Introduction

The growth of animal tissues is known to depend on nutritionally induced circulating hormones of the insulin/insulin-like growth factor 1 (IGF-1) family [1–3]. In addition, tissue growth can be mechanically induced via the Hippo signalling pathway [4–8]. How these two mechanisms of tissue growth control are integrated during animal development remains unclear. The fruit fly Drosophila offers a tractable model system to explore how insulin/IGF-1 signalling (IIS) and the Hippo pathway interact to influence cell proliferation in multiple epithelial tissues during development. In Drosophila oogenesis, the developing oocyte in each egg chamber is fed cytoplasm by endoreplicating germline nurse cells that grow enormously in size by cell growth without cell division [9–13]. The somatic follicle cell epithelium that surrounds the germline grows by coordinated cell growth and cell division—cell proliferation—to precisely encapsulate each egg chamber during early stages of oogenesis (stages 1–7) [14]. Proliferation of follicle cells is thought to occur in response to mechanical stretching caused by the growth of the underlying germline [15, 16]. Furthermore, growth of the entire egg chamber is known to depend on adequate nutrient intake by the adult female [17]. After fertilisation of the egg, embryogenesis proceeds within the eggshell without the need for feeding or growth. At the end of embryogenesis, the hatching larvae is able to feed and grow via a similar subdivision of tissues into those growing without cell division (the endoreplicating larval tissues) and those growing by cell proliferation (the mitotic ‘imaginal discs’ that give rise to the future adult body or ‘imago’) [11, 12, 18]. The growth rate of the entire larva, including both endoreplicating and proliferating tissues, once again depends upon nutritional intake and thus contributes to the determination of adult body size [1, 2, 19].

The IIS pathway was found to be induced by nutritional cues in Drosophila [1–3, 20–23]. Nutrient intake stimulates production of circulating Drosophila insulin-like peptides (Dilps), which activate the insulin/IGF-1 receptor (InR), leading to the recruitment of the phosphatidyl inositol-3-kinase (PI3K—composed of two subunits, p60 and p110), which coverts phosphatidyl inositol phosphate (PIP)2 to PIP3 at the plasma membrane of signal-receiving cells (a process antagonised by the phosphatase and tensin homolog PTEN) [3, 20–30]. Induction of PIP3 leads to the recruitment of two Pleckstrin Homology (PH)-domain–containing kinases, phosphoinositide-dependent protein kinase 1 (PDK1) and Akt (also called protein kinase B [PKB]), enabling PDK1 to phosphorylate Akt, which is then responsible for signal transduction [31–37] and tissue growth control [38–41].

One important downstream effector of Akt is the Target of Rapamycin (TOR) complex 1 (TORC1), which promotes cell growth by inducing translation via Ribosomal protein S6 kinase (S6K) [39, 42, 43] and inactivating the translational inhibitor, Elongation initiating factor 4E binding protein (4E-BP) [43, 44]. TORC1 also induces ribosome biogenesis via the RNA Polymerase (Pol) II transcription factors Myc [45–49] and DNA replication element binding factor (DREF) [50] and the RNA Pol I activators Transcription intermediary factor 1A (TIF1A/Rrn3) [51, 52] and TIF1B/SL1-Upstream binding factor (UBF) [53], as well as by inactivating the RNA Pol III inhibitor Maf1 [54, 55]. TORC1 is composed of the TOR kinase, Raptor, and Lst8 subunits [56–59] and is activated by Ras-homolog enriched in the brain (Rheb) GTPase [60–62] at the lysosome [63, 64] and inhibited by the Tuberous Sclerosis Complex TSC1/2 GTPase-activating proteins (GAPs) [65–67] and the 40-kDa proline-rich Akt substrate (PRAS40) [68–70]. Akt appears to activate TORC1 by phosphorylating TSC2 to displace it from the lysosome [71, 72], although mutation of two Akt phosphorylation sites in TSC2 does not affect growth in Drosophila [73], suggesting that other Akt targets, such as PRAS40, may compensate. Importantly, a negative feedback loop operates in which TORC1 activation promotes cell growth but also functions to reduce phosphorylation and activation of Akt itself [74, 75].

Another target of Akt is the transcription factor Forkhead box O (FOXO), which, upon IIS activation, becomes phosphorylated by Akt to induce its retention in the cytoplasm [76–78]. Drosophila foxo mutants are viable and normally sized, but foxo is required for complete suppression of tissue growth upon reduced insulin/IGF-1–PI3K–Akt signalling, which sends FOXO to the nucleus to inhibit cell proliferation and promote apoptosis [79, 80]. Thus, Akt can act independently of the TORC1 complex to control cell proliferation rather than cell growth, raising the question of whether additional Akt substrates may exist that contribute to the control of tissue growth by the IIS pathway in proliferating tissues. The importance of this question is underscored by the fact that activation of PI3K–Akt signalling (in PI3K catalytic subunit alpha [PIK3CA] constitutively active mutant or in PTEN inactivating mutant tumours) is a widespread cause of malignant human cancer, while activation of TORC1 (in TSC1 or TSC2 mutant tumours) causes benign hamartomas [75, 81].

The Hippo signalling pathway was discovered by Drosophila genetics to restrict the rate of cell proliferation and final tissue size in developing wing and eye imaginal discs but does not affect cell growth in endoreplicating tissues [4–6]. The upstream kinase Hippo (Hpo) activates the downstream kinase Warts (Wts), which then directly inhibits the nuclear localisation of Yorkie (Yki), a transcriptional coactivator for Scalloped (Sd)/TEA domain (TEAD) DNA-binding proteins [82–84]. In mutants for hpo or wts, Yki translocates to the nucleus to switch Sd/TEAD from a repressor to an activator, which then drives cell proliferation and tissue growth in the developing wing and eye imaginal discs [83–86]. When mutant clones for yki are induced in the wing disc, cell proliferation is inhibited, and cells are eliminated by apoptosis [82, 87]. These dramatic phenotypes strongly suggest that the Hippo pathway might have a physiological role in tissue growth control in proliferating tissues, but it is still unclear how dynamic regulation of Yki nuclear localisation is controlled as tissues initiate or arrest their proliferation during development.

A physiological role for the Hippo pathway as a sensor of mechanical stress and strain in vivo was recently identified in Drosophila [88, 89]. Here, we explore how mechanical regulation of Hippo pathway integrates with nutritional control of tissue growth via the IIS pathway in the ovarian follicle cell epithelium and wing imaginal disc. Previous work indicated that genetic alteration of IIS can influence Hippo pathway signalling, but the physiological relevance of this observation and the molecular mechanisms responsible are still not fully understood [90]. Here, we propose that insulin/IGF-1 act via Akt to inhibit Hippo pathway signalling and promote nuclear localisation of Yki in response to mechanical stress and strain. In the follicle cell epithelium, Yki responds primarily to mechanical strain driven by growth of the germline (itself insulin/IGF-1– and TORC1-dependent) but also requires PI3K–Akt signalling in follicle cells. In the wing imaginal disc, Yki responds to Fat–Dachsous signalling [4–8] and mechanical stress in columnar cells [89, 91–94], as well as to mechanical strain in flattened peripodial cells [88, 95], and we uncover an additional physiological regulation by nutritionally regulated IIS via PI3K–Akt. We further confirm a recent report that TORC1 is not required in the wing disc for nuclear localisation of Yki and that loss of TORC1 in fact drives Yki to the nucleus [96], which we attribute to feedback effects on PI3K–Akt. Finally, we confirm our findings in the mammalian skin epithelium, an example of a tissue in which both IIS and mechanically regulated Yes-associated protein (YAP) are crucial to promote cell proliferation.

Results

We first examined early Drosophila oogenesis, where growth of the germline was shown to mechanically induce proliferation of surrounding follicle cells during stages 2–6 [15, 16]. We find that expression of the Yki reporter gene expanded.lacZ (ex.lacZ) is elevated and that green fluorescent protein (GFP)-tagged Yki localises to the nucleus in proliferating follicle cells of early-stage egg chambers (Fig 1A and 1B and S1 Fig). Since ex.lacZ suffers from perdurance of the lacZ-encoded beta-galactosidase protein, which can obscure effects over short timescales, we chose to focus on Yki–GFP localisation as a more dynamic readout. We noticed that Yki nuclear localisation exhibited some female-to-female variation, evident in the quantification of the nuclear/cytoplasmic ratio of wild-type females (Fig 1B). We wondered whether this variation might relate to the growth rate of the egg chambers. To test this notion, we arrested germline growth beyond stage 4 using Ovo dominant (OvoD) mutants, which we find prevents mechanical stretching of follicle cells and inhibits Yki nuclear localisation (Fig 1B). Note that in OvoD egg chambers, some nuclear Yki is still detectable prior to stage 4, but after stage 4, Yki is strongly cytoplasmic, as shown in the quantification (Fig 1B). Yki is known to promote follicle stem cell maintenance [97] and polar cell fate [98], but whether Yki is also involved in stretch-induced cell proliferation in follicle cells has not been explored. We find that Yki is essential for normal follicle cell proliferation because silencing of yki expression specifically in follicle cells by trafficjam.Gal4 (tj.Gal4)-driven transgenic Upstream activator sequence (UAS).yki-RNA interference (RNAi) reduces the number of phospho-Histone-H3–positive mitotic cells in each ovariole and profoundly disrupts oogenesis without affecting cell fate specification of follicle cells (marked by expression of Eyes Absent [Eya] and Broad) or follicle cell apoptosis (marked by Drosophila Death Caspase 1 [Dcp1] staining) (Fig 1C and 1D). Note that the insufficient number of follicle cells in the absence of Yki leads to germline apoptosis (Dcp1-positive) [16] (Fig 1D). Silencing of Sd by expression of UAS.sd-RNAi causes a milder phenotype, reducing follicle cell proliferation without disrupting germline growth during early stages (Fig 1C and 1D and S2 Fig). These results show that Yki subcellular localisation is dynamically patterned during early proliferative follicle cell development, becoming nuclear-localised in order for the Yki–Sd transcription factor to promote follicle cell proliferation during early egg chamber growth.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Yki localises to the nucleus to drive stretch-induced follicle cell proliferation in Drosophila egg chambers.

A) Expression of the Yki–Sd reporter gene ex.lacZ in early egg chambers is revealed by the pattern of nuclear beta-galactosidase immunostaining. Note that ex.lacZ is expressed during the stages in which mechanical stretch drives follicle cell proliferation. B) Expression and subcellular localisation of a Yki–GFP knockin line in early egg chambers reveals nuclear localisation of endogenous Yki during the stages of stretch-induced follicle cell proliferation. Quantification of the nuclear/cytoplasmic ratio of Yki–GFP is shown on the right (n > 30 per genotype; t test p < 0.05). See supplementary file S1_Data.xlsx for underlying data. C) Mitotic cells (pH3+) in the follicle cell epithelium are reduced upon expression of Yki–RNAi or Sd–RNAi. Numbers of follicle cells are reduced by both Yki–RNAi or Sd–RNAi, more weakly in the latter case. The Yki–RNAi phenotype is so strong that germ line cells are often affected by insufficiency in follicle cell numbers. Follicle cell fate, marked by expression of Eya, is not affected. The graph shows quantification of the average number of mitotic cells (pH3+) per egg chamber in each of the genotypes above (n > 10 egg chambers per genotype). For the Yki–RNA quantification, weaker phenotypes that still form individual egg chambers were used to measure pH3+ cells per egg chamber. See supplementary file S1_Data.xlsx for underlying data. D). Apoptosis, marked by Dcp1-positive cells, does not occur in follicle cell epithelia expressing Yki–RNAi or Sd–RNAi but does occur in germline cells affected by insufficiency in follicle cell numbers upon expression of Yki–RNAi. Follicle cell fate, marked by expression of Broad at stage 6 onward, is not affected by either Yki–RNAi or Sd–RNAi. Scale bars approximately 10 μM. Dcp1, Drosophila Death Caspase 1; ex.lacZ, expanded.lacZ; Eya, Eyes Absent; GFP, green fluorescent protein; IR, Inverted Repeat Hairpin RNAi; n/c, nuclear/cytoplasmic ratio; OvoD, Ovo dominant; pH3+, phospho-Histone-H3–positive; RNAi, RNA interference; Sd, Scalloped; WT, wild type; Yki, Yorkie.

https://doi.org/10.1371/journal.pbio.3000509.g001

The fact that the loss of Sd causes a weaker phenotype than loss of Yki is expected because Sd acts as default repressor such that loss of Yki prevents activation and leads to constitutive repression of Yki–Sd target genes, while loss of Sd prevents either activation or repression of target genes [86, 99]. Nevertheless, we sought to confirm that our UAS.sd-RNAi line was indeed causing a full loss of Sd function. Since Sd acts as a direct binding partner for Yki in the nucleus, it promotes Yki nuclear localisation [100–102]. We find that expression of UAS.sd-RNAi is sufficient to completely abolish Yki nuclear localisation in the follicle cell epithelium of both early- and late-stage egg chambers (S3A and S3B Fig). We therefore characterised the UAS.sd-RNAi phenotype in more detail and found that the resulting adult female animals are completely sterile because of arrest of egg chamber development at stage 10 of oogenesis (S3B Fig). When we examine such arrested stage 10 egg chambers, we observe that they lack sufficient follicle cells to cover the surface of the egg chamber, leading to subsequent apoptosis of germline cells (Dcp1-positive) (S3C Fig). These results show that strong Sd loss of function disrupts follicle cell proliferation such that there are insufficient cells to cover the germline and ensure its survival beyond stage 10 of oogenesis. This Sd loss-of-function phenotype confirms a physiological requirement for regulation of the Hippo pathway during follicle cell proliferation in vivo.

Since both follicle cell proliferation and germline growth are known to depend profoundly on nutritionally induced IIS [16, 17, 103, 104], we tested whether nutritional regulation of IIS in the germline cells might affect Yki subcellular localisation in the overlying follicle cell epithelium. We first compared the localisation of Yki–GFP in ovarioles from control (fed) adult females with those subjected to nutrient restriction for 24 h. We found that nutrient restriction caused a dramatic reduction in the nuclear localisation of Yki–GFP in follicle cells (Fig 2A). To test whether the effects of nutrient restriction were mediated by IIS to the germline, we induced null mutant clones of akt³ in the germline. We found that, as expected, cell growth is dramatically impaired in akt³ mutant germline cells and that this leads to relocalisation of Yki–GFP to the cytoplasm in the overlying follicle cell epithelium, which exhibits reduced cell proliferation (Fig 2B). The effect of Akt on germline growth is mediated by TORC1 because expression of UAS.TOR-RNAi in germline cells with the maternal tubulin.Gal4 driver reduces germline cell growth and is sufficient to reduce mechanical stretching and nuclear localisation of Yki–GFP in the surrounding follicle cell epithelium (S4 Fig). Thus, IIS acts on the germline to drive cell growth and thus indirectly induces mechanical stretching of the follicular epithelium and nuclear localisation of Yki in follicle cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Insulin–PI3K–Akt signalling stimulates germline cell growth to drive mechanical stretching of follicle cells and Yki–GFP nuclear localisation.

A) Yki–GFP localises to the nucleus in early-stage ovarian follicle cells but relocalises to the cytoplasm upon nutrient restriction of adult females for 24 h. Quantification of the nuclear/cytoplasmic ratio is shown on the right (n > 30 per genotype, t test p < 0.05). Scale bars approximately 10 μM. See supplementary file S1_Data.xlsx for underlying data. B) Yki–GFP localises to the nucleus in early-stage ovarian follicle cells but relocalises to the cytoplasm upon induction of akt³ mutant clones in the germline, marked by the absence of nlsRFP. Note that follicle cells are wild type. Scale bars approximately 10 μM. C) Surface views of akt³ mutant germline clones from (B) allow quantification of n/c. Scale bars approximately 3 μM. See supplementary file S1_Data.xlsx for underlying data. GFP, green fluorescent protein; nlsRFP, nuclear red fluorescent protein; n/c, nuclear/cytoplasmic ratio; pH3, phospho-Histone H3; PI3K, phosphatidyl inositol-3-kinase; Yki, Yorkie.

https://doi.org/10.1371/journal.pbio.3000509.g002

We next tested whether IIS in the follicle cell epithelium itself affects Yki subcellular localisation because this pathway has previously been shown to act via PI3K and PDK1 to promote Yki activity in imaginal discs [90]. We sought to examine the role of the key downstream effector of these kinases, Akt, whose potential role was previously overlooked because RNAi of Akt in S2 cells did not completely abolish the effect of insulin on Yki phosphorylation [90]. We therefore induced mutant clones of akt³ in the follicle cell epithelium and found that Yki–GFP relocalises to the cytoplasm in these clones, even following strong cellular stretching at late stages of oogenesis (Fig 3A–3F).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Insulin–PI3K–Akt signalling is required for nuclear localisation of Yki–GFP in the follicle cell epithelium.

A) Yki–GFP localisation to the nucleus is detectable in follicle cells at stage 4 of oogenesis. Scale bars approximately 10 μM. B) Yki–GFP localisation to the nucleus is weak or absent in follicle cells at stage 4 of oogenesis that are mutant for akt³. Note that germline cells are wild type. Scale bars approximately 10 μM. C) Yki–GFP localisation to the nucleus is weak or absent in follicle cells at stage 7 of oogenesis that are mutant for akt³. Scale bars approximately 10 μM. D) Yki–GFP localisation to the nucleus is weak or absent in follicle cells at stage 10 of oogenesis that are mutant for akt³. Scale bars approximately 10 μM. E) Yki–GFP localisation to the nucleus is weak or absent in follicle cells at stage 12 of oogenesis that are mutant for akt³. Scale bars approximately 10 μM. F) Yki–GFP localisation to the nucleus is weak or absent in follicle cells at stage 14 of oogenesis that are mutant for akt³. Scale bars approximately 10 μM. GFP, green fluorescent protein; nlsRFP, nuclear red fluorescent protein; Yki, Yorkie.

https://doi.org/10.1371/journal.pbio.3000509.g003

PI3K–PDK1–Akt signalling acts by regulating the canonical Hippo pathway because an Hpo dimerisation reporter, normally detectable only in columnar follicle cells and not in stretched cells [88], is strongly up-regulated upon inhibition of PI3K or Akt or overexpression of PTEN (Fig 4A–4C). The effect of Akt depends on canonical Hippo pathway signalling because double mutant akt³ wts^X1 clones still exhibit nuclear Yki–GFP localisation, even in columnar cells (Fig 4D). Furthermore, Yki nuclear localisation in highly stretched cells can be inhibited by treatment with two different PDK1 inhibitors, BX-795 and BX-912, whose effects are blocked by coexpression of UAS.Hpo-RNAi or UAS.Wts-RNAi (Fig 4E). The effect of Akt in follicle cells is not mediated by TORC1 because expression of dominant-negative UAS.TOR^TED or UAS.TOR-RNAi or induction of mutant clones for rheb does not prevent nuclear localisation of Yki–GFP (Fig 4F and 4G). Together, the above results show that IIS, via PI3K–PDK1–Akt, directly regulates the canonical Hippo pathway and is a prerequisite for nuclear localisation of Yki in both proliferating early-stage follicle cells and flattening late-stage stretch cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Insulin–PI3K–PDK1–Akt signalling to the follicle cell epithelium inhibits Hpo to drive nuclear Yki–GFP.

A) Expression of an Hpo KD BIFC sensor is detectable at the apical membrane of columnar follicle cells at stage 10 of oogenesis and is dramatically up-regulated upon coexpression of a dominant-negative PI3K p60 subunit (UAS.p60). A KA Hpo BIFC sensor is also dramatically up-regulated compared to the KD version. Scale bars approximately 10 μM. B) Expression of PTEN up-regulates the KD Hpo BIFC sensor in columnar follicle cells at stage 10 of oogenesis. Scale bars approximately 10 μM. C) Expression of Akt–RNAi up-regulates the KD Hpo BIFC sensor in columnar follicle cells at stage 10 of oogenesis. D) Yki–GFP localisation to the nucleus is weak or absent in follicle cells at stage 10 of oogenesis that are mutant for akt3 but is nuclear-localised in akt³ wts^X1 double mutant clones. Scale bars approximately 10 μM. E) Yki–GFP localises to the nucleus in stretch cells at stage 10 of oogenesis. Inhibiting PDK1 activity acutely with BX-795 or BX-912 blocks Yki–GFP nuclear localisation in a manner that depends on the canonical Hippo pathway kinases Hpo and Wts, whose knockdown with RNAi restores Yki–GFP nuclear localisation even in the presence of PDK1 inhibitors. F) Yki–GFP localises to the nucleus even upon inhibition of TORC1 by expression of dominant-negative TOR-TED or TOR-RNAi in stretch cells at stage 10 of oogenesis. G) Yki–GFP localises to the nucleus even upon inhibition of TORC1 by induction of rheb mutant clones at stage 14 of oogenesis. BIFC, bimolecular fluorescence complementation; DN, dominant-negative; GFP, green fluorescent protein; Hpo, Hippo; inh., inhibited; IR, Inverted Repeat Hairpain RNAi; KA, kinase-active; KD, kinase-dead; PDK1, phosphoinositide-dependent protein kinase 1; PI3K, phosphatidyl inositol-3-kinase; PTEN, phosphatase and tensin homolog; Rheb, Ras-homolog enriched in the brain; RNAi, RNA interference; tj.Gal4, trafficjam.Gal4; TOR, Target of Rapamycin; TORC1, TOR complex 1; UAS, Upstream activator sequence; Wts, Warts; Yki, Yorkie.

https://doi.org/10.1371/journal.pbio.3000509.g004

To confirm our findings in another Drosophila tissue, we examined the proliferating Drosophila larval wing imaginal disc, particularly the wing pouch, where the Hippo pathway has been shown to respond to gradients of Fat–Dachsous signalling [4–8] as well as physiological mechanical inputs, Ajuba–Wts-mediated stress sensing in the columnar cells of the wing pouch, and Hippo-pathway–mediated strain sensing in the overlying peripodial cells [89, 94] (S5–S7 Figs). Since experimental manipulation of IIS has been reported to influence Hippo pathway signalling via PI3K–PDK1 in the wing disc [90], we wondered whether physiological changes in IIS might affect Yki nuclear localisation during wing development. Towards the end of the third instar of larval development, feeding ceases, and larvae begin to wander out of the food as they prepare to form pupae and undergo metamorphosis. Wing disc cell proliferation slows and subsequently arrests at this point, with final size determined by a combination of reduced IIS from the brain and a surge of ecdysone signalling that triggers pupariation and metamorphosis [21, 23, 105]. We tested whether this physiological change in IIS might regulate Yki–GFP as the wing disc slows and arrests proliferation at the end of the third instar. We find that nuclear Yki–GFP localisation is prominent during early stages but is reduced by the late third instar stage when cessation of larval feeding occurs (Fig 5A and S7 Fig). Reducing Akt signalling by expression of akt–RNAi or subjecting early third instar larvae to nutrient restriction for 24 h was sufficient to reduce Yki–GFP nuclear localisation (Fig 5B). Furthermore, the pattern of Yki nuclear localisation in early third instar wing discs can be maintained in late third instar wing discs by expressing insulin receptor (UAS.InR) or constitutively active forms of PI3K (UAS.dp110caax) or Akt (UAS.Akt-myr) in the posterior compartment with the hedgehog.Gal4 (hh.Gal4) driver (Fig 5C). We confirm that hh.Gal4-driven manipulation of PI3K–Akt signalling and manipulation of Yki activity in wing discs generates similar tissue growth phenotypes in adult wings, illustrating the importance of cell proliferation control in larval imaginal discs for determination of final tissue size (Fig 5D). These results support the notion that the interplay of Fat–Dachsous signalling and mechanical cues with nutritionally induced IIS, acting via PI3K–PDK1–Akt, controls Yki nuclear translocation to promote tissue growth in the Drosophila wing.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Yki becomes nuclear in response to both mechanical inputs and nutritionally induced insulin/IGF-1 signals in the developing wing.

A) Yki–GFP localises to the nucleus in early second and third instar larval wing imaginal discs in a pattern that closely resembles that of mechanical stress and strain (see also S7 Fig). Yki–GFP localises much less strongly to the nucleus in late ‘wandering’ third instar larval wing imaginal discs. Note the cytoplasmic localisation in the central region and weak nuclear localisation at the dorsal ventral boundary and at the periphery at late L3. Quantification is shown on right (n > 10 per time point). Scale bars approximately 5 μM. See supplementary file S1_Data.xlsx for underlying data. B) Silencing of akt expression by UAS.akt-RNAi hairpin transgene prevents nuclear Yki–GFP localisation in the L2 wing disc. Reducing Akt signalling by nutrient restriction for 24 h also prevents nuclear localisation of Yki–GFP at early L3. Quantification is shown on right (n > 10 per time point). Scale bars approximately 5 μM. See supplementary file S1_Data.xlsx for underlying data. C) Activating PI3K–Akt signalling by overexpression of insulin receptor (hh.Gal4 UAS.InR), a membrane-targeted version of the PI3K catalytic subunit (hh.Gal4 UAS.dp110caax), or membrane-targeted Akt (hh.Gal4 UAS.akt-myr) in the P compartment is sufficient to maintain the early pattern of Yki–GFP nuclear localisation in the posterior compartment of late third instar wing discs. Compare A and P compartments. Scale bars approximately 5 μM. D) Adult wings showing similar tissue overgrowth and undergrowth phenotypes in the posterior compartment when hh.Gal4 is used to silence expression of pten or hpo (which causes P-compartment overgrowth) as well as akt and yki (which causes P-compartment undergrowth). A, anterior; AEL, after egg laying; dp110caax, Drosophila p110-caax; GFP, green fluorescent protein; hh.Gal4, hedgehog.Gal4; Hpo, Hippo; IGF-1, insulin-like growth factor 1; InR, insulin/IGF-1 receptor; n/c, nuclear/cytoplasmic ratio; P, posterior; PI3K, phosphatidyl inositol-3-kinase; PTEN, phosphatase and tensin homolog; RNAi, RNA interference; UAS, upstream activator sequence; Yki, Yorkie.

https://doi.org/10.1371/journal.pbio.3000509.g005

A recent report suggested that IIS also acts via TOR to regulate Yki [96]. Interestingly, inhibition of TOR caused a strong ectopic localisation of Yki to the nucleus, even in late third instar imaginal discs [96]. We confirmed these findings with our Yki–GFP line, which strongly accumulates in the nucleus when dominant-negative UAS.TOR^TED is expressed in the posterior compartment of the wing disc (S8A Fig). The increase in nuclear Yki–GFP is due to disruption of the TORC1–Akt negative feedback loop [74] because the effects of UAS.TOR^TED are reversed by treatment with PDK1 inhibitors (S8B Fig). Thus, IIS promotes cell proliferation in the wing disc via PI3K–PDK1–Akt inhibition of the Hippo pathway and additionally promotes cell growth via TORC1, which also provides negative-feedback–limit Akt signalling. Accordingly, ectopic activation of IIS drives both cell growth and division, producing larger tissues with larger cells [3, 106], while ectopic activation of Yki produces larger tissues composed of increased numbers of normally sized cells [82].

We note that some differences exist between the spatiotemporal pattern of Yki–GFP localisation and the pattern of the Yki reporter genes ex.lacZ and bantam.lacZ (ban.lacZ) in the wing disc (Fig 6A). For example, the strong nuclear localisation of Yki–GFP at the Dorsal–Ventral (D–V) boundary is not reflected in either ex.lacZ or ban.lacZ expression, and both target genes are instead repressed at the D–V boundary (Fig 6A). Recent results explain this discrepancy because Notch signalling at the D–V boundary was found to potently repress the ability of activated Yki to transcriptionally induce target gene expression in the wing disc [107]. Thus, the combination of Fat–Dachsous signalling and mechanical and nutritional inputs determines Yki nuclear localisation in the wing, while other signalling pathways, such as Notch, can modulate the transcriptional output of Yki–Sd on specific target genes. Nevertheless, the resulting pattern of ex.lacZ or ban.lacZ—low medially and high in a circumferential pattern at the periphery (Fig 6A)—matches the pattern of Fat–Dachsous signalling and mechanical stretching previously proposed to complement the action of compartment boundary organisers [108, 109] and thus level out the rate of cell proliferation across the wing pouch [110–114].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Patterns of Sd–GFP, Yki–GFP, and Yki target genes in the wing disc and loss of nuclear Yki–GFP in akt³ clones.

A) Wing imaginal discs at early, mid, and late L3 showing expression patterns for Sd–GFP, Yki–GFP and two Yki target genes: ban.lacZ and ex.lacZ. Note the correlation between Yki–GFP nuclear localisation and its target genes at early L3, at which both are elevated in a circumferential pattern. Note also that the pattern of lacZ (beta-galactosidase) staining is maintained into late-stage L3 because of the perdurance of beta-galactosidase protein. Finally, both ban.lacZ and ex.lacZ target genes are normally repressed at the D–V boundary by Notch signalling, which prevents nuclear Yki–GFP from inducing them in this territory of the wing. Thus, only the circumferential activation of Yki–GFP is normally functional in the wing pouch, consistent with its proposed role in wing growth. B) Late L3 wing imaginal disc peripodial epithelial cells are strongly stretched and retain nuclear Yki–GFP. Note that this nuclear Yki–GFP localisation still depends on the low levels of Akt signalling occurring at this stage because the localisation becomes cytoplasmic in akt³ mutant clones (marked by the absence of nlsRFP). ban.lacZ, bantam-lacZ; D–V, Dorsal–Ventral; ex.lacZ, expanded.lacZ; GFP, green fluorescent protein; nlsRFP, nuclear red fluorescent protein; Sd, Scalloped; Yki, Yorkie.

https://doi.org/10.1371/journal.pbio.3000509.g006

To examine the mechanism by which Akt regulates the Hippo pathway in Drosophila, we tested whether a conserved Akt consensus phosphorylation site first identified in the human Hpo kinase homolog Mammalian Sterile 20 kinase (MST)2 in vitro [115–118] is required for control of tissue growth in Drosophila. The Akt phosphorylation site (T132) is located two amino acids downstream of the catalytic aspartate in the Hpo (and in human MST1/2) kinase domains, suggesting that Akt phosphorylation of might directly inhibit Hpo kinase activity (S9A–S9D Fig). A pan-Akt substrate phosphospecific antibody recognises the monomeric form of the Hpo kinase but not the active dimer, suggesting that phosphorylation by Akt may prevent dimerisation of Hpo (S9B Fig). We therefore tested whether mutation of the T132 residue would cause ectopic activation of the Hpo kinase in vivo. Using a landing site system for controlled expression of transgenes, we compared the effect of expressing wild-type Hpo versus phosphomutant HpoT132A in follicle cells. We find that wild-type Hpo mildly reduces the number of follicle cells, while HpoT132A dramatically reduces follicle cell numbers, exposing many gaps in the coverage of the germline (S9E Fig). Accordingly, Yki–GFP nuclear localisation is strongly inhibited by HpoT132A expression, even in stretched cells (S9E Fig). In the wing, we find that phosphomutant HpoT132A dramatically inhibits tissue growth compared with wild-type Hpo expression (S9F, S9G and S10 Figs). These results suggest that this highly conserved Akt phosphorylation site in Hpo–MST1/2 kinases (S11 Fig) may be functional in vivo, although Akt may also employ other mechanisms to regulate Hippo pathway signalling to control cell proliferation and tissue growth.

Because the Hippo pathway appears to effectively integrate PI3K–Akt signalling with either mechanical stress sensing (via Rho-kinase [Rok]–Ajuba–Wts) in the columnar cells of the wing disc or mechanical strain sensing (via reduced Hpo kinase activation) in follicle cell epithelium [89, 94], we sought to confirm that the Rok–Ajuba–Wts mechanism is indeed dispensable in follicle cells. Accordingly, mutant clones for rok² or ajuba^ΔII induced in the follicle cell epithelium (marked by absence of red fluorescent protein [RFP]) has no effect on nuclear localisation of Yki–GFP in mechanically stretched follicle cells (S12 Fig). These results confirm that the Rok–Ajuba pathway is not required for regulation of the Hippo pathway by mechanical strain, and thus, that the Hippo pathway uses distinct mechanisms to sense stress and strain in Drosophila. Nevertheless, an input from PI3K–Akt is essential for Yki activation in response to either stress or strain, regardless of the mechanisms involved.

In mammals, the skin epithelium is an example of a tissue in which both IIS via Akt [119–125] and mechanically regulated YAP and Transcriptional Activator with a PDZ domain (TAZ) [126–130] are known to promote keratinocyte cell proliferation, tissue regeneration after wounding, and tumour formation. Previous work identified a key role for PI3K and PDK1 in promoting nuclear localisation of YAP in skin keratinocytes [126] and other cell types [131]. We now find that treatment with Akt inhibitor is sufficient to prevent YAP nuclear localisation in keratinocytes, even when plated at a low density to induce mechanical stimulation via integrin adhesions (Fig 7A). We further find that treatment of densely packed keratinocytes (with cytoplasmic YAP) with IGF-1 is sufficient to promote YAP nuclear localisation in all but the most crowded cells, similar to inhibition of the Hpo orthologs MST1/2 (Fig 7B). Thus, activation of PI3K–Akt has a similar effect on YAP as inhibition of Hippo pathway signalling in these cells. Finally, in mouse skin, activation of Akt by skin-specific deletion of PTEN is sufficient to promote YAP nuclear localisation and tumour-like tissue overgrowth, similar to that induced by skin-specific expression of nuclear-localised YAP^5SA (Fig 7C). Together, these results indicate that IIS via Akt inhibits the canonical Hippo pathway to promote nuclear localisation of YAP and tissue growth in mammalian skin.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. YAP integrates mechanical cues with PI3K–Akt signalling to promote cell proliferation in mammalian skin.

A) Human skin keratinocytes (HaCaT) cultured in sparse conditions to induce nuclear localisation of YAP were treated with either control (DMSO) or Akt inhibitor (MK2206) at 1 μM for 4 hours prior to fixation and immunostaining for YAP, which becomes cytoplasmic upon Akt inhibition. B) Human skin keratinocytes (HaCaT) cultured in dense conditions to induce cytoplasmic localisation of YAP were treated with control (DMSO), IGF-1, or MST1/2 kinase inhibitor (XMU-MP-1) for 4 hours prior to fixation and immunostaining for YAP, which becomes nuclear upon IGF-1 activation or MST1/2 inhibition. C) Mouse skin from wild-type, skin-specific conditionally induced PTEN knockout or skin-specific conditionally induced nuclear NLS–YAP–5SA–overexpressing animals were immunostained for YAP (brown). Note similar nuclear localisation of YAP and tumour-like tissue overgrowth in both PTEN knockout and NLS–YAP–5SA skin. IGF-1, insulin-like growth factor 1; inh., inhibited; MST, Mammalian Sterile 20 kinase; NLS, nuclear localisation signal; PI3K, phosphatidyl inositol-3-kinase; PTEN, phosphatase and tensin homolog; YAP, Yes-associated protein.

https://doi.org/10.1371/journal.pbio.3000509.g007

Discussion

Our results show that the Hippo pathway has a physiological function in integrating hormonal IIS with mechanical inputs and polarity cues to control tissue growth in two different proliferating epithelia of Drosophila as well as in mammalian skin. Although manipulation of the Hippo pathway was previously shown to strongly affect cell proliferation and tissue growth [4–6], as well as to respond to the Fat–Dachsous planar polarity pathway [4–8], mechanical stress [89, 91–94], and mechanical strain [88, 95], it was previously unclear whether the Hippo pathway was also physiologically regulated by hormonal insulin/IGF-1 signals to control tissue growth during development. Our findings support the view that Yki/YAP normally relocalise to the nucleus to drive cell proliferation in epithelial cells subject to both mechanical stimuli and growth factor signalling via the PI3K–PDK1–Akt pathway. Below, we discuss how the Hippo pathway integrates these cues in different tissue types and different animal species. We then consider how evolutionary changes in the structure of the epidermis between Drosophila and mammals impact the mechanisms involved [8].

In the Drosophila ovarian follicle cell epithelium, our results suggest that growth of the germline stretches the surrounding follicle cells to induce nuclear localisation Yki, which activates Sd-mediated transcription to promote cell proliferation (Figs 1 and 2). Note that previous investigation with ex.lacZ failed to identify an effect of germline growth on follicle cell Yki activity, most likely because of perdurance of the lacZ gene product, beta-galactosidase [16]. The primary mechanism of mechanotransduction in this tissue involves the canonical Hippo pathway, which is activated by Crumbs (Crb)-mediated clustering of apical Hpo kinase containing complexes to promote Hpo dimerisation and transactivation and inactivated by mechanical strain to decluster the same complexes [88]. In addition, our results show that follicle cells must receive an insulin/IGF-1 signal that activates PI3K–PDK1–Akt to inhibit the Hpo kinase, possibly via direct phosphorylation on a conserved Akt phosphorylation motif or via other mechanisms that are still to be elucidated (Figs 3 and 4; S9–S11 Figs). In the absence of sufficient PI3K–PDK1–Akt signalling in follicle cells, the Hpo kinase becomes abnormally active, and Yki is retained in the cytoplasm even following mechanical strain (Fig 4). We note that PI3K–PDK1–Akt signalling directly affects Hpo kinase activity but that Hpo kinase activity itself promotes dimerisation because a kinase-active Hpo dimerisation sensor is much more strongly dimerised than the kinase-dead version [88]. Thus, the dimerisation and transactivation mechanism of the Hpo kinase enables it to integrate both mechanical inputs and hormonal insulin/IGF-1 signals in follicle cells. We note that the Hippo pathway continuously senses mechanical strain in the follicle cell epithelium [88] but only controls cell proliferation during the early stages (1–7) of oogenesis (Figs 1–3), after which Notch receptor and ecdysone receptor (EcR) activation terminates proliferation by inducing a switch to endoreplication [132, 133].

During Drosophila wing disc growth, cessation of larval feeding in the late third instar stage reduces IIS and leads to decreased Yki nuclear localisation as cell proliferation first slows and then ultimately arrests upon ecdysone-induced metamorphosis (Fig 5A). Our findings confirm previous work linking IIS via PI3K and PDK1 with regulation of the Hippo pathway in the wing [90] and extend it by providing a molecular mechanism—Akt-mediated regulation of Hpo signalling—and demonstrating the physiological significance of this mechanism in tissue growth control (Figs 5A–5D, 6A and 6B). Our results confirm that nuclear localisation of Yki occurs in both mechanically stressed cells (domains with high Rok and Myosin-II [Myo-II] along the D–V boundary) and in mechanically strained cells (strongly flattened peripodial cells), as well as cells with high Fat–Dachsous signalling (at the periphery of the wing pouch) (Figs 5A–5C, 6A and 6B; S7 Fig). Stress sensing occurs via Rok and Myo-II, which are known to promote localisation of the Wts/Large tumour suppressor (LATS) inhibitor Ajuba/LIM domain 1 (LIMD1)/Thyroid receptor interacting protein 6 (TRIP6) to adherens junctions, which promotes Yki activity in the wing disc [89, 94, 134–136] (S7B Fig). Strain sensing in peripodial cells and also follicle cells occurs via a distinct mechanism involving dilution of apical Hippo pathway components [88, 137], and thus, follicle cells do not require Rok or Ajuba to activate Yki (S12 Fig). Note that the strong Ajuba-driven nuclear localisation of Yki along the D–V boundary of the wing disc does not actually lead to pro-proliferative target gene activation because of antagonism by Notch signalling [107]. Thus, the Yki target gene bantam, which, along with Myc, mediates much of the effect on Yki on cell proliferation [87, 138, 139], is activated in a circumferential pattern at the periphery of the wing pouch and repressed medially during the early larval stages (Fig 6A)—in agreement with proposed models for mechanical stretch-induced proliferation in this tissue [110–114] and with gradients of Fat–Dachsous signalling [4–8]. Note also that Ajuba-mediated Yki activation was recently proposed to decline over time in the wing disc because of reduced mechanical stress and/or increased compression of cells [89], and our finding that increased PI3K–Akt signalling restores Yki nuclear localisation (Fig 5C and S7B Fig) suggests that integration of both mechanics and hormonal input is necessary to explain the temporal regulation of Yki.

The role of Myc as a downstream target of PI3K–Akt and Yki is particularly interesting because Myc not only stimulates cell proliferation and drives cell competition but also provides positive feedback on PI3K–Akt signalling itself [140]. Thus, nutritional regulation of growth in Drosophila involves insulin/IGF-1–mediated signalling via PI3K–Akt, which can stabilise the Myc protein but also drive expression of the myc gene via activation of Yki–Sd (in the context of sufficient mechanical input) to sustain a feed-forward loop that drives further signalling and growth. Starvation-induced loss of IIS or insufficient mechanical input due to developmental defects might thus rapidly slow growth until nutrition is located or normally coordinated development is restored.

In mammalian skin, squamous epithelial cells lack a classical apical domain typical of columnar epithelia. Instead, basal layer cells contact the underlying extracellular matrix via basal integrin adhesions, which are important for stem cell proliferation [141–143] and to promote YAP/TAZ activation via multiple mechanisms, including activation of Focal Adhesion Kinase (FAK) and Rous Sarcoma Virus Oncogene (Src)-family kinase signalling [7, 8, 93, 126]. Src can directly phosphorylate YAP on three tyrosine residues to promote its activation [144] and was also reported to inhibit the LATS1 kinase in mammals [145]. Integrin signalling can also activate PAK1 to inhibit membrane association of the key upstream Hippo pathway activator Merlin/Neurofibromatosis2 (NF2) [146–153]. Notably, integrins are generally dispensable for regulation of the Hippo pathway in Drosophila species, which have a columnar epidermis and can thus regulate Hippo pathway signalling via the apical domain as described above [8]. Accordingly, Drosophila Yki lacks the tyrosine residues phosphorylated by Src in mammalian YAP, and Drosophila Merlin lacks the PAK1 phosphorylation site from mammalian Merlin/NF2.

Despite this divergence in mechanotransduction mechanisms between Drosophila and mammals, the Akt site in Hpo is conserved in MST1/2 [115–118], and our results indicate that integration of mechanical signalling and IIS is also a fundamentally conserved feature across metazoa. As in Drosophila, growth-factor–induced PI3K signalling is necessary for nuclear localisation of mammalian YAP [93, 126, 131], and our data show that this effect requires Akt activity and that IGF-1 is sufficient to promote nuclear localisation of YAP in response to mechanical stimulation in human keratinocytes (Fig 7). In mice, activation of PI3K–PDK1–Akt in conditional knockout PTEN^flox/flox skin [154] also promotes YAP nuclear localisation and skin proliferation, similar to conditionally induced expression of activated form of YAP (nlsYAP^5SA) (Fig 7), confirming that YAP acts as an integrator of mechanical cues and IIS via PI3K–PDK1–Akt in skin cells both in culture and in vivo.

In conclusion, a key physiological function for the Hippo pathway is to integrate input from epithelial cell polarity determinants, mechanical cues, and hormonal IIS to promote cell proliferation and tissue growth in both Drosophila epithelia and mammalian skin. Although mechanisms of mechanotransduction via the Hippo pathway differ in the monolayered columnar epidermis of Drosophila versus the stratified squamous epidermis of mammals, growth factor signalling acts via a common PI3K–PDK1–Akt-mediated mechanism to inhibit the Hippo pathway and thus promote tissue growth. Thus, the Hippo signalling pathway could be considered as one of the key effectors of Akt in tissue growth control, alongside inhibition of FOXO and the classical inhibition of TSC1/2 to activate TORC1 (S13 Fig). Notably, our results show that activation of Akt alone is not sufficient to activate Yki because some type of mechanical stimulus is also necessary. In the absence of significant mechanical inputs, Akt acts primarily via TORC1 and FOXO, and the postmitotic Drosophila pupal retina appears to exemplify this mode of action—a phenomenon that helped distinguish and delineate PI3K–Akt and Hippo–Yki as separate pathways in pioneering genetic screens. Nevertheless, the widespread coupling of PI3K–Akt and Hippo pathway signalling in many tissues is potentially of major significance for human cancer because activation of PI3K–Akt signalling (in PIK3CA or PTEN mutant cells) causes formation of highly proliferative malignant tumours, while mutation of TSC1 or TSC2 leads to formation of relatively benign tumours [75, 81]. Given the increasing evidence for a potent role of YAP/TAZ in malignant tumours, future work should examine whether the Hippo pathway might be the primary effector of PI3K–Akt signalling in human cancer and thus a critical target for therapy.

Materials and methods

Drosophila genotypes

The Drosophila genotypes used were as follows:

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Supporting information