The Hippo signalling pathway coordinates organ growth and limits developmental variability by controlling dilp8 expression

Coordination of organ growth during development is required to generate fit individuals with fixed proportions. We recently identified Drosophila Dilp8 as a key hormone in coupling organ growth with animal maturation. In addition, dilp8 mutant flies exhibit elevated fluctuating asymmetry (FA) demonstrating a function for Dilp8 in ensuring developmental stability. The signals regulating Dilp8 activity during normal development are not yet known. Here, we show that the transcriptional co-activators of the Hippo (Hpo) pathway, Yorkie (Yki, YAP/TAZ) and its DNA-binding partner Scalloped (Sd), directly regulate dilp8 expression through a Hpo-responsive element (HRE) in the dilp8 promoter. We further demonstrate that mutation of the HRE by genome-editing results in animals with increased FA, thereby mimicking full dilp8 loss of function. Therefore, our results indicate that growth coordination of organs is connected to their growth status through a feedback loop involving Hpo and Dilp8 signalling pathways.

T he classic reciprocal grafting experiments of Twitty and Schwind 1 on salamanders have established 480 years ago that organs follow autonomous growth programs. The principle of autonomous organ growth was later confirmed in various animals including insects. When young larval imaginal discs are transplanted in heterologous environments, they reach final sizes that are comparable to those achieved in situ 2 . To achieve individuals of correct size and proportions, mechanisms allowing organ growth to be coordinated and coupled with the developmental programme are essential. Evidence that such coordination mechanisms exist comes from studies showing that artificially slowing down growth of a subset of discs delays metamorphosis and reduces the growth rate of unperturbed discs, allowing perturbed discs to complete their growth programs before entry into metamorphosis [3][4][5] . We and others recently identified the hormone Dilp8 as a central organizer of organ growth coordination 6,7 . Dilp8 is secreted from abnormally growing organs and acts remotely on the central brain complex to delay entry into metamorphosis 6,7 . Dilp8 signals through the Leucine-rich repeat-containing G-protein-coupled receptor 3 (Lgr3) in a pair of bilateral brain (GCL) neurons activating a neuroendocrine circuit that ultimately suppresses synthesis of the moulting hormone ecdysone [8][9][10][11] . Interestingly, mutations in the dilp8 and lgr3 loci produce animals that exhibit FA. FA, measured as the variance between left and right bilateral traits within individuals, is an assessment of stochastic developmental variations 12 . Therefore, the dilp8/lgr3 axis carries additional function in adjusting organ size and ensuring developmental stability [7][8][9] . Reducing lgr3 levels in the GCL neurons recapitulates the lgr3 mutant phenotype consistent with organ growth being adjusted through a central relay 8 . The ability of dilp8 to fine-tune organ growth suggests that its expression should be controlled by signals central to organ size assessment mechanisms.
In this study, to gain insight into how dilp8 expression might be coupled with organ growth, we searched for regulators of dilp8 expression among 120 candidates recovered in a genetic screen for molecules coupling disc growth with developmental transitions 6 . The condition used for this screen corresponds to a disc-specific knockdown of the syntaxin Avalanche (Avl; rn4avl RNAi), which generates neoplastic growth and a Dilp8-dependent delay in larva-to-pupa transition. We inferred that reducing the function of molecules regulating dilp8 expression should rescue the delay of rn4avl RNAi animals. Indeed, altering JNK signalling efficiently rescues the delay by suppressing the upregulation of dilp8 transcription observed in rn4avl RNAi animals 6 . JNK signalling induces dilp8 transcription in response to various stresses, including wound healing and tumour formation. This likely represents an important checkpoint mechanism allowing the organism to recover before entering metamorphosis, but may not be important for coordinating organ growth in normally developing animals 6 .

Results
dilp8 expression requires the co-activators Yorkie and Scalloped. In addition to JNK signalling, we identified the Hpo pathway as an important regulator of dilp8 expression. The Hpo pathway is an important regulator of organ growth and is thought to play a central role in organ size assessment 13,14 . The core kinase module of the Hpo pathway includes the Hpo (Mst1/2 in humans) and Warts/Lats kinases, which suppress activation of the transcriptional co-activator Yorkie (Yki; YAP/TAZ in humans). When the Hpo pathway is inactive, Yki and its DNA-binding partner Scalloped (Sd) activate target genes and promote organ growth [15][16][17][18][19][20] . We observed that reducing levels of the transcriptional co-activators Yki or Sd efficiently rescues the developmental delay in rn4avl RNAi animals and normalizes dilp8 transcript levels (Fig. 1a,b). Given the substantial evidence that crosstalk takes place between the Hpo and JNK signalling pathways 21,22 , we tested whether Yki can regulate dilp8 expression independently of JNK signalling. Indeed, we found that dilp8 expression is still significantly upregulated by Yki overexpression in flies that are mutant for the JNK kinase Hemipterous (Hep; Fig. 1c-i). We next tested whether overexpression of yki is sufficient to activate dilp8 transcription. Using an enhanced green fluorescent protein trap inserted in the first intron of the dilp8 gene as a reporter for native dilp8 expression, we could observe increased levels of enhanced green fluorescent protein in yki-overexpressing clones (Fig. 2a,b). Consistent with this, Dilp8 protein levels were also elevated in these clones (Fig. 2c,d). In agreement with Yki-regulating gene expression through association with its co-activator Sd, we found that reducing Sd levels in Yki-overexpressing clones abolished the Yki-dependent upregulation of Dilp8 ( Fig. 2e-j). We next examined dilp8 levels in clones carrying mutations in genes encoding upstream components of the Hpo pathway including expanded (ex), hpo and warts (wts). As expected, reducing the activity of upstream Hpo pathway members, which induces Yki activation, also increases Dilp8 protein levels ( Fig. 2k-p).
A Hippo-responsive element controls dilp8 expression by Yki/Sd. Genome-wide CHIPseq analyses using anti-Yki antibodies recently identified a number of potential Yki target genes 17,23 . Interestingly, these CHIPseq data identified a 600-bp promoter fragment localized 1.5 kb upstream of the coding region in the dilp8 locus. Close examination of this fragment reveals three potential Sd-binding sites (hereafter referred to as Hpo-responsive element (HRE)) 23 , suggesting that dilp8 expression might be directly activated by a Yki/Sd heterodimer (see region map in Fig. 3a). To directly test this, we performed DNA pull-down assays by mixing lysates from cells expressing tagged Sd (Sd-Flag) and a 600bp DNA fragment of the dilp8 promoter region centred around the HRE (dilp8 promoter fragment: dilp8-PF). A region in the diap promoter that is known to bind Sd (diap-PF) was used as positive control. We found that Sd-Flag binds the dilp8-PF and diap-PF with similar efficiency (Fig. 3b). Moreover, the binding of Sd-Flag to the dilp8-PF is abolished on targeted mutation of the three putative Sd-binding sites (dilp8-PFD123; Fig. 3b, Supplementary  Fig. 1 and see the 'Methods' section). To test the functional relevance of this binding assay, we analysed the potential of each of these fragments to promote transcription of the luciferase reporter gene in the presence of Sd and Yki. Consistent with the DNA pulldown result, both diap-PF and dilp8-PF, but not dilp8-PFD123, were able to activate transcription in the presence of Yki and Sd (Fig. 3c). To study the transcriptional regulation of dilp8 by Yki in vivo, we next generated transgenic fly lines carrying constructs harbouring either the full dilp8 promoter (dilp8-full-prom), dilp8-PF, dilp8-PFD123 or intron 1 of the dilp8 gene used as a negative control, all fused to the lacZ coding sequence (Fig. 3a). We found that both dilp8-full-prom and dilp8-PF were able to promote lacZ expression in Yki-overexpressing clones ( Fig. 3d-o). Moreover, the ability of Yki to induce lacZ expression depends on the integrity of the HRE, since mutation of the three Sd-binding sites prevented all lacZ expression (Fig. 3p-s). Altogether these results are consistent with Yki-activating dilp8 gene expression through Sd bound to the HRE in the dilp8 promoter.
Yki/Sd ensures developmental stability through dilp8 expression. We ultimately wanted to know whether the Yki-dependent regulation of Dilp8 function is central to the mechanism allowing ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13505 organ growth coordination. To test the biological relevance of the HRE in the dilp8 promoter, we carried out gene editing using a combination of CRISPR-induced double-strand breaks and ends-out homologous recombination 24 to replace the endogenous dilp8-PF with either dilp8-PFD123 or a control dilp8-PF at the dilp8 locus (see Supplementary Figs 2 and 3 and see the 'Methods' section). As expected, Yki-dependent induction of dilp8 expression was efficiently suppressed in flies carrying mutations in the HRE (dilp8-PFD123; Fig. 4a-d). By contrast, JNK-dependent regulation of dilp8 expression was not compromised in this genetic background ( Fig. 4e-h). We observed that dilp8-PFD123 flies exhibit increased levels of FA, although flies carrying a replacement with the control dilp8-PF do not (Fig. 4i, Supplementary Fig. 4). Finally, wing size distribution is neither modified in dilp8-PFD123 flies, nor in dilp8KO/KO flies compared with control animals ( Supplementary Fig. 5). This suggests that the FA phenotype is not due to a disruption of a general size-control mechanism operating independently in each disc. Overall, our data demonstrate that Yki-dependent regulation of dilp8 expression plays a critical role in limiting developmental variability. It also contrasts with our observation that hep75 mutant flies do not exhibit elevated FA ( Supplementary Fig. 6). These results are consistent with a central role of Yki in coordinating organ growth and limiting developmental variability through its effect on dilp8 expression.

Discussion
On the basis of a flurry of experiments, different working models have been proposed to explain organ-autonomous size control operating in complex organisms. One unifying view is that the Hpo pathway plays an instrumental role in assessing and regulating organ growth. The ability of the Hpo pathway to integrate both short-and long-range signals makes it an ideal candidate for sensing organ size and regulate growth accordingly. Importantly, we demonstrate here, direct molecular and functional links between Hpo/Yki signalling and dilp8 expression during normal development. Furthermore, suppressing the sole input of Yki on dilp8 expression suffices to increase developmental instability and recapitulates the effect of a dilp8 loss-of-function mutation. The coupling between Hpo pathway activity and dilp8 expression allows the growth status of organs to be transduced into a Dilp8 signal. Dilp8 acts through its receptor Lgr3, residing in two pairs of bilateral neurons to control the levels of the steroid hormone ecdysone. This in turn affects the timing of maturation, allowing for a checkpoint mechanism on the developmental transition [8][9][10] . When ectopically expressed, Dilp8 also affects organ growth 6 , suggesting that it may also participate in a continuous feedback ensuring proper organ growth during development. Whether ecdysone acts downstream of Dilp8/Lgr3 to mediate such continuous feedback on organ growth remains to be determined. Several reports indicate that ecdysone is required for imaginal tissue growth both in flies and in Lepidoptera 4,5,25-26 , and previous data suggest that organ growth coordination relies on systemic effects mediated by ecdysone. Interestingly, Taiman, a co-activator of ecdysone receptor mediates Yki-dependent dilp8 expression, therefore suggesting that the ecdysone signal itself feeds back on dilp8 expression 27 .
On the basis of previous experimental data, two non-exclusive models can be proposed for organ growth coordination during development. In a developmental checkpoint model, each tissue autonomously follows its growth trajectory and deviations that may occur due to developmental noise are resolved at the end of the growth period by adjusting the time of the next developmental transition. Alternatively, continuous adaptation of growth rates and/or organ size could take place during development, allowing organs to adjust while growing. Although experimental evidence exists for both mechanisms in response to large growth disruptions, it is not clear whether they apply for small perturbations usually encountered by larval organs during development. Further analysis will be necessary to decipher how the Yki/Dilp8 coupling contributes to final organ size adjustment and limits developmental noise. Animals were reared at 25°C (or 26.5°C for Fai measurement) on fly food containing, per litre: 17 g inactivated yeast powder, 83 g corn flour, 10 g agar, 60 g white sugar and 4.6 g Nipagin.
Immunostainings of larval tissues. Tissues dissected from larvae in 1 Â phosphate-buffered saline (PBS (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 , 1.47 mM KH 2 PO 4 , pH 8)) at the indicated hours after egg deposition (AED) were fixed in 4% formaldehyde (Sigma) in PBS for 20 min at room temperature, washed in PBS containing 0.1% Triton-X-100 (PBT), blocked for 2 h in PBT containing 10% fetal bovine serum (PBS-TF) and incubated overnight with primary antibodies at 4°C. The next day, tissues were washed, blocked in PBS-TF and incubated with secondary antibodies at 1/500 dilution (Cy3-conjugated donkey   anti-rat or Alexa Fluor 647 goat anti-rat, Cy3-conjugated donkey anti-chicken from Jackson ImmunoResearch) for 2 h at room temperature. After washing, tissues were mounted in Vectashield containing DAPI for staining of DNA (Vector Labs). Fluorescence images were acquired using a Leica SP5 DS ( Â 20 and Â 40 objectives) and processed using Adobe Photoshop CS5 or Image J. Wing areas were measured using Image J.
Plasmids. dilp8, diap and actin promoter sequences were PCR-amplified from genomic DNA (DNA preparations made using DNeasy Blood and Tissue kit (Qiagen)) and cloned in the pENTR/D-Topo vector (Invitrogen) using the following primers: for pENTR-dilp8-PF: sense, 5 0 -GTA TGATAGCAGGGTTCTG TG-3 0 and antisense, 5 0 -AGG TTG TGA AAT TGT ATT TAT GAA TGA-3 0 , for pENTR-diap-PF: sense, 5 0 -ATT TGC CCT TGT CTG TAG TTG C-3 0 and antisense, 5 0 -ATC GAGCAATGGATAAACCGTGAG-3 0 , for pENTR-act-PF: sense, 5 0 -GGACTC GTCGTACTCCTGCTTG-3 0 and antisense, 5 0 -CCC ATC TAC GAG GGT TAT GC-3 0 . To generate the pENTR-dilp8-PF D123 plasmid, mutations were introduced in pENTR-dilp8-PF using the kit QuickChange site-directed mutagenesis kit (Stratagene) and the following primers: Dilp8 MutSd S1: dilp8 editing by accelerated homologous recombination. To generate a dilp8 null mutant, an approach combining the CRISPR technique and homologous recombination was used as described in ref. 24. Double-strand breaks were induced by the CRISPR technique using single-stranded guide (sg)RNAs and a Cas9-encoding plasmid. For optimal targeting of the dilp8 locus, sgRNA target sequences were selected as 20-nt sequences preceding an NGG PAM sequence in the genome (GN20GG). To generate pCFD4{dilp8 KO }, gRNAs targeting exon I and exon II in the dilp8 locus were cloned into the tandem gRNA expression vector, pCFD4 (kind gift from Simon Bullock (Addgene plasmid # 49411)), using the dilp8-5 0 -KO-pCFD4-FORWARD: 5 0 -TAT ATA GGA AAG ATA TCC GGG TGA ACT TCg cca cta aaa tga gtt caa GTT TTA GAG CTA GAA ATA GCA AG-3 0 and dilp8-3 0 -KO-pCFD4-REVERSE: 5 0 -ATT TTA ACT TGC TAT TTC TAG CTC TAA AAC tca ggc aac aga ctc cga tgac GAC GTT AAA TTG AAA ATA GGT C-3 0 (dilp8-specific sequences are in lower case) primers as described in http://www.crisprflydesign.org/wp-content/uploads/2014/06/Cloning-with-pCFD4.pdf. For homologous recombination, two homology arms were amplified from genomic DNA using the following primers: for homology arm I, sense: 5 0 -GGT ACC TGA TGG GCA GCT CGT CGG GCA GCT TC-3 0 and antisense: 5 0 -GCG GCC GCG TAA ATG GAT CTG TGT CCC TGG GA-3 0 , for homology arm II, sense: 5 0 -ACT AGT AAC TCA TTT TAG TGG CGT TCT AGG ATC A-3 0 and antisense: 5 0 -AGA TCT CAA ATG ATT ACC GCA TTA AAG CTA ATC AC-3 0 (in bold are the added restrictions sites used for cloning into pTV2). The resulting PCR products were digested and cloned into the pTV2 vector (kind gift from Cyril Alexandre and Jean-Paul Vincent, the Francis Crick Institute) to generate pTV2{dilp8 KO }. To facilitate homologous recombination, embryos were injected with pTV2{dilp8 KO } in the presence of pCFD4{dilp8 KO } and a CAS9containing plasmid. To confirm gene targeting the following primers were used: dilp8KO S: 5 0 -GGA CGG GTT AAC CAT TCA GCA AGT TAG-3 0 and dilp8KO Rev: 5 0 -CGC GAA CTC GAT GGA CCT TCT GTC GC-3 0 . After confirmed targeting, the majority of the targeting vector was removed by crossing to a strain expressing a constitutively active Cre (Bloomington stock BL1092; procedure outlined in ref. 24. The resulting strain is referred to as dilp8 KO in the manuscript. To generate a dilp8 mutant deleting a 600 bp fragment of the promoter encompassing the three scalloped-binding sites (dilp8 pTV2DPF ), gRNAs directed against sequences upstream and downstream of the scalloped-binding sites were  cloned into pCFD4 as described above using the following primers: dilp8 HR S: 5 0 -TAT ATA GGA AAG ATA TCC GGG TGA ACT TCg aac ttt cac tgt cgt ttc tgt tGT TTT AGA GCT AGA AAT AGC AAG-3 0 and dilp8 HR REV: 5 0 -ATT TTA  ACT TGC TAT TTC TAG CTC TAA AAC gtg gcg cga ttc ttt aga tcG ACG TTA  AAT TGA AAA TAG GTC-3  Quantitative RT-PCR. Larvae were collected at the indicated number of hours AED. Whole larvae or dissected larval wing discs were frozen in liquid nitrogen. Total RNA was extracted from whole larvae or dissected tissues using a QIAGEN RNeasy lipid tissue minikit or microkit according to the manufacturer's protocol. RNA samples (2 mg per reaction) were treated with DNase and reverse-transcribed using SuperScript II reverse transcriptase (Invitrogen), and the generated cDNAs were used for real-time RT-PCR (StepOne Plus; Applied Biosystems) using PowerSYBRGreen PCR mastermix (Applied Biosystems), with 8 ng of cDNA template and a primer concentration of 300 nM. Samples were normalized to levels of ribosomal protein (rp)49 transcript levels. Three separate biological samples were collected for each experiment and triplicate measurements were performed. Primers were designed using PrimerExpress software (Applied Biosystems) as follows: dilp8S 5 0 -CGA CAG AAG GTC CAT CGA GT-3 0 , dilp8R 5 0 -GTT TTG CCG GAT CCA AGT C-3 0 .
Western blotting. Proteins were resolved by SDS-PAGE using 12% gels (NuPAGE Novex gel, Invitrogen) and transferred electrophoretically to polyvinylidene difluoride membranes (Amersham). The membranes were incubated for 1 h in blocking buffer (PBS, 5% milk) and incubated overnight at 4°C in the same buffer containing primary antibodies at 1:1,000 dilutions (mouse anti-Flag F3165 (Sigma)). Membranes were washed three times in PBS-T, blocked for 1 h, and probed with secondary antibodies in blocking buffer for 1 h at room temperature. After three washes in PBS-T, chemiluminescence was observed using the ECL-Plus western blotting detection system (Amersham Biosciences). Images were generated using the Fujifilm Multi Gauge software. The uncropped western blot (for Fig. 3) can be found in Supplementary Fig. 1.
Data availability. All data are available in the article or its Supplementary Files or available from the authors on request.