% (Fig 6N; S1 Data). To resupply bantam, GAL4-dependent transgene expression was not used, since bantam overexpression itself induces growth [58,59], rendering synergistic outcomes difficult to interpret. Instead, a transgene was constructed that expresses GFP at low levels under the direct control of the Tubα1 promoter, with the 3'UTR containing a minimal 100 bp genomic bantam fragment that includes the 21 bp bantam microRNA [58]. The resulting Tubα1.GFP-bantam transgene does not itself stimulate overgrowth (Fig 6N; S1 Data) but produces enough bantam to rescue a homozygous null bantam animal to adulthood, producing a normally sized and patterned animal (S9O and S9P Fig). This suggests that a single copy of Tubα1.GFP-bantam produces a level of bantam activity comparable to homozygosity for the endogenous bantam gene. Strikingly, when introduced as a single copy into the hdc.G4>UAS.rheb background, disc size increased 70% (Fig 6N; S9N Fig; S1 Data), similar to the effect of co-overexpressing DIAP with rheb, although the addition of the Tubα1.GFP-bantam did not fully suppress cell death (compare Caspase staining in S9L and S9M Fig). Finally, combining co-overexpression of rheb and dashed white line; counterstained with Hoechst, blue). Yki accumulation (red), imaged at the nuclear plane, is unchanged in the clone. (B) Yki nucleocytoplasmic distribution is similar in wild-type and Tsc1discs. Effective separation of cell fractions was confirmed by Tubulin (cytoplasm) and Histone 3 (nucleus). (C, D) Wing discs carrying Tsc1 Q87X mutant clones (marked and imaged as in A): expression of the Yki target genes fj-lacZ (C) and ex-lacZ (D) (red) is not affected. (E, F) Compared to wild type discs (E), expression of Rheb in a stripe under ptc.GAL4 control (F) causes a reduction in DIAP accumulation (red), as well as bantam micro-RNA activity, the latter indicated by relief of repression of a bantam-GFP sensor (green) [54]; peak activity of the ptc.Gal4 driver is indicated by expression of β-galactosidase from a UAS.lacZ transgene, magenta). (G) Protein extracts of wild type, Tsc1 Q87X /Tsc1 PA23 , and homozygous wts P2 (positive control) discs probed for DIAP1 protein reveal that DIAP1 is strongly reduced in Tsc1discs. β-actin was used as a loading control. (H) Phospho-Yki S168 levels Tsc1homozygous mutants discs are not elevated (and are in fact mildly reduced) compared to wild type control discs (total Yki and β-actin were used as loading controls; CIP treatment was used to ensure the correct product was being observed). (I) Tsc1discs do not show a reduction in Yki enrichment at a Yki responsive enhancer in the bantam locus compared to wild-type discs. IgG mock IP and enrichment at the PDH (pyruvate dehydrogenase) locus were included as controls. (J, K) hdc-GAL4 wing discs that either do (K, experimental) or do not (J, control) express a UAS. rheb transgene, labelled for active caspase III (green): both discs are approximately the same size (counterstained with Hoechst, blue), but the experimental disc shows pronounced Caspase activity in contrast to the control. Control of Drosophila Wing Growth by TOR and Yorkie diap1 with Tubα1.GFP-bantam led to still more growth, with the wing disc more than doubling in size (110% larger; Fig 6H and 6N; S1 Data).
In sum, the effects of reintroducing DIAP1 and bantam reveal a novel effect of superphysiological TOR activity: the repression two key Yki target genes, albeit by Yki-independent mechanisms, to induce cell death and counterbalance the excessive tissue growth that would otherwise result from TOR hyperactivity. We suggest that this regulatory circuit normally safeguards against organ overgrowth and overriding it exposes a hidden capacity of superphysiological TOR signaling to stimulate tissue hyperplasia.

Discussion
Organ size depends on organ-intrinsic mechanisms that couple cell proliferation to the specification of pattern, as well as on external signals such as nutrients that scale overall body size [6][7][8][9]. Our study reveals one way in which these two distinct developmental phenomena can be integrated in the developing Drosophila wing, via a circuit linking nutrient-controlled TOR (and upstream InR) activity to Yki target gene access. A key finding is our discovery that, in addition to the well-known cytoplasmic anchoring mechanism employed by the Wts/Hpo tumour suppressor pathway to constrain Yki nuclear access, the InR/TOR pathway operates by a wholly different mechanism to allow nuclear Yki to activate its target genes. Inhibiting each pathway results in the accumulation of Yki in the nucleus, but with opposite effect. Nuclear Yki resulting from reduced Wts/Hpo activity drives Yki target gene expression and promotes growth. But, strikingly and counterintuitively, nuclear Yki resulting from reduced InR/TOR activity causes a loss of target gene access and impedes growth.
Based on our results, and as summarised in Fig 7A, we propose that reduced TOR signaling causes Yki to bind a secluding factor that sequesters Yki in the nucleus but diverts it from acting in complex with Sd to bind target loci, thereby reducing gene transcription. We posit that, by combining TOR-dependent relief from Yki nuclear seclusion with canonical Wts/Hpo phosphoregulation of Yki nuclear access, wing cells are able to integrate nutrient levels (via TOR) with patterning/morphogenetic inputs (via Wts/Hpo) to achieve a level of Yki-Sd activity that matches metabolic capacity and scales wing size appropriately. Elucidating this mechanism further will depend on identifying the proposed NSF and determining how its action is mediated by Yki's N-terminus and WW domains. NSF may bind Yki that shuttles into the nucleus, titrating it away from target genes; one possibility, raised by a reviewer, is that NSF acts by modifying Yki in some way, stabilizing the protein and preventing its degradation (thereby contributing to its observed accumulation in the nucleus) with a modification that blocks it from accessing target loci.
A model for how TOR and Wts/Hpo inputs are integrated during normal Drosophila wing growth is depicted in Fig 7B. When the native TOR pathway is signaling at its peak, the level of Wts/Hpo activity dictates the amount of unphosphorylated Yki that is free to enter the nucleus. TOR activity liberates the Yki-Sd complex from a default state of nuclear seclusion, permitting nuclear Yki-Sd to access and transcriptionally activate loci that promote proliferative growth. In parallel, TOR also stimulates cellular metabolic processes such as protein translation (mediated by targets such as dS6K), which are preconditions for tissue growth. Both outputs of TOR are required for wing cells to divide and gain mass under Yki-Sd control. Given the central role of TOR signaling in nutrient sensing, [5,10], we further propose that this circuit may contribute to the capacity of the developing wing to scale downwards under conditions of nutrient deprivation. Specifically, under nutrient-limiting conditions, we suggest that the consequent reduction in TOR activity renders the Yki-Sd complex subject to nuclear seclusion, thereby decreasing transcription of Yki-Sd target genes that promote growth. Reduced TOR activity would also diminish metabolism, matching metabolic capacity with the reduced activity of Yki-Sd target genes. Such a circuit would provide a sensitive means for developing tissues to integrate systemic and local inputs into growth. Our finding that excess activity of TOR appears to cause excess growth that is compensated by cell death via reduced DIAP1/bantam expression (Fig 7C) reveals that the circuit linking TOR and Yki is additionally stabilized by negative feedback, constraining organ growth in the event of erroneous, superphysiological TOR activity. Hence, a previously unknown facet of nutrient sensing by TOR is the incorporation of a self-limiting tumor suppression mechanism.
While our study delineates a circuit that integrates TOR and Yki activity during Drosophila wing growth, we note that it is only in this tissue that we find unequivocal evidence for nuclear seclusion of Yki, the means by which we posit this integration is achieved. Blocking TOR signaling in two other imaginal disc types, the eye and leg, does not appear to affect either Yki localization or target gene expression (S8I- S8K Fig). The same is also true even for other cell populations within the wing imaginal disc, notably those giving rise to the wing hinge and body wall. Hence, nuclear seclusion of Yki does not appear to be a general consequence of inhibiting TOR activity, and may occur primarily (or possibly only) in presumptive wing cells.
Intriguingly, the wing primordium is unique amongst imaginal disc tissues in expressing high levels of Sd under the control of Vg, a transcriptional coactivator that forms a complex with Sd to "select" the wing state. Although Yki-Sd complexes regulate growth control downstream of the Wts/Hpo pathway in the other Drosophila imaginal discs, Sd is expressed in these other tissues at tonic, low levels. Further, in the wing, Vg-Sd and Yki-Sd coexist, regulating distinct cohorts of target genes, and our present evidence suggests that inhibiting TOR does not alter the capacity of the Vg-Sd complex to gain access to, or activate, its distinct set of target genes (e.g., vg and sd itself). Hence, nuclear seclusion of Yki may represent an organ-specific mechanism to integrate nutrient-regulated InR/TOR and morphogen-dependent Wts/Hpo activities under conditions in which Sd plays a major, independent role in specifying cell state. between the nucleus and cytoplasm in response to phosphorylation (P) by Wts, which targets Yki to cytosolic tethers. Wing disc intrinsic signals (e.g., the morphogens Dpp and Wg) drive wing growth by down-regulating Wts activity, allowing a small proportion of Yki to escape phosphorylation-dependent tethering, enter the nucleus, and bind to its growth-promoting target genes in complex with Sd. Inhibiting the InR/TOR pathway results in up-regulation or activation of a putative NSF that sequesters unphosphorylated Yki in the nucleus and impedes binding of the Yki-Sd complex to its target genes, reducing Yki-Sd dependent growth. (B) Proposed integration of InR/TOR and Wts/Yki signaling to scale wing growth. In well-fed animals (left panel), TOR activation by wing disc extrinsic, nutrient-dependent signals facilitates Yki-dependent tissue growth by two parallel means: (i) by inhibiting NSF and thereby potentiating Yki nuclear access and target gene expression, and (ii), by up-regulating cell physiological functions such as dS6-Kinase and possibly many others, to match growth potential to the level of Yki target gene activity. Reductions in InR/TOR activity diminish both outputs, leading to reduced cell proliferation and tissue growth. (C) Excess growth caused by superphysiological activation of TOR is offset by Yki-independent down-regulation of the antiapoptotic factors DIAP1 and bantam, promoting cell death and safeguarding the developing wing against hyperplasia. doi:10.1371/journal.pbio.1002274.g007

Control of Drosophila Wing Growth by TOR and Yorkie
It remains possible that TOR activity regulates nuclear seclusion of Yki to control growth in other tissues, and possibly in other animals, albeit in complex with other transcriptional effectors and primarily through the control of target gene access rather than nuclear access. Alternatively, TOR may limit growth in these other systems exclusively through its regulation of cellular metabolism, without additionally modulating the level of Yki activity. Studying how global chromatin occupancy by Yki changes in response to TOR pathway activity, and varies by tissue type, will be needed to evaluate these scenarios.
To our knowledge, nuclear seclusion represents a novel paradigm of transcription factor regulation. Although experimental overexpression of nuclear Yki cofactors (e.g., Sd [18,52]) can exert similar effects on Yki localization and target gene expression, thus far it has not been demonstrated that such a mechanism functions as a biologically relevant mode of transcription factor regulation operating in vivo. Presently, the identity of the factor mediating Yki nuclear seclusion is unknown. However, in both yeast [60] and mammals [61,62], mTOR functions as a classical kinase that phosphorylates diverse substrates, leading to their retention in the cytoplasm. We hypothesize that our proposed NSF is one such substrate. A potential avenue of future exploration will be systematically testing how TOR inhibition influences the increasing number of identified nuclear modulators of Yki-some of which, like Sd, affect Yki localization and function when overexpressed (e.g., Hipk [63,64], MASK [65,66] and Tgi [48,67]). As we have demonstrated for the regulation of Yki by TOR, nuclear seclusion offers a complementary mechanism for controlling transcription factors that are already subject to phosphorylation-dependent partitioning between the cytosol and nucleus. Nuclear seclusion may therefore be used more widely to achieve signal integration, both during development as well as in mature tissues.

Quantification of Wing and Clone Sizes and Cell Numbers
All measurements were made blind, without knowledge of the genotype. When performing multiple t tests involving the same treatments, Bonferroni-corrected p-values were used to assess the level of significance. Clone and wing sizes were measured in Adobe Photoshop using the "lasso" tool, and cell numbers counted using the "count" tool. For whole wing cell number estimates, setae were counted within a 170 x 170 pixel box within the posterior intervein region. Total wing cell number = (wing size in pixels / (170 x 170)) x cell number in 170 x 170 pixel box.
For ChIP at the bantam locus, chromatin was prepared from~150 wild type and Tsc1 Q87X / Tsc1 PA23 mutant wing discs, and subjected to pulldown with ant-Yki or IgG control antibodies as explained above. An~7 kb region around the microRNA was screened for Yki enrichment, and a strong peak was determined within a region 1,861-1,621 bp 5' to the 21 bp hairpin, corresponding to positions 640419-640646 on chromosome 3L of the April 2006 assembly of the Drosophila melanogaster genome (BDGP R5/dm3). This region was targeted in qPCR using the primers: bantam3F 5'-CGGGAACAGTCATAAAAGTTG C/bantam3R 5'-CTTTGCCTGTTCTGCCAT CC. Note that this region lies inside the C12 enhancer region bound by Yki identified by Oh and Irvine [56].
tubulin-bantam. EGFP with a 3'UTR containing a 100 bp fragment encompassing the bantam hairpin [58] was inserted as a KpnI-XbaI fragment into p(Tuba1>DsRed, y 2 >)AttB, and the construct was inserted into the genome at 25C7. The >DsRed, y 2 > cassette was excised to generate Tuba1>ban. This flipped out form was introduced into the null ban Δ1 homozygous background to assess rescue.

Quantification of Transgenic Yki expression
To quantify Yki expression from the tub-Yki-GFP transgene, the cassette was removed in clones, and discs were stained for total Yki. The staining intensities of five wild type regions and five regions of clonal tissue were quantified in ImageJ and averaged. The ratio between clonal and wild type tissue was calculated as 1.69. When yki and ΔP60 were coexpressed, blade size was still 57% smaller than wild type (L, M, the slight size increase being manifested in a modest increase in cell number (N). The accelerated rates of cell growth and cell division caused by Yki expression are thus equally hindered by ΔP60 coexpression. (P-X) Standard Error of the Mean images of heads composed predominantly of mutant tissue generated using the EGUF/hid technique [68]. (P-S) Images are to scale; (P) wild type (control). (Q) Tor ΔP causes a "pinhead" phenotype composed of fewer and smaller cells. (R) ex e1 caused substantial overgrowth of the head (note that overgrowth of the eye field does not occur in ex mutants [71,72], but the rest of the head capsule is strongly affected). (S) ex e1 Tor ΔP again produces a pinhead phenotype, indistinguishable from (J). (T-X) Images are to scale. (T) akt q , (U) foxo 25 akt q (V) wts X1 , (W) akt q wts X1 (X) foxo 25 akt q wts X1 . InR signalling activates TOR at least in part via the kinase Akt [73], but Akt also controls the phosphorylation and cytoplasmic sequestration of the transcription factor FOXO [74][75][76]. FOXO is dephosphorylated when InR/Akt activity is diminished and accumulates in the nucleus to suppress cell growth and proliferation [76,77]. Loss of akt decreases head size (T), an effect which can be partially rescued by simultaneous removal of foxo (U) [77]. akt loss was able to block the wtsovergrowth phenotype (V, W), and importantly, overgrowth was still mostly suppressed when foxo was also removed (X). These results show that blocking InR pathway activity inhibits Yki-driven cell proliferation largely in a FOXO-independent manner. (TIF) . Numbers denote disc size ratios compared to wild type; in (D), the bottom italicised value is a comparison with the ds6k genotype. Asterisks denote significances from t tests ( Ã = p < 0.05, ÃÃ = p < 0.01, ÃÃÃ = p < 0.001, n. s. = not significant). Number of discs measured = 12 (wt), 5 (wts), 13 (ds6k), 12 (ds6k wts). Yki-driven overgrowth of the wing (caused by removal of wts) can be suppressed by removal of S6-Kinase (dS6K; compare discs in (B) and (D), a TOR target that catalyses capdependent mRNA translation, but which is not involved in the nuclear sequestration of Yki (S2H Fig). (TIF) F) fj-lacZ expression (magenta) is still repressed in the Ptc domain (green in rightmost panels of E and F) by expression Tor TED with ptc-Gal4> in wts X1 clones (E; marked by absence of GFP in second panel), and wts X1 /wts P2 homozygous mutant wing discs (F). (G, H) An Sd-GFP protein trap line (green) shows similar expression in wild type discs (G) and discs expressing Tor TED  Overexpressing bantam using the EP line EP(3)3622 with ptc-GAL4 (G; domain labelled by coexpressed ß-galactosidase; magenta), strongly diminishes expression of the bantam sensor (G') without noticeable effects on DIAP1 protein levels (G; red). Likewise, in H, bantam homozygous mutant clones (labelled by absence of magenta ß-galactosidase) show a very strong increase in bantam sensor levels (green) but no discernable change in DIAP1 protein levels (red). This indicates that native DIAP1 levels are regulated independently of bantam in the wing. (J-N) hdc. GAL4 discs labelled for cleaved Caspase III (green) and Hoechst (blue). hdc>rheb discs (K) are the same size as wild type (J) but show highly elevated levels of cell death. When diap1 and rheb were coexpressed, cell death is fully suppressed (L), and disc size increases by 77%; introducing tub.GFP-bantam while coexpressing rheb increases disc size by 70% (M), without obviously suppressing cell death. Simultaneously combining rheb, diap1, and tub.GFP-bantam together caused the disc to more than double in size (N). (O, P) Wings from adult female flies of the genotypes: bantam -/+ (O) and bantam -/bantamtub.GFP-bantam (P). A single copy of tub.GFP-bantam rescues a bantam homozygous mutant animal, producing a normally sized adult. (TIF)