Two Membrane-Associated Tyrosine Phosphatase Homologs Potentiate C. elegans AKT-1/PKB Signaling

Akt/protein kinase B (PKB) functions in conserved signaling cascades that regulate growth and metabolism. In humans, Akt/PKB is dysregulated in diabetes and cancer; in Caenorhabditis elegans, Akt/PKB functions in an insulin-like signaling pathway to regulate larval development. To identify molecules that modulate C. elegans Akt/PKB signaling, we performed a genetic screen for enhancers of the akt-1 mutant phenotype (eak). We report the analysis of three eak genes. eak-6 and eak-5/sdf-9 encode protein tyrosine phosphatase homologs; eak-4 encodes a novel protein with an N-myristoylation signal. All three genes are expressed primarily in the two endocrine XXX cells, and their predicted gene products localize to the plasma membrane. Genetic evidence indicates that these proteins function in parallel to AKT-1 to inhibit the FoxO transcription factor DAF-16. These results define two membrane-associated protein tyrosine phosphatase homologs that may potentiate C. elegans Akt/PKB signaling by cell autonomous and cell nonautonomous mechanisms. Similar molecules may modulate Akt/PKB signaling in human endocrine tissues.

In humans, dysregulation of insulin and IGF-1 signaling plays a prominent role in disease pathogenesis. Patients with type 2 diabetes mellitus exhibit resistance to insulin [10]; similar insulin resistance is observed in mice harboring mutations in the insulin receptor and downstream components of insulin signaling [11][12][13][14]. Downstream components of IGF-1 signaling have been implicated in cancer pathogenesis based on the identity of homologous transforming retroviral oncoproteins [15,16] as well as the existence of gene amplifications [17][18][19] and somatic mutations [20][21][22][23] in primary tumors and tumor cell lines.
daf-2/InsR mutants were first identified based on their increased tendency to enter an alternative larval developmental stage called the dauer stage (daf refers to a dauer formation phenotype) [45]. In replete growth conditions, C. elegans undergoes four larval molts prior to reaching reproductive adulthood [46]. Under conditions of high population density, high temperature, or starvation, early larvae bypass the normal second and third larval stages and instead develop into the alternative dauer larva. Dauers are morphologically distinct from normal L3 larvae, exhibiting radial and pharyngeal constriction, decreased pharyngeal pumping, and cuticular specializations called alae. In addition, they increase intestinal fat storage and exhibit extended longevity. Upon improvement of ambient conditions, dauers recover to the L4 larval stage and proceed to reproductive adulthood [47].
The tumor suppressor function of the 3-phosphoinositide phosphatase PTEN [20,21], the frequent somatic mutation of phosphatidylinositol (PI) 3-kinase in human cancers [23], and the discovery of a germline Akt2 loss-of-function mutation in a family with autosomal dominant insulin-resistant diabetes [58] underscore the central role of this pathway in human disease. In view of the striking structural and functional conservation of insulin-like signaling throughout metazoan phylogeny [59], it is likely that the identification of novel DAF-2/InsR signaling components will illuminate not only mechanisms of developmental regulation in C. elegans but also the pathogenesis of common human diseases such as cancer and diabetes.
To identify genes encoding such molecules, we performed a genetic screen for mutants that enhance the dauer-constitutive phenotype of an akt-1 null mutant (Eak screen). Seven genetic loci enhance akt-1 when mutated; we report the molecular identity of three eak loci here. These genes may encode elements of a membrane-associated complex that functions in two endocrine cells to potentiate insulin-like signaling. A similar activity may modulate insulin and IGF-1 signals in human endocrine tissues.

A Sensitized Genetic Screen for Akt/PKB Signaling Components
We hypothesized that a genetic screen performed in a weak dauer-constitutive daf-2/InsR pathway mutant background might allow the identification of DAF-2/InsR signaling components which, when mutated alone, might not have phenotypes. akt-1(mg306) was isolated in a genetic screen for mutations that affect the production of or response to elevated 3-phosphoinositide levels and contains a C!T transition that generates a nonsense mutation in the pleckstrin homology domain (see Materials and Methods). In contrast to daf-2/InsR, age-1/PI3K, and pdk-1 loss-of-function (lf) mutants, which form dauers at 25 8C or lower [28,30,60], akt-1(lf) alleles have a Hid (high-temperature-induced dauer) phenotype [61], developing reproductively at 25 8C but forming dauers at 27 8C (unpublished data and Figure 1). Consistent with a role for AKT-1 in DAF-2/InsR signal transduction [31,43], the 27 8C dauer-constitutive phenotype of akt-1 mutants is suppressed by a daf-16/FoxO (lf) mutation [61].

Synopsis
Insulin and insulin-like growth factor (IGF) signaling regulates critical physiological processes in a wide variety of multicellular organisms. In humans, dysregulation of IGF signaling underlies the pathogenesis of cancer and diabetes. In the nematode Caenorhabditis elegans, the DAF-2 insulin-like pathway regulates development, metabolism, and longevity. All known components of DAF-2 insulinlike signaling are structurally and functionally conserved in mammals, suggesting that insights gained from studying this pathway in C. elegans may shed light on pathogenetic mechanisms underlying cancer and diabetes. In this study, the authors describe a genetic screen designed to identify novel components of DAF-2 insulin-like signaling in C. elegans. They have characterized three genes that may encode parts of a novel multimolecular membraneassociated complex that potentiates DAF-2 insulin-like signaling in two neuroendocrine cells, the XXX cells. Two of these genes encode proteins similar to mammalian protein tyrosine phosphatases. These results suggest that protein tyrosine phosphatase-like molecules may transduce IGF signals in mammalian endocrine cells and highlight the role of endocrine circuits in the pathogenesis of cancer and diabetes.
background (unpublished data and Figure 1A). Mapping and complementation analysis indicate that these 21 mutants define seven Eak genes.
In an akt-1(þ) background, eak mutants had minimal dauerconstitutive phenotypes at 25 8C ( Figure 1A) and relatively weak dauer-constitutive phenotypes in comparison to akt-1(mg306) animals at 27 8C ( Figure 1B). The ability of eak mutants to enhance dauer arrest phenotypes of age-1/PI3K, akt-1, and pdk-1 loss-of-function mutants is consistent with their functioning in a parallel pathway, and their low penetrance dauer arrest phenotype at 25 8C would explain why these mutants were not isolated in previous screens for dauer arrest mutants performed at this temperature.
To address the issue of whether all three eak gene products function in the same pathway, we constructed eak-x;eak-y double mutants and tested them for dauer arrest at 25 8C and 27 8C. Lack of enhanced dauer arrest in double mutants would suggest that eak gene products function together in the same complex or pathway. eak-4;sdf-9 and eak-6;sdf-9 double mutants did not arrest as dauers at 25 8C ( Figure 1A) and did not exhibit enhanced dauer arrest at 27 8C ( Figure 1B), indicating that the EAK proteins likely function together in the same complex or pathway.
Most dauer-constitutive mutants in the daf-2/InsR pathway also have extended organismal longevity [6,29,30,66]. Therefore, we performed longevity assays on all eak mutants. Consistent with a recent report [32], the akt-1(mg306) mutation extended median life span by less than 10%, which is a smaller extension of life span than has been seen with other loss-of-function dauer-constitutive mutants in the daf-2/ InsR pathway. Interestingly, no eak mutants tested extended median or maximum life span significantly ( Figure S4). Furthermore, eak alleles did not enhance life span extension of akt-1(mg306), although they all strongly enhanced the dauer formation phenotype of akt-1(mg306) ( Figure 1A). Thus, eak signaling does not subserve longevity control.

Two eak Genes Are Related to Protein Tyrosine Phosphatases
Nine mutants define the eak-5 gene. Single nucleotide polymorphism (SNP) mapping [67] localized eak-5 to an approximately 480-kb genomic interval between cosmid F48F5 and the right telomere of Chromosome V. sdf-9, identified in a screen for enhancers of the dauer-constitutive phenotype of unc-31(e169) [62], lies within this interval. Sequencing of sdf-9 exons and splice junctions in all nine eak-5 alleles identified eight distinct point mutations (Table   1), and an akt-1 sdf-9 double mutant constructed using the sdf-9(ut187) allele identified in the unc-31(e169) enhancer screen [62] had a strong dauer-constitutive phenotype at 25 8C ( Figure 1A), indicating that eak-5 is allelic to sdf-9.
The eak-6 gene is defined by one allele, mg329. SNP mapping localized eak-6(mg329) to an approximately 210-kb genomic region on Chromosome I between cosmids F52F12 and B0379. BLASTP analysis of SDF-9 against the C. elegans Wormpep database (http://www.sanger.ac.uk/Projects/ C_elegans/WORMBASE/current/wormpep.shtml) identified a predicted homolog encoded by open reading frame F10G8. 4. This open reading frame lies within the eak-6 genomic interval defined by our SNP mapping. Sequencing of predicted exons and splice junctions of F10G8.4 in eak-6(mg329) identified a G!A transition resulting in an opal nonsense mutation near the predicted amino-terminus of the protein ( Figure 2A). Two of nine transgenic lines containing a genomic PCR fragment including the putative promoter, open reading frame, and 3' untranslated regions (UTR) of F10G8.4 exhibited rescue of the dauer arrest phenotype of an eak-6(mg329);akt-1(mg306) double mutant (unpublished data), supporting the argument that F10G8.4 is eak-6.
EAK-6 and SDF-9 both have amino acid similarity to PTPs ( Figure 2B and [62]). Conservation is strongest in the ten motifs that are conserved among 37 human PTPs [68]. Whereas SDF-9 does not retain the canonical catalytic cysteine residue found in all known PTPs and is therefore predicted to be catalytically inactive [62,69], the catalytic cysteine is conserved in EAK-6. Despite this difference, sdf-9 and eak-6 mutants have a similar phenotype ( Figure 1A and 1B). Among 19 amino acid residues that are invariant among 113 vertebrate PTPs [68], 15 are conserved in EAK-6 and 13 are conserved in SDF-9 ( Figure 2B). PTP activity assays on both epitope-tagged EAK-6 expressed in and immunoprecipitated from cultured human cells and a GST-EAK-6 fusion protein expressed in Escherichia coli revealed no hydrolytic activity on the substrate p-nitro-phenylphosphate (PNPP), a phosphotyrosine analog (unpublished data). Thus, EAK-6 and SDF-9 may be inactive phosphatase homologs that bind to tyrosine phosphoproteins. However, we cannot rule out the possibility that EAK-6 has PTP activity that is not detectable in the PNPP assay. Table 1. eak-5 Is Allelic to the Synthetic Dauer Formation Gene sdf-9 sdf-9 residues mutated in each eak-5 allele are numbered according to the annotated sequence of YAC Y44A6D (http://www.ncbi.nlm.nih.gov/entrez/viewer. fcgi?db¼nucleotide&val¼3217937). Predicted amino acid changes are numbered according to the predicted primary sequence of SDF-9 [62]. DOI: 10.1371/journal.pgen.0020099.t001

eak-4 Encodes a Novel Protein with an N-Myristoylation Signal
Three alleles, mg326, mg328, and mg348, define the eak-4 locus. SNP mapping narrowed the eak-4 interval to an approximately 560-kb genomic region on Chromosome IV between cosmids B0001 and T23B5. Cosmid rescue assays identified F53B2 as the rescuing cosmid, and rescue experiments with genomic PCR fragments corresponding to Figure 2. Predicted Primary Amino Acid Sequence of EAK-6 and Similarity with PTPs (A) EAK-6 amino acid sequence. EAK-6 sequence was derived from full-length cDNA amplified by RT-PCR from wild-type C. elegans total RNA. The PTP domain is denoted in boldface. The residue mutated in eak-6(mg329) is boxed and the predicted change indicated above the mutated residue. Amino acids encoded by an alternatively spliced exon present in EAK-6L but not in EAK-6S (see Materials and Methods) are underlined. (B) EAK-6 homology with PTPs. The PTP domain of EAK-6 is aligned with that of SDF-9 and 3 human PTPs. Sequences used in the alignment are based on domains defined by Pfam [101] and correspond to amino acids 40 to 276 of PTPN1 (PTP1B), 57 to 308 of EAK-6, 31 to 283 of SDF-9, 265 to 500 of PTPRA (receptor-type PTPa), and 273 to 520 of PTPN11 (SHP-2). Alignment was performed using ClustalX 1.8 and MacBoxshade 2.15. Conserved and identical residues are shaded. Ten motifs conserved among 37 vertebrate PTPs [68] are underlined, and 19 residues that are invariant among 113 vertebrate PTP domains are boxed. The four invariant residues that are not identical in EAK-6 (R45, Q85, H214, and G220, numbered according to the PTPN1 primary sequence) are denoted with dots. The catalytic cysteine residue [69] is denoted by an asterisk. Two conserved residues that are not conserved in EAK-6, D181 and Q262, are denoted by arrowheads. EAK-4 has amino acid similarity to four other C. elegans proteins, denoted by cosmid gene names F14H8.2, F14H8.4, F14H8.5, and T19C3.7 ( Figure 3). The function of these proteins is not known. A glycine residue that is conserved among EAK-4, F14H8.2, F14H8.4, and F14H8.5 is mutated in mg348, which is the strongest eak-4 allele based upon the 20 8C dauer-constitutive phenotype of eak-4;akt-1 double mutants (unpublished data). A heteroallelic strain with eak-4(mg348) in trans to a deficiency did not exhibit enhanced dauer arrest ( Figure S5), suggesting that eak-4(mg348) is a null allele. Notably, EAK-4, F14H8.2, F14H8.4, and F14H8.5 all have canonical N-myristoylation sequences [70], suggesting that they may associate with membranes.
eak Promoters Drive Transcription Specifically in the XXXL/R Cells A functional SDF-9::GFP fusion protein is expressed specifically in the XXXL/R cells [62]. Promoter fusions of eak-4 and eak-6 to GFP [71] were specifically expressed in two cells in the head that we identified as the XXXL/R cells based on their variable positions and neurite-like morphology [62]. Most eak-6p::GFP-containing animals also exhibited GFP expression in a third cell identified as the pharyngeal M1 motor neuron based on the position of its cell body and axonal process [72].
To confirm this result, we constructed an sdf-9 promoter fusion to red fluorescent protein [73], coinjected either eak-4p::GFP or eak-6p::GFP with sdf-9p::RFP, and assayed for colocalization of GFP and RFP in transgenic animals. The sdf-9p::RFP fusion was expressed exclusively in two head cells with position and morphology consistent with their identification as the XXX cells [62]. In both strains containing GFP and RFP reporter constructs, GFP and RFP colocalized ( Figure 4A), indicating that eak-4, sdf-9, and eak-6 are all expressed in XXXL/R.

EAK-4, SDF-9, and EAK-6::GFP Fusion Proteins Localize to the Plasma Membrane
In order to visualize the subcellular localization of EAK proteins, we made full-length translational GFP fusion constructs and expressed them in wild-type animals. Both EAK-4::GFP and EAK-6::GFP fusion proteins were expressed specifically in XXX, as was an SDF-9::GFP fusion protein [62]. They were also variably expressed in the intestine (unpublished data), a common site of artifactual GFP expression [74,75]. Coexpression of translational GFP fusions with an sdf-9p::RFP promoter fusion indicated that all three GFP fusions localize to the plasma membrane of the XXX cells ( Figure 4B). Membrane localization was also apparent in intestinal cells (unpublished data).
We determined the role of the N-myristoylation consensus motif in EAK-4 plasma membrane localization by constructing an EAK-4::GFP mutant in which the invariant glycine residue at position 2 is mutated to alanine (G2A). In contrast to wild-type EAK-4::GFP, which was localized to the plasma membrane ( Figure 4B and 4C), the EAK-4 G2A mutant GFP fusion protein exhibited diffuse cytoplasmic localization ( Figure 4C), indicating that an intact N-myristoylation motif is required for EAK-4 plasma membrane localization.
Expression of AKT-1::GFP in XXXL/R Rescues Dauer Arrest in an eak-4;akt-1 Double Mutant In stark contrast to the specific expression of EAK::GFP proteins in the XXX cells ( Figure 4A and 4B), a functional AKT-1::GFP fusion protein is expressed widely in postem- elegans Homologs Sequences represent the entire predicted amino acid sequences of all five genes. EAK-4 sequence was derived from full-length cDNA amplified by RT-PCR from wild-type C. elegans total RNA. Alignment was constructed using ClustalX 1.8 and MacBoxshade 2.15. Shaded residues indicate amino acid identity and conservation. Mutated residues in three alleles of eak-4 are boxed, and the predicted amino acid change is indicated above the box. The conserved G at residue 2 and S/T/A (boxed) at residue 6 in the N-myristoylation consensus sequence [70] are denoted by an asterisk and a vertical arrow, respectively. DOI: 10.1371/journal.pgen.0020099.g003 bryonic animals [31]. To determine the importance of AKT-1 function specifically in XXXL/R, we expressed the akt-1 open reading frame and 3' UTR under the control of the eak-4 promoter and asked whether this transgene could rescue dauer arrest in an eak-4(mg326);akt-1(mg306) double mutant. In three of three transgenic lines assayed, animals harboring the eak-4p::AKT-1 transgene bypassed dauer arrest and grew reproductively, whereas nontransgenic siblings formed dauers ( Figure 5). Control transgenic animals expressing an eak-4p::GFP transcriptional fusion construct did not bypass dauer (unpublished data). These results show that AKT-1 expression in the XXX cells is sufficient to rescue the akt-1(mg306) dauer arrest phenotype and indicate that the XXX cells are a major site of AKT-1 function in C. elegans.
A DAF-16A::GFP Fusion Protein Is Not Expressed in XXXL/R In order to assess whether eak gene products regulate DAF-16/FoxO cell autonomously, we first determined whether a functional DAF-16A::GFP fusion protein under the control of the daf-16a promoter is expressed in XXXL/R. We constructed a strain harboring integrated DAF-16A::GFP and sdf-9p::RFP transgenes and assayed for colocalization of GFP and RFP. Surprisingly, although DAF-16A::GFP is widely expressed [40], it did not colocalize with RFP ( Figure 6), indicating that it is not expressed at high levels in XXXL/R. Therefore, eak gene products may regulate DAF-16/FoxO nonautonomously.

Discussion
We have used an akt-1 enhancer screen to identify three genes encoding proteins that potentiate AKT-1 signaling. It is noteworthy that this screen has yielded neither weak alleles of known dauer-constitutive genes encoding components of DAF-2/InsR, DAF-7/TGF-b, or DAF-11/GC pathways nor strong alleles of the many mutants with weak dauerconstitutive phenotypes [61,63,76]. The fact that we have isolated multiple alleles of five of the seven genes identified in our screen ( Figure 3, Table 1, and unpublished data) without identifying alleles of most mutants with weak dauer phenotypes argues that the target for akt-1 enhancement is very specific and that most mutants with weak dauer phenotypes do not strongly enhance akt-1(mg306). Therefore, eak-4, sdf-9, and eak-6 likely do not enhance akt-1(mg306) simply by virtue of their weak dauer-constitutive phenotype; rather, these genes probably encode components of a complex or pathway that cooperates specifically with AKT-1.
sdf-9 was also identified in a genetic enhancer screen using the weak dauer arrest mutant unc-31(e169) as a genetic background [62]. unc-31 encodes a homolog of CAPS, a protein required for calcium-induced dense core vesicle exocytosis [77]. Since insulin is stored in and secreted from dense core vesicles in pancreatic islet beta cells [78], it is likely that UNC-31/CAPS and AKT-1 function in the same insulin signaling pathway. This is underscored by the suppression of the weak dauer arrest phenotypes of akt-1 and unc-31 mutants by daf-16/FoxO mutations ( Figure 1C and 1D [61]) and may explain why sdf-9 mutants were isolated in both screens. akt-1 sdf-9 and unc-31;sdf-9 double mutants have similar dauer arrest phenotypes at 25 8C ( Figure 1A and [62]). Surprisingly, we have not identified any akt-2 mutants in this screen. We expected akt-2 mutants to emerge from this screen, given that akt-1;akt-2 double mutants [79] as well as animals subjected to simultaneous RNAi of akt-1 and akt-2 [31,32] undergo dauer arrest. Furthermore, eak mutants do not enhance dauer arrest of akt-2 mutants ( Figure S2B), indicating that they may act in the same pathway. However, none of the 21 alleles isolated in the Eak screen is X-linked (the akt-2 gene lies on the X chromosome). In the course of constructing an akt-1(mg306);akt-2(ok393) double mutant, we noted that akt-1;akt-2 animals formed nonconditional dauers in the F 3 generation but were maternally rescued for dauer arrest in the F 2 generation (unpublished data), explaining why akt-2 alleles were not isolated in this F 2 screen. It is possible that other components of DAF-2/InsR signaling would emerge from an F 3 akt-1 enhancer screen.
In contrast to previous results demonstrating a strong dauer arrest phenotype for multiple sdf-9 alleles [62], we have only observed weak dauer phenotypes for all eak-4, sdf-9, and eak-6 alleles tested (Figure 1 and unpublished data). The alleles tested included the ut163 and ut187 alleles of sdf-9 previously described ( Figure 1A and 1B; [62]). As dauer assays at 27 8C are exquisitely sensitive to small changes in environmental conditions [63], we attribute these differences to small discrepancies between assay conditions in different laboratories. This may also explain the disparate effects of XXX laser ablation on dauer formation reported in the literature [62,65,80].

The XXX Cells as a Site of DAF-2/InsR Function
The XXX cells are annotated as hypodermal cells in the head [81] that abut the pseudocoelom and have neuronal characteristics [62,80,82]. Laser ablation of the XXX cells causes partial dauer arrest [62], indicating that the XXX cells normally function to inhibit dauer arrest. The specific expression of EAK-4, SDF-9, and EAK-6 in the XXX cells (Figure 4), taken together with our finding that AKT-1 expression in XXX suffices to rescue the dauer arrest phenotype of an eak-4;akt-1 double mutant ( Figure 5), indicates that the XXX cells may be a major site of DAF-2/ InsR function in C. elegans. However, the output of the XXX cell in the regulation of dauer arrest also depends on insulin signaling in other cells, since the dauer arrest phenotype caused by ablation of the XXX cells is suppressed by mutations in daf-16/FoxO but not by mutations in daf-3/SMAD4 [62], This is also consistent with our observation that DAF-16::GFP expressed from the daf-16a promoter is not expressed in XXX ( Figure 6). It is not known whether DAF-2/InsR is expressed in XXX.
EAK Proteins May Inhibit DAF-16/FoxO Nonautonomously eak-4, sdf-9, and eak-6 mutants strongly enhance the akt-1 null phenotype ( Figure 1A), suggesting that EAK proteins function in parallel to AKT-1. In addition, eak-4;sdf-9 and eak-6;sdf-9 double mutants do not exhibit more severe phenotypes compared to the respective single mutants ( Figure 1A and 1B), indicating that EAK-4, SDF-9, and EAK-6 function in the same pathway or complex. Parallel signaling of AKT-1 and EAK proteins could occur either at the organismal level, whereby AKT-1 signaling in non-XXX cells would converge with EAK signals in XXX to inhibit DAF-16/FoxO cell nonautonomously, or at the intracellular level, whereby AKT-1 signaling would converge with EAK signals in XXX to inhibit DAF-16/FoxO cell autonomously. The ability of AKT-1 expressed specifically in XXX to rescue an eak-4;akt-1 double mutant ( Figure 5) supports the notion that AKT-1 functions at least in part by signaling in parallel to EAK-4, SDF-9, and EAK-6 in XXX. However, the observation that dauer arrest caused by XXX laser ablation is suppressed by a daf-16 loss-of-function mutant [62] and the lack of DAF-16A::GFP expression in XXX ( Figure 6) suggest that EAK proteins regulate DAF-16A nonautonomously. It is possible that AKT-1 phosphorylates critical substrates in XXX distinct from DAF-16/FoxO. Alternatively, since there are at least three DAF-16/FoxO isoforms that may be transcribed from distinct promoters (WormBase Web site, http://www. wormbase.org, release WS150, November 30, 2005), a DAF-16/FoxO isoform distinct from DAF-16A may be a target of AKT-1 and EAK signals in XXX.

Mechanisms of SDF-9 and EAK-6 Membrane Localization
In contrast to EAK-4, which likely localizes to the plasma membrane through N-myristoylation ( Figure 4C), the mechanisms underlying the membrane localization of SDF-9 and EAK-6 are not clear. Neither SDF-9 nor EAK-6 possesses an N-myristoylation motif or a predicted transmembrane domain. SDF-9 and EAK-6 could associate with the plasma membrane by binding to membrane-associated proteins such as EAK-4. Alternatively, given that SDF-9 (and possibly EAK-6) has an inactive PTP domain that might bind to phosphotyrosine residues, they could bind to membraneassociated tyrosine phosphoproteins such as DAF-2/InsR. The ability of the DAF-18 ortholog PTEN to hydrolyze both phosphotyrosine [84] and phosphoinositides [85] suggests a third possible mechanism of SDF-9 and EAK-6 membrane localization via direct binding to phospholipids. We addressed each model experimentally. First, coprecipitation assays on N-terminal epitope-tagged EAK-4, SDF-9, and EAK-6 expressed in all pairwise combinations in cultured 293T cells failed to reveal direct physical interactions among the three proteins despite high levels of protein expression (unpublished data). Furthermore, epitope-tagged SDF-9 and EAK-6 did not coprecipitate tyrosine phosphoproteins after exposure of transfected cultured 293T cells to IGF-1 (unpublished data). Finally, radiolabeled SDF-9 and EAK-6 synthesized in vitro did not bind to phosphoinositides immobilized on nitrocellulose (Seth Field and Lewis Cantley, personal communication). The observation that SDF-9::GFP exhibits plasma membrane localization in the absence of intact DAF-2/InsR signaling ( Figure S6) is consistent with the lack of SDF-9 binding to tyrosine phosphoproteins in 293T cells and suggests that SDF-9 membrane association is independent of tyrosine phosphorylation. We cannot rule out the possibility that biologically relevant EAK proteinprotein or protein-lipid interactions require additional components that exist in vivo but were not present in the assays described. Notably, one (or more) of the four eak genes that remain to be cloned (eak-1, eak-2, eak-3, and eak-7) may encode such a molecule.

SDF-9 and EAK-6 Are Similar to Tyrosine Phosphatases
The similarity of SDF-9 and EAK-6 amino acid sequences to PTPs ( Figure 2B and [62]) suggests that SDF-9 and EAK-6 may comprise a PTP that activates downstream signaling through dephosphorylation of tyrosine residues. Although SDF-9 lacks the cysteine residue that is critical for PTP catalytic activity [69], EAK-6 retains this cysteine ( Figure 2B) and may possess phosphatase activity. Many receptor-type PTPs harbor tandem catalytic domains, one of which frequently does not have PTP activity [68]. Similarly, SDF-9 and EAK-6 might function as a heteromeric PTP associated with the plasma membrane. The finding that EAK-6 expressed in mammalian cells or bacteria lacks PTP activity on the substrate PNPP does not preclude the possibility that it may have activity on specific phosphotyrosine residues in the context of a fulllength protein.
Alternatively, it is possible that EAK-6 has phosphohydrolase activity on nonphosphotyrosine or nonprotein substrates. Among 19 invariant residues in the catalytic domains of 113 vertebrate PTPs [68], 15 are conserved in EAK-6 ( Figure 2B). Two of the four invariant residues not conserved in EAK-6, R45 and G220 (PTPN1/PTP1B numbering), are thought to be involved in phosphotyrosine binding [68]. In EAK-6 these are replaced with isoleucine and alanine, respectively. Changes in these residues could reflect changes in EAK-6 substrate specificity.
A third possibility is that EAK-6 does not possess catalytic activity. One of the four invariant residues not conserved in EAK-6, H214, is a glutamine residue in EAK-6 ( Figure 2B). This histidine is proposed to lower the pK a of the catalytic cysteine [68]; mutation of H214 to alanine reduces the catalytic activity of PTPN1/PTP1B approximately 80-fold [86]. Two other conserved residues, D181 and Q262, are replaced by glutamate and leucine, respectively, in EAK-6 Figure 7. Models of EAK-4, SDF-9, and EAK-6 Function (A) Cell autonomous signaling in XXX. A schematic of one XXX cell is shown. EAK proteins function in parallel with AKT-1 and in the same pathway as AKT-2 to promote nondauer development by potentiating DAF-9/CYP27A1 function either directly or indirectly. DAF-9/CYP27A1 synthesizes 3-keto-7(5a)-cholestenoic acid, a ligand that promotes reproductive development by inhibiting the nuclear hormone receptor DAF-12 [91]. DAF-16/FoxO is denoted with dashed lines, since DAF-16A does not appear to be expressed in XXX. It is not known whether other DAF-16/FoxO isoforms or DAF-2/InsR are expressed in XXX. (B) Nonautonomous signaling from XXX. A schematic of the anterior portion of an animal is shown with the head pointing left. The pharynx is also shown. XXX cells are denoted by red ovals, and DAF-16/FoxO-expressing cells are denoted by green ovals. EAK proteins in the XXX cells generate signals that regulate the synthesis or secretion of a hormone that inhibits DAF-16/FoxO function in other cells. DOI: 10.1371/journal.pgen.0020099.g007 ( Figure 2B); D181E and Q262A mutations reduce PTPN1/ PTP1B catalytic activity approximately 600-fold and approximately 80-fold, respectively [86]. Interestingly, the PTPN1/ PTP1B H214A mutation reduces K m approximately 5-fold, and the D181E and Q262A mutations reduce K m approximately 10-fold each [86]. Therefore, although substitutions in these conserved residues in EAK-6 may decrease hydrolytic activity, they also may increase binding affinity for substrate. These data are consistent with a model of EAK-6 and SDF-9 functioning as inactive PTP domains that bind to tyrosine phosphoproteins and regulate their interactions with other proteins in a manner similar to STYX family proteins [87]. In this scenario, EAK-6 and SDF-9 could also potentiate DAF-2/ InsR signaling by serving as adaptor proteins that increase the local concentration of associated proteins during cascade activation. Resolution of these uncertainties awaits a more detailed analysis of EAK-6 catalytic activity, binding to tyrosine phosphoproteins, and structure.

Bifurcation of DAF-2/InsR Signaling into Dauer and Longevity Outputs
Temporal and spatial specificity of DAF-2/InsR signaling may underlie differential effects on dauer formation and longevity. Inhibition of DAF-2/InsR signaling during early development enhances dauer formation but has no effect on organismal longevity, and inhibition of DAF-2/InsR signaling in early adulthood is sufficient to extend life span [88]. eak-4, sdf-9, and eak-6 promoter and translational fusions are all expressed continuously from late embryogenesis through early adulthood (unpublished data and [62]); thus, temporal regulation of EAK-4, SDF-9, and EAK-6 expression is not likely to explain the normal life span of eak-4, sdf-9, and eak-6 mutants. The specificity of eak-4, sdf-9, and eak-6 expression in XXX suggests that DAF-2/InsR signaling in the XXX cells may have specific dauer regulatory functions that have no impact on organismal longevity. This model is supported by a recent analysis of tissue-specific functions of DAF-16/FoxO indicating that longevity is primarily regulated by intestinal DAF-16/ FoxO, whereas dauer arrest is regulated by neuronal DAF-16/ FoxO [89]. Similarly, neuronal DAF-2/InsR regulates life span, whereas intestinal DAF-2/InsR controls metabolism [90]. In light of the observation that DAF-16A::GFP is not expressed in XXX (Figure 6), it will be of great interest to determine whether DAF-16A::GFP localizes to the nucleus in eak;akt-1 double mutants and, if so, whether there is tissue-specificity of nuclear localization.

A Model for EAK Protein Function
The data presented in this work are consistent with a model whereby EAK-4, SDF-9, and EAK-6 function in a single complex or pathway at the XXX plasma membrane in parallel with AKT-1 to inhibit dauer formation. Overexpression of the steroid hydroxylase DAF-9/CYP27A1, which is normally expressed in XXX, suppresses dauer arrest in sdf-9 mutants [62], suggesting that DAF-9 functions downstream of or parallel to SDF-9. In XXX, SDF-9 may function with DAF-9/ CYP27A1 to promote the synthesis and/or secretion of dafachronic acids, which are high-affinity ligands for the nuclear hormone receptor DAF-12 [62,91]. Therefore, high EAK protein activity may inhibit dauer formation by potentiating DAF-9/CYP27A1 activity, either directly or indirectly ( Figure 7A). EAK proteins may also have non-autonomous inhibitory effects on other cells that express DAF-16/FoxO ( Figure 7B). EAK regulation of DAF-16/FoxO probably does not occur through DAF-9/CYP27A1, since daf-9 is epistatic to daf-16 for dauer arrest [64,65].
This work identifies two membrane-associated PTP homologs in the endocrine XXX cells that modulate insulin-like signaling in C. elegans. In view of the striking structural and functional similarities in insulin/IGF signaling among metazoa, further studies of insulin/IGF signaling in C. elegans should contribute to our understanding and management of growth factor dysregulation in human disease.

Materials and Methods
eak mutant isolation, SNP mapping, and sequencing of mutant alleles. akt-1(mg306) animals were mutagenized with ethyl methanesulfonate using standard procedures [92]. Four P 0 animals were placed on each of 100 6-cm NGM agar plates [93] and removed after overnight egglay at 25 8C. F 2 dauers were picked to new plates for recovery after visualization under a Nikon SMZ800 dissecting microscope. Recovered dauers were retested for the dauer phenotype at 25 8C. Dauers that bred true were then outcrossed once with N2 (wild-type) animals and subjected to secondary assays. Mutants were tested for suppression by feeding RNAi of daf-3/SMAD4 and daf-16/ FoxO as described [56]. Those that were suppressed by daf-16/FoxO RNAi but not by daf-3/SMAD4 RNAi were outcrossed again, and F 2 dauers were picked for recovery, singled for egglay, and picked to worm lysis buffer containing 50 mM KCl, 10 mM Tris (pH 8.3), 2.5 mM MgCl 2 , 0.45% NP-40, 0.45% Tween 20, 0.01% gelatin, and 60 lg/ml proteinase K. After incubation at À70 8C for at least 10 min, 60 8C for 1 h, and 95 8C for 15 min, 2.5 ll of single worm lysate was PCRamplified in 100 mM Tris (pH 8.3), 500 mM KCl, 15 mM MgCl 2 , and 0.01% gelatin with Taq polymerase (Roche, Basel, Switzerland) using primers flanking the akt-1(mg306) point mutation. akt-1(mg306) animals harbor a C-to-T transition at nucleotide 31730 of cosmid C12D8 that creates a TaqI restriction fragment length polymorphism. Mutants for which most or all F 2 dauers were homozygous for akt-1(mg306) were considered true Eak mutants and were analyzed further. All mutants were outcrossed at least four times prior to detailed phenotyping.
SNP mapping was performed essentially as described [67]. Mutants were mated with the Hawaiian C. elegans isolate CB4856, and single F 1 cross-progeny were allowed to lay eggs overnight at 25 8C. Individual F 2 dauers were picked for recovery, subjected to egglay, and picked to 100 ll of worm lysis buffer. Lysates were incubated as described above. Then 2.5 ll of single worm lysate was PCR-amplified as described above using genomic primers (Invitrogen, Carlsbad, California, United States) spanning SNPs between CB4856 and the wild-type N2 strain. SNPs were detected by restriction enzyme digestion or sequencing of PCR products.
Dauer assays. Two or three gravid animals were picked to individual 6-cm NGM plates, allowed to lay eggs for several hours, and removed. Plates were shifted to the assay temperature, and dauers were scored 48 to 60 h thereafter. All assays of 27 8C dauer arrest were scored in blinded fashion.
RNAi. Feeding RNAi was performed as described [94], with minor modifications. Six-well plates containing NGM agar þ 5 mM IPTG were spotted with 400 ll of overnight cultures of E. coli HT115 harboring double-stranded RNAi expression plasmid L4440 [95] or L4440 containing daf-12/NHRor daf-16/FoxO-specific inserts. Overnight cultures were grown in 2XYT medium containing 50 lg/ml carbenicillin. After overnight incubation of RNAi plates at room temperature (to allow IPTG induction of double-stranded RNA synthesis), two L4 animals were picked to each well, grown at 15 8C, removed after egglay, and shifted to 27 8C. Dauers were scored approximately 60 h thereafter.
Life span assays. L4 animals were picked to seeded NGM agar plates containing 0.1 mg/ml 5-fluorodeoxyuridine (ten to 20 animals per plate), incubated at 25 8C, and scored every 1 to 2 d for vitality as described [90]. Animals that did not respond to prodding were scored as dead and removed.
cDNA isolation. C. elegans total RNA was isolated as described [96]. cDNA was amplified from total RNA using the SuperScript III RT-PCR Kit (Invitrogen). The 5' and 3' cDNA ends were identified using a 5'/3' RACE kit (Roche). PCR products were cloned into pCR4-TOPO (Invitrogen), and Qiaprep miniprep plasmid DNA (Qiagen) was sequenced using primers flanking the cloned insert.
RT-PCR of wild-type C. elegans total RNA revealed the presence of two alternatively spliced eak-6 mRNAs differing by the absence or presence of a single exon. The presence of the exon in the long cDNA isoform, eak-6L, results in a 17-amino-acid in-frame insertion Cterminal to the PTP domain of EAK-6 ( Figure 2A). cDNA analysis also revealed a GeneFinder misprediction in WormBase (WormBase web site, http://www.wormbase.org, release WS120, March 1, 2004) of the exon/intron boundary between exons 3 and 4 (unpublished data). The 5' and 3' RACE identified a 5' cDNA end 13 nucleotides upstream of the translation initiation codon and a 3' cDNA end 303 nucleotides downstream of the translation termination codon but did not identify a trans-spliced SL1 leader (unpublished data).
GFP and RFP reporter constructs. For promoter fusion constructs, promoter fragments were generated by amplifying 5' upstream sequences between the putative translational start codon and the nearest boundary of the gene immediately upstream. The following promoter fragments were amplified: eak-4p: nucleotides 13809 to 14696 of cosmid F53B2 [numbering corresponds to cosmid sequences obtained from the National Center for Biotechnology Information Web site (http://www.ncbi.nlm.nih.gov/)]; sdf-9p: nucleotides 10417 to 16170 of YAC Y44A6D; eak-6p: nucleotides 4381 to 5426 of cosmid F10G8. PCR primers were tailed with BglII linkers. Promoter fragments were purified using the Qiaquick PCR Purification Kit (Qiagen), digested with BglII, and repurified. eak-4 and eak-6 promoter fragments were subcloned into BamHI-digested GFP reporter plasmid pPD95.67 (a gift from Andrew Fire) to generate eak-4p::GFP and eak-6p::GFP, respectively; the sdf-9 promoter fragment was subcloned into BamHI-digested RFP [73] reporter plasmid pPD95.75_mRFP3 (a gift from Ho Yi Mak) to generate sdf-9p::RFP. Fragment orientation was confirmed by restriction digestion. Plasmids were purified using Qiagen columns.
Protein fusion constructs were made using overlap extension PCR [97,98]. PCR fragments encompassing the putative promoter and open reading frame up to but not including the translation termination codon were fused to a PCR fragment containing GFP and the unc-54 3' untranslated region (amplified from the GFP expression vector pPD95.75, a gift from Andrew Fire). The following gene-specific fragments were amplified for protein fusions: EAK-4: nucleotides 12780 to 14696 of cosmid F53B2; SDF-9: nucleotides 10417 to 18862 of YAC Y44A6D; EAK-6: nucleotides 4381 to 7152 of cosmid F10G8. To construct the EAK-4::GFP G2A N-myristoylation mutant, changes were made in the wild-type EAK-4::GFP primers to encode a glycine-to-alanine missense mutation at amino acid 2 of EAK-4. Fusion PCR products were purified using the Qiaquick PCR Purification Kit (Qiagen).