Drosophila lowfat, a novel modulator of Fat signaling

Recent studies have linked together the action of several tumor suppressors into a Fat-Hippo-Warts signaling network (reviewed by Reddy and Irvine, 2008). These genes play a crucial role in growth control from Drosophila to mammals, as exemplified by the ever-increasing number of cancers that have been associated with mutations in pathway genes(Steinhardt et al., 2008; Zeng and Hong, 2008). Fat-Warts signaling regulates growth through a transcriptional co-activator protein, called Yorkie (Yki) in Drosophila and YAP in vertebrates(Dong et al., 2007; Huang et al., 2005). In addition, Fat influences a distinct planar cell polarity (PCP) pathway(reviewed by Reddy and Irvine,2008; Strutt,2008). Planar cell polarity is the polarization of cells within the plane of a tissue, and can include both polarized structures, like hairs and bristles, and polarized behaviors, such as cell division and cell intercalation (Strutt,2008).

Fat is a large member of the cadherin family, and acts as a transmembrane receptor (reviewed by Reddy and Irvine,2008). Fat influences the subcellular localization of both the unconventional myosin Dachs and the FERM-domain protein Expanded, and through these proteins ultimately regulates the kinase Warts(Bennett and Harvey, 2006; Cho et al., 2006; Feng and Irvine, 2007; Mao et al., 2006; Silva et al., 2006; Tyler and Baker, 2007; Willecke et al., 2006). Warts then inhibits Yki by phosphorylating it: phosphorylated Yki is retained in the cytoplasm, but unphosphorylated Yki enters the nucleus to promote the transcription of target genes (Dong et al.,2007; Oh and Irvine,2008; Zhao et al.,2007). The Fat PCP pathway is less well characterized, but it is partially dependent upon Dachs (Mao et al., 2006), and also involves Atrophin (Grunge), a transcriptional co-repressor that can bind to the Fat cytoplasmic domain(Fanto et al., 2003).

The only Fat ligand identified is Dachsous (Ds), which like Fat is a large,atypical cadherin (Clark et al.,1995), and which influences the phosphorylation of Fat by Discs overgrown (Feng and Irvine,2009; Sopko et al.,2009). ds mutants have phenotypes similar to, but weaker than, those of fat mutants, raising the possibility that there might be other ligands, or other means of regulating Fat. The Golgi kinase Four-jointed (Fj) also regulates Fat signaling, but presumably acts by modulating Fat-Ds interactions (Ishikawa et al., 2008; Reddy and Irvine, 2008). Intriguingly, the two known Fat pathway regulators(ds and fj) are expressed in gradients in developing tissues(Clark et al., 1995; Villano and Katz, 1995). The vectors (directions) of these gradients parallel vectors of PCP, and experimental manipulations of ds and fj indicate that, at least in some tissues, their graded expression can direct PCP(Adler et al., 1998; Casal et al., 2002; Simon, 2004; Strutt and Strutt, 2002; Yang et al., 2002; Zeidler et al., 1999). The graded expression of ds and fj also influences the transcriptional branch of the pathway and wing growth, but in this case it is the slope rather than the vector of their gradients that appears to be instructive (Cho et al., 2006; Cho and Irvine, 2004; Reddy and Irvine, 2008; Rogulja et al., 2008; Willecke et al., 2008).

Although thus far most components of Fat signaling have been identified through genetic studies in Drosophila, protein interaction screens are an alternative approach with which to identify components of signaling pathways. A genome-wide yeast two-hybrid screen identified the product of the CG13139 gene as both a candidate Fat-interacting protein and a candidate Ds-interacting protein (Giot et al., 2003). This gene, which we have named lowfat(lft), encodes a small protein of unknown structure and biochemical function. It shares sequence similarity with two vertebrate genes, Limb expression 1 (Lix1) and Lix1-like (Lix1l; Fig. 1A). Lix1 was first identified in chickens through a differential screen for genes expressed during early limb development (Swindell et al., 2001). Subsequent analysis in mice revealed that Lix1 is actually expressed more broadly(Moeller et al., 2002). Lix1l has been defined only by its sequence similarity to Lix1. The biological functions of these genes have not been described, although genetic mapping of a feline spinal muscular atrophy identified LIX1 as a candidate gene(Fyfe et al., 2006).

While a basic outline of Fat signaling has emerged, many steps remain poorly understood. Here, we show that lft is a modulator of Fat signaling, and identify a cellular requirement for Lft in establishing normal levels of both Fat and Ds. Our observations identify transcriptional regulation of lft as a potential mechanism for modulating Fat signaling through its post-translational regulation of Fat and Ds protein levels. We also establish human LIX1L as a functional homolog of Lft, and LIX1 and LIX1L as Fat-interacting proteins, thus identifying a likely cellular function of vertebrate Lix1 genes as modulators of Fat signaling. This linkage raises the possibility that other Fat pathway components could be candidate susceptibility loci for spinal muscular atrophy.

Unless otherwise noted, crosses were conducted at 25°C. Gal4 lines employed included ptc-Gal4, en-Gal4, act-Gal4[3^rdchromosome] and tub-Gal4[LL7]. ds and fat mutant stocks employed have been described previously(Cho et al., 2006; Cho and Irvine, 2004).

A null mutation in lft was created using ends-out homologous recombination-mediated gene targeting(Gong and Golic, 2003). The targeting vector included a 5000-bp left arm and a 3680-bp right arm,amplified by PCR from wild-type (Oregon-R) genomic DNA and cloned into pW25 (Gong and Golic,2003). The left arm 3′ end is 40-bp upstream of the lft start codon, and the right arm 5′ end is 16-bp upstream the lft stop codon. Third chromosome transgenic lines, W25-TG2 and W25-TG4, were crossed to hs-Flp; hs-I-SceI/TM3, and heat shocked at 38°C for one hour three days after egg laying. Progeny with mosaic eyes were crossed to hs-Flp-70 lines, and their progeny with non-mosaic eyes were balanced over CyO. Southern blotting and PCR were performed to confirm correct targeting. The targeted line lft^TG2 was used for all experiments.

Primers for creating the targeting construct were as follows:

Left arm, CG13139-960 5′-GGTCCATTGCGGCCGCGCTGCCTGCGAGCTACGGTGCTCAAAA-3′ and CG13139-5964 5′-GACGGTACCGGTTTCGGGTTTCGTTTTCAGCACAAA-3′;

Right arm, CG13139-7013 5′-TGAGGCGCGCCCGGCTACCATTGATGATTA-3′CG13139-10775 5′-CCGGACCGGGTGGAAGAAT-3′.

TILLING was performed by the Seattle TILLING Project(http://tilling.fhcrc.org). The screened region covered 1464 bp, including part of the promoter region and the first 214 codons. The primers sequences were 5′-TGGTCCGTTCTCCTGGATAAAATAAAAGTG-3′ (left primer) and 5′-ATTATCGTGCTCCCTGGCAATCCAAT-3′ (right primer).

For the creation of conventional mutant clones, lft^TG2FRT40A, ds^UAO71 FRT40A/CyO Kr-Gal4 UAS GFP, fat^G-rvFRT40A/CyO GFP, fat^G-rv lft^TG2 FRT40A/CyO GFP or ds^UA071 lft^TG2 FRT40A/CyO GFP were crossed to y w hs-FLP[122]; Ubi-GFP FRT40A/CyO.

For the creation of MARCM clones, lft^TG2 FRT40A;UAS-d:V5[9F], fat⁸ FRT40A; UAS-lft:FLAG[6] or ds^UAO71 FRT40A; UAS-lft:FLAG[6] were crossed to y w hs-FLP tub-Gal4 UAS-GFP/FM7; tub-Gal80 FRT40A/CyO.

For the examination of wing disc growth, en-Gal4 UAS-GFP/CyO;UAS-dcr2/TM6B flies were crossed to RNAi ds (vdrc36219), RNAi lft and RNAi ds (vdrc36219); RNAi lft /TM6B flies, and cultured at 28.5°C.

Two methods were used to establish transgenic lines expressing FLAG-tagged lft. For P-mediated transformation, pUAST-Flag:lft was created, and insertions were isolated on the second (UAS-FLAG:lft[H]) and third(UAS-FLAG:lft[G]/TM6B, UAS-FLAG:lft[F]/TM6B and UAS-FLAG:lft[6]/TM6B) chromosomes. In order to compare the activities of lft versus its mammalian homologs, we used phiC31-mediated site-specific integration to insert transgenes into the attP site at 68A(Groth et al., 2004). Plasmids pUASTattB-3xFlagCG13139, pUASTattB-3xFlagLIX1 and pUASTattB-LIX1L were used to create the transgenic fly lines UAS-FLAG:lft[attP68A],UAS-FLAG:LIX1[attP68A], and UAS-FLAG:LIX1L[attP68A],respectively.

To investigate the consequences of reducing lft on fatmRNA expression, en-Gal4, UAS-GFP/CyO; dcr2/TM6B flies were crossed to UAS-RNAi lft (NIG13139R-1), and, as a control, to UAS-RNAi fat(vdrc 9396), and cultured at 28.5°C. To investigate regulation of lft mRNA, en-Gal4, UAS-GFP/CyO; dcr2/TM6B flies were crossed to UAS-RNAi fat, UAS-RNAi warts (vdrc 9928), UAS-RNAi Notch (NIG 3936R-3) or w^- controls and cultured at 28.5°C. Regulation by Notch was also confirmed by crossing UAS-ECN:FLAG(dominant negative Notch) to ptc-Gal4. Knockdown of Wg signaling was lethal, so regulation by Wg signaling was investigated by crossing UAS-sgg to ptc-Gal4 UAS-GFP;UAS-Gal80^ts/TM6B. Flies were kept at 18°C, and then shifted to 29°C for 48 hours to allow expression of Sgg (GSK3β)before dissecting.

For rescue experiments, lft^TG2 FRT40A; tub-Gal4/TM6Bwas crossed to lft^TG2 FRT40A; UAS-FLAG:lft[F]/TM6B,lft^TG2 FRT40A; UAS-FLAG:lft[attP68A], lft^TG2 FRT40A;UAS-FLAG:LIX1[attP68A] or lft^TG2FRT40A;UAS-FLAG:LIX1L[attP68A].

To examine the influence of ft or ds mutant clones on FLAG:Lft localization, hs-Flp, arm-lacZ/CyO; act-Gal4/TM6B was crossed to ft⁸ FRT40A/CyO; UAS-FLAG:lft[G]/TM6B or ds^UA071 FRT40A /CyO; UAS-FLAG:lft[G]/TM6B.

Discs were fixed and stained as described previously(Cho and Irvine, 2004), using mouse anti-Wg [1:800, 4D4, Developmental Studies Hybridoma Bank (DSHB)], rat anti-Fat (1:400) (Feng and Irvine,2009), rat anti-Ds (1:200, M. Simon, Stanford University,Stanford, USA), mouse anti-V5 (1:400, Invitrogen), mouse anti-Flag (1:600,Sigma), mouse anti-Diap1 (1:500, gift of B. Hay, Cal Tech, Pasadena, USA), rat anti-Elav (1:20, 7E8A10, DSHB), mouse anti-Pros (1:50, MR1A, DSHB), goat anti-β-gal (1:1000, Biogenesis) and rat anti-E-Cad (1:40, DSHB). Fluorescent stains were captured on a Leica TCS-SP5 confocal laser scanning microscope. For horizontal sections, maximum projection using Leica software was employed to allow visualization of staining in different focal planes.

In situ hybridization was carried out as described previously(Rauskolb and Irvine, 1999). For lft, an antisense RNA probe derived from the full-length coding region of lft was used, and discs from lft^TG2were used as a negative control. For fat, an antisense RNA probe derived from cDNA encoding the intracellular domain of Fat was used, and a sense probe was used as a negative control.

Photoshop and Image J were used for measurements of wing areas and distances. Prism (Graphpad) was used for statistical analyses.

Co-immunoprecipitation experiments were performed as described previously(Cho et al., 2006), using cell lysates prepared in radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% NP40, 0.1% SDS, 0.5% Na deoxycholate). Cell debris was precipitated by centrifugation with a table-top centrifuge at 15,700 g for 15 minutes. The supernatant was mixed with anti-FLAG M2 beads (Sigma); after overnight incubation, beads were washed seven times with RIPA buffer and then boiled in SDS-PAGE loading buffer.

Wing imaginal discs for the western blotting experiment were collected from wild-type (w^-), lft^TG2, and the progeny of act-Gal4/TM6B crossed to UAS-FLAG:lft[F],UAS-FLAG:lft[6]. Flies were allowed to lay for 5-6 hours, and wing discs were collected 96 hours later. Wing discs were dissected in ice-cold HyQ CCM3 serum-free medium (Hyclone, catalog number SH30065.01), and approximately 30 discs were pelleted at 1500 g for 4 minutes and then flash frozen in dry ice/ethanol and stored at -80°C.

For chemiluminescence western blotting, we used mouse anti-V5-HRP (1:6000,Invitrogen), mouse anti-FLAG M2-HRP (1:100,000, Sigma), mouse anti-α-Tubulin (1:4000, Sigma) and rat anti-Fat (1:4000). For quantitative western blotting, immunofluorescent secondary antibodies were used [anti-mouse IgG IRDye700 (LiCor) and anti-rat IgG IRDye800 (Rockwell)],and gels were captured on a Li-Cor Odyssey infrared imaging system and analyzed using Li-Cor software.

pUAST-Fat-TM-ICD:V5 was constructed from pUAST-fat-STI-4(Feng and Irvine, 2009) by digesting with KpnI and XbaI to remove an existing triple epitope tag, and then ligating with oligonucleotides(5′-CGGTAAGCCTATCCCTAACCCTCTCCTCGGTCTCGATTCTACGCGTACCGGTCATCATCACCATCACCATTGAGTTTAAGAATTCT-3′and 5′-CTAGAGAATTCTTAAACTCAATGGTGATGGTGATGATGACCGGTACGCGTAGAATCGAGACCGAGGAGAGGGTTAGGGATAGGCTTACCGGTAC-3′)to insert V5 and His tags.

pUAST-Ds-TM-ICD:V5 was constructed by PCR amplifying the Ds transmembrane and intracellular domains from genomic DNA (using the forward primer 5′-GCCTTTCCGCGAAGAAGAGCCGGTGGTTCGTCAAGTGGTTCCATT-3′ and the reverse primer 5′-GCAGGTACCCATCCGTGTCCCCACATTTCCCCTCTGACTT-3′). The PCR product was digested with SapI and KpnI, and ligated into SapI/KpnI cut pUAST-fatSTI-4, to create a fusion gene utilizing the Fat signal peptide but the Ds transmembrane and cytoplasmic domains. The C-terminal tags were then exchanged as described above for Fat-TM-ICD-V5.

pUAST-TM-EGFP:V5 was constructed by PCR amplifying EGFP from pmaxEGFP(Amaxa) using the forward primer 5′-GCACCGCGGAACTAGTGCCACCATGCCCGCCATGAA-3′ (adding a SacII site) and the reverse primer 5′-GCAGGTACCTCGAGCTCGAGATCTGGCGAA-3′ (adding a KpnI site), and digesting the PCR product with SacII/KpnI. This fragment was then cloned into SacII/KpnI cut pUAST-Ft-TM-ICD, which leaves the transmembrane domain and five amino acids of the predicted Fat cytoplasmic domain. The C-terminal tags were then exchanged as described above.

pUAST-FAT4-TM-ICD:V5 was constructed from pUAST-FAT4-TM-ICD:FLAG (Y. Feng)using a PCR product (forward primer,5′-CTGAAGCCTCGAAGGTACCACGGTCGCAGGGCC-3′; reverse primer,5′-GGGGTACCTCAACCGGTACGCGTAGAATCGAGACCGAGGAGAGGGTTAGGGATAGGCTTACCCACATACTGTTCTGCT-3′)to exchange the existing FLAG tag for a V5 tag.

pUAST-Fat-TM-ICD-ΔC:V5 was constructed by PCR amplifying the portion of the fat intracellular domain to be retained (using the forward primer, GGGAATTCGTTAACAGATCTGCGGCCGCATGGAGAGGCTA and the reverse primer,TCTAGATTATCAACCGGTACGC -GTAGAATCGAGACCGAGGAGAGGGTTAGGGATAGGCTTACCCTCGAATCCATCGTA), digesting with EcoRI and XbaI, and then using this fragment to replace the corresponding region of Fat-TM-ICD:V5. The resulting construct lacks the C-terminal 99 codons of Fat.

pUAST-TM-EGFP+Ft-C:V5 was constructed by PCR amplifying the C-terminal 99 codons from pUAST-Fat-TM-ICD:V5 (using the forward primer,GGGGTACCCTGGCCGCCGCCTCATCATTTCGCGGAT and the reverse primer,GGGGTACCTCCCACGTACTCCTCTGGAGCC). This PCR product was then digested with KpnI, and ligated into KpnI cut pUAST-EGFP:V5.

lft constructs were generated from a full-length cDNA, amplified by RT-PCR from wild-type (Oregon-R) larvae using a one-step RT-PCR kit (Qiagen) (using CG13139-UPinfrm,5′-GTACCCGGGGATGGTCTATCCCGAAGAACCTTTT-3′ and CG13139-lower,5′-CCGGCTGCAGTTAATCATCAATGGTAGCCGAGTTAA-3′). This PCR product,together with a triple FLAG epitope tag at the 5′ end, was cloned into pUAST and pUASTattB using XhoI and XbaI sites. The constructed plasmids were named pUAST-FlagCG13139 and pUASTattB-3xFlagCG13139,respectively. Human cDNAs of LIX1 and LIX1-like were obtained from the ATCC and cloned by PCR (using the primers hlix1up,5′-GACGGTACCAGGCCTATGGACAGAACCTTGGAATCTCT-3′ and hlix1lw,5′-GACGCTAGCGGGCTTGGCCTTGCTAGTGATA-3′ for human LIX1; and hlix1Lup, 5′-GACGGTACCAGGCCTATGGAGACTATGCGAGCGCA-3′ and hlix1Llw,5′-GACGCTAGCGGGTGGATGCCTAGCAGTTGGAA-3′ for human LIX1-like) into pUAST-FlagCG13139 using KpnI and NheI/XbaI sites, replacing the lft insertion. The constructed plasmids were named pUASTattB-lix1 and pUASTattB-lix1-like. All plasmid constructs were verified by DNA sequencing.

To evaluate whether the reported interaction between Lft and Fat and Ds(Giot et al., 2003) could be reproduced in Drosophila cells, epitope-tagged Lowfat protein(FLAG:Lft) was expressed in cultured S2 cells together with tagged fragments of Fat or Ds. As Lft was predicted to encode a cytoplasmic protein, we focused on examining interactions between Lft and polypeptides including the intracellular and transmembrane domains of Fat and Ds (Fat-TM-ICD:V5 and Ds-TM-ICD:V5, Fig. 1B), but excluding their extracellular domains. Immunoprecipitation of Lft:FLAG specifically and reproducibly co-precipitated Fat-TM-ICD:V5 or Ds-TM-ICD:V5,but not a control protein (TM-EGFP:V5; Fig. 1D). Thus Lft can bind to both Fat and Ds in Drosophilacells.

Pair-wise BLASTP analysis of the Fat and Ds cytoplasmic domains identified a small region of similarity between them(Fig. 1C)(Clark et al., 1995). Deletion of the C-terminal 99 amino acids of Fat (Fat-TM-ICD-ΔC:V5), which includes this region, substantially reduced Fat-Lft binding(Fig. 1D), implying that this region contributes to their physical association. However, as binding was not completely eliminated, the interaction between Lft and Fat apparently also involves additional regions of the cytoplasmic domain. Nonetheless, the Fat C-terminal region makes a crucial contribution to the association with Lft, as its addition onto GFP (TM-EGFP+Ft-C:V5) conferred to this protein a modest but reproducible ability to bind Lft (Fig. 1D). Thus, Lft is a Fat- and Ds-binding protein, and this binding is mediated in part through the C terminus of Fat, which exhibits some sequence similarity to a region of Ds.

To investigate biological requirements for lft, we first reduced lft expression by RNAi, using a UAS-hairpin transgene (UAS-RNAi lft; NIG-13139R-1). Ubiquitous expression of this lftRNAi transgene under act-Gal4 control resulted in flies with slightly shorter wings (Fig. 2C), but no evident phenotypes in other organs. The reduced length of the wing was most obvious in the middle, where the distance between the anterior and the posterior cross-veins was decreased (Fig. 2C,J). Reduction in the distance between cross-veins is a diagnostic Fat pathway phenotype, as it has been observed in viable alleles of all of the genes identified to date as functioning specifically within the Fat branch of the Fat-Hippo-Warts pathways [i.e. fat, ds, fj,approximated (app) and dachs](Mao et al., 2006; Matakatsu and Blair, 2008; Villano and Katz, 1995; Waddington, 1940). The observation of this phenotype with lft RNAi thus suggests that it is a component of the Fat pathway.

RNAi often only partially reduces gene function, hence we sought to isolate mutations in lft. Two strategies were used, both of which were successful. In one approach, we used homologous recombination-mediated gene targeting (Gong and Golic,2003) to create a lft allele in which the entire coding region was deleted (Fig. 2A). This deletion allele of lft (lft^TG2) is homozygous viable and fertile, and the only obvious phenotype was a reduced wing length and a shorter cross-vein distance(Fig. 2D,J). Measurements revealed an average wing area that was 82% of that in wild-type wings, and an average cross-vein distance that was 59% of that in wild-type wings(Fig. 2J, data not shown). This wing phenotype was stronger than the lft RNAi phenotype, and similar to that observed in null alleles of fj or app, or in hypomorphic alleles of fat or dachs. The reduced size of the wing implies that the regulation of wing growth by Fat signaling could be affected, which would suggest that there is an influence on Fat-Warts signaling. At the same time, the shape of the wing was also affected, as the length was affected more than the width, especially in the middle of the wing. Wing shape can be influenced by the Fat PCP pathway(Baena-Lopez et al., 2005). The orientation of wing hairs, however, which also reflects PCP, was not significantly affected in lft mutants(Fig. 3D). We also examined lft mutant clones for effects on PCP, or the transcription of downstream targets of Fat-Warts signaling, including Diap1, Wingless and Expanded, but no significant effects were observed (see Fig. S1 in the supplementary material; data not shown). Sequence analysis implied that there are no other lft-like genes in Drosophila. These observations suggest that lft could contribute to normal Fat signaling during wing development, but that the requirement for lftis relatively mild.

In parallel to the creation of a deletion allele of lft, we employed the Seattle TILLING Project(http://tilling.fhcrc.org/)to identify point mutations in lft. TILLING screens for nucleotide changes in mutagenized chromosomes regardless of phenotypic effect(Till et al., 2003). Seven mis-sense mutations in the lft coding region were identified by TILLING of a 1464-bp region, corresponding to the first 214 codons of lft. Two of these resulted in obvious wing phenotypes as transheterozygotes with lft^TG2(Fig. 2E,J). Measurements of the distance between cross-veins identified lft³⁷⁰⁹ as similar to lft^TG2, whereas lft³⁷⁶²exhibited a slightly milder reduction in cross-vein length. Another allele, lft⁰⁴⁵¹, exhibited an even weaker phenotype(Fig. 2J). All of these alleles change amino acids that are conserved among Lft and its human homologs LIX1 and LIX1L (Fig. 1A). The other four mis-sense mutations did not exhibit significant wing phenotypes. Lft and its vertebrate homologs are highly conserved, but structurally novel, and their biochemical function is unknown. The characterization of these TILLING alleles identified amino acids that are or are not required for normal Lft function independently of their evolutionary conservation.

Vertebrate Lix1 was first identified and named as a gene expressed in developing limbs (Swindell et al.,2001), but subsequent studies have revealed that it is also expressed elsewhere (Fyfe et al.,2006; Moeller et al.,2002). Expression of Drosophila lft was examined by in situ hybridization to mRNA. lft was broadly expressed in developing imaginal discs, including wing, leg and eye, and was also expressed within the neuroepithelia of the optic lobes of the brain(Fig. 4, data not shown). These are all places where fat and ds are expressed. Comparison with control imaginal discs from lft^TG2 mutants indicated that although lft is expressed throughout the wing and eye disc, the levels of expression vary. In the eye imaginal disc, lft expression is highest along the morphogenetic furrow(Fig. 4C), and, in the wing imaginal disc, lft expression is highest near the dorsoventral (DV)compartment boundary (Fig. 4A). The DV compartment boundary is a site of Notch activation and a source of Wg expression, and the upregulation of lft expression in the wing was eliminated by the downregulation of Notch or Wg signaling (see Fig. S2 in the supplementary material). By contrast, lft is not subject to feedback regulation by Fat signaling, as its expression was not affected by the downregulation of fat or warts (Fig. S2 in the supplementary material).

Fat is expressed broadly throughout imaginal discs, but its expression is not uniform. Consistent with earlier reports(Strutt and Strutt, 2002; Yang et al., 2002), we observed, using a Fat-specific sera (Feng and Irvine, 2009), that in the wing imaginal disc Fat protein staining is elevated in the region fated to give rise to the wing blade (the wing pouch), especially near the DV boundary, and that in the eye disc Fat staining is strongest near the morphogenetic furrow(Fig. 5A,C). Although fat mRNA distribution is also not uniform at late third instar (see Fig. S3 in the supplementary material)(Garoia et al., 2000), it does not match the strong increase in protein levels along the DV boundary or morphogenetic furrow in comparison to other regions of these discs, suggesting that Fat levels are regulated post-transcriptionally. The correlation between regions of imaginal discs in which lft expression is elevated and regions in which Fat protein staining is elevated raised the possibility that Lft might influence Fat protein levels or localization.

Indeed, Fat protein staining was clearly reduced in wing and eye imaginal discs from lft mutants (Fig. 5B,D), especially in regions where peak levels of Fat staining are observed in wild type. To provide a direct comparison between Fat levels in wild-type versus lft mutant cells, Fat staining was examined in discs with lft mutant clones. In eye discs, and in the wing pouch region of the wing disc, lft mutant clones were associated with a strong decrease in Fat levels (Fig. 5E,H). In the region of the disc fated to give rise to the wing hinge, lft mutant clones had little effect on Fat levels(Fig. 5G), although this apparently reflects Lft perdurance, as Fat levels could be affected by lft RNAi in the hinge (see Fig. S3 in the supplementary material),and also appeared to be reduced in the hinge within lft mutants(Fig. 5B). Thus, Lft increases Fat levels, and its effects are most obvious in regions where the highest levels of Fat and lft are normally observed.

To investigate whether Fat protein staining could also be influenced by increased Lft, a FLAG epitope-tagged UAS-lft transgene was created. Expression of UAS-lft under tub-Gal4 control rescued the wing phenotype of lft^TG2 mutants, confirming that FLAG:Lft provides Lft function (Fig. 2G,J). Expression of UAS-lft under ptc-Gal4control elevated Fat protein staining, especially in the hinge and notal regions of the wing disc, where endogenous levels of lft are relatively low (Fig. 6A,B). To confirm that the visible changes in Fat staining associated with mutation or overexpression of Lft are reflective of differences in Fat protein levels, Fat was examined by quantitative western blotting of lysates from wing discs. A 2.2-fold decrease in Fat levels was detected in lft mutant discs when compared with wild-type discs (Fig. 6J), which, because this is an average over the entire disc,underestimates the decrease in peak regions. A 3.0-fold increase in Fat protein levels occurred in discs overexpressing Lft under act-Gal4control (Fig. 6J).

To confirm that these effects of lft on Fat protein levels are post-transcriptional, fat mRNA levels were examined by in situ hybridization in discs in which lft levels were reduced by RNAi, or increased by overexpression. Expression of the lft RNAi construct under en-Gal4 control reduced Fat protein levels, but did not significantly reduce fat mRNA levels (see Fig. S3E,H in the supplementary material). Moreover, expression of lft under ptc-Gal4 control did not increase fat mRNA levels (see Fig. S3B,C in the supplementary material). Thus, the influence of Lft on Fat is post-transcriptional.

The observation that Lft binds to Ds as well as to Fat raised the possibility that Lft might also influence Ds levels. Indeed, although endogenous levels of Ds are quite low in the wing pouch, a reduction in Ds protein staining at the membrane could be observed within lft mutant wing clones (Fig. 5F), and also in eye disc clones (not shown). When Lft was overexpressed, Ds protein staining was increased in both the hinge and the pouch(Fig. 6I). The influence of Lft on Fat and Ds is specific, because mutation or overexpression of Lft did not detectably influence levels of E-cadherin or Notch (not shown). To confirm that Lft could independently influence both Fat and Ds, clones of cells overexpressing Lft but mutant for fat or ds were stained for expression of Ds or Fat, respectively. Strong upregulation of Fat, and weak upregulation of Ds, was observed in such clones within both the wing pouch and the wing hinge (see Fig. S4 in the supplementary material).

To gain further insight into the mechanism by which Lft influences Fat, we employed antibodies against the FLAG epitope tag to localize Lft expressed in imaginal discs from UAS-lft transgenes. Endogenous Fat and Ds proteins are preferentially localized to the sub-apical membrane, just apical to the adherens junctions. FLAG:Lft was detected at the sub-apical membrane,overlapping Fat and Ds staining, but was also distributed broadly throughout the cytoplasm (Fig. 6A,H). The profile of FLAG:Lft staining detected varied depending upon the expression level and the region of the disc. When expressed in the wing imaginal disc under ptc-Gal4 control, strong cytoplasmic staining of FLAG:Lft was detected in the wing pouch, but in parts of the wing hinge FLAG:Lft was preferentially detected at the sub-apical membrane(Fig. 6A,B,E). Because Ds is expressed at high levels in the wing hinge and low levels in the wing pouch,these differences suggest that the localization of FLAG:Lft to the sub-apical membrane could depend upon the availability of its binding partners. Indeed,localization of FLAG:Lft to the sub-apical membrane was reduced in fat or ds mutant clones (see Fig. S5 in the supplementary material). Conversely, when fat or ds were overexpressed under ptc-Gal4 control, FLAG:Lft levels were substantially increased(Fig. S5 in the supplementary material). Thus, Lft and its binding partners,Fat and Ds, have reciprocal effects on the levels and localization to the sub-apical membrane of one another.

Although the lft mutant phenotype is relatively mild, the Fat ligand ds also has a mutant phenotype that appears weaker than that of fat mutants. Intriguingly, lft and ds are expressed in partially complementary domains in wing discs, as ds is expressed at highest levels in proximal cells, whereas lft expression is highest in distal cells. Thus, we explored the consequences of loss of both lft and ds. ds mutant animals[ds^UA071/Df(2L)ED94] can survive to adulthood, but ds lft double mutant flies [ds^UA071lft^TG2/Df(2L)ED94 lft^TG2] did not survive. To determine whether an additive phenotype of lft and ds could also be detected for wing growth, we examined imaginal discs in which their levels were reduced by RNAi. By expressing UAS-RNAi transgenes specifically in the posterior (P) half of the disc under en-Gal4 control and comparing the relative sizes of the anterior (A) and P compartments, we could control for variations in developmental stage that might otherwise confound precise measurements of disc growth. In wild type, the P compartment of the wing disc was 80% of the size of the A compartment. Expression of lftRNAi under en-Gal4 control resulted in a modest, but statistically significant, increase in the relative size of the P compartment, to 87% of A compartment size (Fig. 7B,E). Expression of ds RNAi alone resulted in a large increase in P compartment size, to 140% of A compartment size(Fig. 7C,E). Co-expression of lft and ds RNAi lines enhanced the overgrowth of the P compartment to 178% of A compartment size(Fig. 7D,E). Thus, lftand ds have additive effects on wing disc growth.

We also examined lft and ds mutant clones for their effects on Fat protein staining. Mutation of ds had distinct effects on Fat in different regions of the disc. In the wing pouch, Fat staining appeared modestly elevated and slightly more diffuse within ds mutant clones (Fig. 5L). Nonetheless,preferential localization to the sub-apical membrane, which visibly outlines cells, remained. By contrast, in the wing hinge, Fat staining remained strong within ds mutant clones, but appeared diffusely distributed on the apical surface (Fig. 5K). This diffuse staining was surrounded by a one-cell-wide halo depleted of Fat staining, which presumably reflects a re-localization of Fat to the membrane at the outer edge of the clone, where it could be bound by Ds in neighboring wild-type cells (Cho and Irvine,2004; Ma et al.,2003; Strutt and Strutt,2002). lft mutant clones resulted in a strong reduction in Fat staining in the wing pouch, but the Fat protein that remained appeared to localize normally (Fig. 5H). lft mutant clones had no obvious effect on Fat staining in the hinge(Fig. 5G). In both the wing hinge and the wing pouch, ds lft double mutant clones exhibited additive effects on Fat staining. Fat was diffusely localized in the wing hinge and levels were reduced (Fig. 5I); Fat levels were also greatly reduced in the wing pouch(Fig. 5J). Similarly, fat and lft had additive effects on Ds localization in the wing pouch, as lft mutant clones reduced Ds levels in the wing pouch, fat mutant clones resulted in diffuse apical localization, and fat lft double mutant clones exhibited Ds staining that was both reduced and diffuse (see Fig. S6 in the supplementary material).

The mild phenotype of lft mutants, despite the substantial reduction in Fat protein levels, suggests that Fat protein is normally present in excess. However, we reasoned that if further reductions in Fat activity could be achieved, such that its levels were closer to the minimal thresholds needed for normal development, then lft mutants should exhibit stronger phenotypes. This was explored by investigating the phenotypes of animals doubly mutant for lft^TG2 and a weak allele of fat, fat¹. The distance between cross-veins was greatly reduced in fat¹ lft^TG2 double mutants(Fig. 3B). In addition the posterior cross-vein was incomplete, and the L2 longitudinal vein was often both incomplete and associated with ectopic vein material, phenotypes that are not observed in either single mutant. Leg growth is only very subtly affected in either lft^TG2 or fat¹ single mutants, but fat¹ lft^TG2 double mutants had shorter legs, and individual leg segments, including the femur and tibia were both shorter and broader (Fig. 3K-N). In addition, fat¹ lft^TG2double mutants had only four tarsal segments instead of the usual five(Fig. 3N), a phenotype that is characteristic of mutations in fat pathway genes. Finally, we did not observe PCP phenotypes in lft mutants, and fat¹mutants had only very subtle PCP phenotypes(Fig. 3)(Fanto et al., 2003), but obvious PCP phenotypes were observed in fat¹lft^TG2 double mutants in both wings and legs(Fig. 3F,J). Thus, under sensitized conditions, an influence of lft mutations on Fat signaling can be detected in multiple organs, and for both growth and PCP phenotypes.

Lft protein appears to be highly conserved with two human homologs, LIX1 and LIX1L. Although LIX1L differs from Lft and LIX1 in that it has a longer,unconserved, N-terminal region, within the central conserved region (amino acids 24-257 of Lft), Lft is more similar to LIX1L (75% amino acid identity)than it is to LIX1 (57% identity), or even than LIX1 is to LIX1L (61%identity; see Fig. 1A). The functional significance of these sequence similarities was examined both in vitro and in vivo.

In co-immunoprecipitation experiments, human LIX1 and LIX1L expressed in Drosophila S2 cells bound to the cytoplasmic domains of Fat and Ds(Fig. 1D). LIX1 and LIX1L binding appeared similar to Lft binding, and involved the same C-terminal region of Fat. LIX1, but not Lft or LIX1L, also appeared to be unstable when expressed without a binding partner in S2 cells, as it was barely detectable when co-expressed with GFP, but was readily detected when co-expressed with Fat or Ds (Fig. 1D). We also examined the ability of these proteins to bind to the cytoplasmic domain of human FAT4, which within its cytoplasmic domain is the closest of the four human FAT proteins to Drosophila Fat. LIX1, LIX1L and Lft could all co-precipitate FAT4 (Fig. 1D).

The interaction between LIX1 and LIX1L and Drosophila Fat was also investigated by comparing their influence on Fat protein levels to that of Lft. Transgenes expressing FLAG-tagged lft, LIX1 and LIX1Lunder UAS control were inserted into the same chromosomal location using phiC31-mediated integration (Groth et al.,2004), such that their expression levels would be similar. Expression of LIX1L under ptc-Gal4 control resulted in an upregulation of Fat protein staining, both in the wing hinge and in the wing pouch, similar to the effects of Lft (Fig. 6C,F). Expression of LIX1 also resulted in an upregulation of Fat protein staining in the hinge, but actually decreased Fat protein staining in the wing pouch (Fig. 6D,G). This apparently complex effect could be interpreted as indicating that LIX1 has weak Lft-like activity. Hence, we suggest that in the hinge, where Lft levels are lower, LIX1 elevates Fat levels by providing partial Lft activity,but in the pouch, where Lft levels are higher, it decreases Fat levels by competing with Lft. Like Drosophila FLAG:Lft, FLAG:LIX1 and FLAG:LIX1L could be detected at the sub-apical membrane, overlapping Fat and Ds staining (Fig. 6C-G). However, in the case of LIX1L, but not LIX1, we also detected strong cytoplasmic staining. Indeed, under identical expression and staining conditions, LIX1 protein was barely detectable in the wing pouch(Fig. 6G), suggesting that, as in S2 cells, it is unstable when not associated with a binding partner.

Finally, we examined the ability of human LIX1 and LIX1L to rescue the lft mutant phenotype. Expression of LIX1 under tub-Gal4control exhibited only a partial rescue of lft(Fig. 2I,J). However, LIX1L rescued the wing phenotype of lft mutants as well as did lftitself (Fig. 2H,J). Thus, human LIX1L is a functional homolog of Drosophila Lft. The difference in the extent of rescuing activity for LIX1 versus LIX1L correlates with their sequence similarity to Lft, and with their distinct effects on Fat protein staining.

Elucidation of the Fat signaling pathway requires the identification and characterization of pathway components. Here, we have identified Lft as a novel, highly conserved modulator of Fat signaling. lft mutants display decreased levels of both Fat and Ds protein staining, and presumably as a consequence exhibit a characteristic Fat pathway phenotype in the wing. In addition, lft can genetically interact with both fat and ds to cause more severe phenotypes. The lft mutant phenotype resembles weak mutant alleles of fat or ds, and lftmutants do not exhibit any additional phenotypes that could not be accounted for by effects on Fat signaling. The expression of lft itself is modulated by other signaling pathways, and differences in lftexpression levels correlate with differences in Fat and Ds protein levels both in wild-type animals, and when lft levels are experimentally increased or decreased. Thus, transcriptional regulation of lftdefines a mechanism for modulating Fat signaling.

Lft influences levels of both Fat and Ds. Because Fat and Ds in turn can influence levels of Lft, and because Fat and Ds also influence the localization of one another to the membrane, we infer that for any one of these three proteins, the influence that it has on the other two includes both direct effects, and indirect effects mediated through the third protein. In addition, the net effect observed for any one protein presumably also reflects the consequences of feedback regulation of its own levels via the other two proteins.

Given the substantial decrease in Fat staining in lft mutants, the phenotype appears surprisingly mild. This observation suggests that Fat is normally present in excess; for example, it could be that only a fraction of Fat is normally active, and that levels of Fat are not normally limiting for pathway activation. This hypothesis was supported by the observation of enhanced Fat pathway phenotypes in combination with fat¹,and would be consistent with the conclusion that Fat acts as a ligand-activated receptor, with only a fraction of Fat normally being present in the active form (Feng and Irvine,2009; Sopko et al.,2009). Complicating this simple explanation is the observation that the levels of the Fat ligand Ds are also reduced in lft mutants. However, because Fat signaling is influenced not only by the amount of Ds, but also by the pattern of Ds (i.e. is Ds expression graded, and how steeply), Ds can have positive or negative effects on Fat activity(Reddy and Irvine, 2008; Rogulja et al., 2008; Willecke et al., 2008). Thus,we suggest that the lft mutant phenotype might be relatively weak because decreased Fat and Ds levels, which would be expected to decrease Fat signaling, are partially offset by a flattening of the Fat and Ds expression gradients, which would be expected to increase Fat-Warts signaling(Reddy and Irvine, 2008; Rogulja et al., 2008; Willecke et al., 2008).

The observation that ds lft double mutants have more severe phenotypes than do ds or lft single mutants indicates that ds and lft can each independently influence Fat. lft and ds both influence Fat levels and localization, but even in the absence of these two genes, there was a visible difference in Fat protein staining between the wing pouch and the wing hinge. This implies that there are additional Fat regulators, and that the expression of these additional Fat regulators is differentially distributed between the wing pouch and the wing hinge. One additional Fat regulator that is differentially expressed between the pouch and the hinge is Fj(Villano and Katz, 1995),although as Fj is thought to act by influencing Fat-Ds interactions, it is not clear whether it could explain the differential Fat staining observed.

It appears that Lft is a major contributor to the normal levels of Fat. As Lft binds to the Fat cytoplasmic domain, it presumably influences Fat protein levels through this direct binding. Different molecular mechanisms for how Lft might influence Fat (and Ds) levels can be envisioned. One attractive possibility, given that Fat and Ds are transmembrane proteins, and that Lft could co-localize with them at the sub-apical membrane, is an effect on endocytosis, but it is also possible that Lft affects them in some other way.

Because Lft is closely related to LIX1 and LIX1L, and indeed LIX1L is functionally homologous to Lft, our studies of Lft identify regulation of mammalian Fat and Ds homologs as the likely cellular functions of LIX1 and LIX1L. Consistent with this inference, these proteins could bind to the cytoplasmic domain of human FAT4, and a BLASTP search with a short sequence motif of Fat common to Ds and FAT4 (WEYLLNWGPSYENLMGVFKDIAELPD, Fig. 1C) identifies these three proteins plus the mammalian Ds homologs DCHS1 and DCHS2 as the five closest matches in protein databases. This sequence motif also exhibits weak similarity to a region of E-cadherin that has been identified as contributing to binding to β-catenin (Clark et al.,1995; Huber and Weis,2001), but there is no obvious primary sequence similarity between Lft and β-catenin, and Lft did not detectably affect E-cadherin staining.

Functional studies of LIX1 and LIX1L in vertebrates have not yet been reported. However, feline LIX1 has been genetically linked to feline spinal muscular atrophy(Fyfe et al., 2006). Direct examination of human LIX1 in spinal muscular atrophy patients did not reveal any mutations (Fyfe et al.,2006; Parkinson et al.,2008). Nonetheless, the linkage of LIX1 and LIX1L to Fat signaling suggests that other members of the Fat signaling pathway should also be examined as potential candidate susceptibility loci for this debilitating disease. Murine Fat4 has been shown to be required for normal PCP in the ear and kidney(Saburi et al., 2008);however, it is also highly expressed in the nervous system, as are murine Lix1 and Dchs genes(Moeller et al., 2002; Rock et al., 2005), consistent with the expectation that these genes will interact in mammals, and might influence nervous system development.