Gas1 Is Related to the Glial Cell-derived Neurotrophic Factor Family Receptors α and Regulates Ret Signaling*

The growth arrest-specific gene 1 (Gas1) protein has been proposed to function during development as an inhibitor of growth and a mediator of cell death and is also re-expressed in adult neurons during excitotoxic insult. Here we have demonstrated that the Gas1 protein shows high structural similarity to the glial cell-derived neurotrophic factor (GDNF) family receptors α, which mediate GDNF responses through the receptor tyrosine kinase Ret. We found that Gas1 binds Ret in a ligand-independent manner and sequesters Ret in lipid rafts. Signaling downstream of Ret is thus modified through a mechanism that involves the adaptor protein Shc as well as ERK, eventually blocking Akt activation. Consequently, when Gas1 is induced, Ret-mediated GDNF-dependent survival effects are compromised.

Gas1 is a glycosylphosphatidylinositol (GPI) 3 -anchored protein, expressed at growth arrest and down-regulated under proliferative conditions (1,2). During development, Gas1 is widely expressed in the nervous system (3) and is associated with growth inhibition and cell death (4). In Gas1 Ϫ/Ϫ mice the cerebellum is smaller than in wild type (5), indicating an unexpected role for Gas1 in proliferation. Moreover, Gas1 is induced by Wnt and interacts directly with sonic hedgehog (SHH), antagonizing SHH patterning function (6). In the adult, Gas1 is expressed during physiological apoptosis in some tissues (7), although not in brain (3). Aberrant Gas1 expression was nonetheless found in adult neurons during excitotoxicity, and its expression in neuroblastoma cells is associated with proapoptotic effects (8). Together, this suggests that Gas1 may have different functions within distinct cell contexts and that its mechanism of action depends on its spatiotemporal expression. No mechanisms have yet been proposed by which Gas1 transmits signals that affect decisions about cell proliferation, growth arrest/differentiation, or cell death.
The GFR␣ (GDNF family receptors ␣) are four GPI-anchored pro-teins that serve as a link between the neurotrophic factors GFL ( GDNF family ligand) and the transmembrane receptor tyrosine kinase Ret (reviewed in Ref. 9). The GFL consist of four proteins, GDNF, neurturin, artemin, and persephin, representing an important class of soluble mediators of neuronal survival, neurite growth, and differentiation. GFL are critical regulators of neurodevelopment and support the survival of midbrain dopaminergic and spinal motor neurons (reviewed in Ref. 10). These effects are mediated through the multicomponent receptor system consisting of GFR␣ and Ret (11,12). Activated Ret binds many different adaptor proteins to activate divergent downstream signaling pathways. Recruitment of Ret to lipid rafts seems to be important for the choice of adaptor molecule (13,14), although the mechanisms are not yet fully understood (reviewed in Ref. 15).
To study the Gas1 mechanism of action, we undertook detailed sequence analysis and found a structural relationship between Gas1 and the GFR␣. Based on this finding, we analyzed Gas1 involvement in GFR␣1-Ret receptor complex formation and downstream effects. Our results suggest that Gas1 is related to the GFR␣ and may have a regulatory function in Ret signaling. In a ligand-independent manner, Gas1 sequesters Ret to lipid rafts, mediates Shc and ERK recruitment to this membrane domain, and blocks Akt activation without affecting ERK activation. This indicates that the proapoptotic function of Gas1 could be associated to its interaction with Ret and to inhibition of a GDNFdependent survival pathway.

EXPERIMENTAL PROCEDURES
Sequence Analysis-For sequence analysis, we related distant protein families via intermediate searches (16) using global hidden Markov model profiles (17). To improve profile quality, we searched against EST (expressed sequence tag) data bases. Alignment was produced with T-Coffee software (18) and was slightly refined manually; it is viewed with the Belvu program (www.cgb.ki.se/cgb/groups/sonnhammer/ Belvu.html). Secondary structure was predicted using PhD (19). The location of predicted signal peptides and transmembrane regions is based on Signal-P and TMHMM2 programs, respectively (20,21).
Phylogenetic Analysis-To determine the phylogenetic distribution of the domains in homologous proteins, the domains were considered as individual entries. We used standard methods based on progressive alignment (22), generating Neighbor-Joining trees (23) to establish topologies in 10,000 bootstrap replicates. As the sequences were small and divergent, we used a probabilistic approach using MrBayes (24), running for 900,000 generations and four independent Markov chains. We sampled 25,000 trees and generated a consensus. Trees were drawn using the TreeView tool (25). For clarity, only the probabilistic unrooted tree is shown (Fig. 2).
Molecular Modeling-Fold recognition analyses were performed using the 3D-Jury metaserver (26). The model was based on the published crystal structure from domain 3 of the GFR␣1 receptor protein (27) and obtained using Swiss-Model (28). The model was evaluated using PSQS (www1.jcsg.org/psqs/). Model and template coordinates were superimposed using the DALI server (www.ebi.ac.uk/ dali/Interactive.html). Illustrations were generated with MOLMOL.
Plasmids-The rat Gas1 expression vector was previously described (8). Human GFR␣1 was obtained by PCR from a substantia nigra cDNA library (BD Biosciences) with the primers: 5Ј-TGAACCCCTAAA-AGCGGAACC and 5Ј-GCATATCCCAAAGCCTTCTGAGTT. Isoform GFR␣1.1 was subcloned in pcDNA3. The Ret sequence coding for amino acids 1-663 (extracellular domain, transmembrane domain, and 5 amino acids of the intracellular domain) was obtained by PCR using the primers 5Ј-ATGGCGAAGGCGACGT and 5Ј-AAACTTGTGG-TAGCAGTGGA and a human pcDNA3-Ret9 plasmid as the template. The amplified product was subcloned in the pCS2ϩMyc expression vector to obtain the plasmid pCS-RetMyc. The Gas1-TM expression construct, which codes for rat Gas1 without the putative amino acids susceptible to GPI modification, fused to the transmembrane domain (TM), and a truncated cytosolic tail (CT12) of human LDL receptor (29) were generated as follows. The TM and CT12 of human LDL-R were amplified by PCR using the primers 5Ј-GTTGGCGC GCCAGGAAG-TAGCGTGAGGGCTCTG-3Ј and 5Ј-CGCTCTAGATTATCAGTT-GATGCTGTTGATGTTC-3Ј with XhoI and XbaI restriction sites, respectively. The PCR product was cloned into pGEM-T Easy, sequenced, and subsequently ligated with rat Gas1, truncated at the C-terminal residue 357, in pcDNA3.
Cell Culture, Transfections, and Treatments-Human embryonic kidney (HEK)293T and mouse neuroblastoma Neuro 2a (N2a) cells were cultured in Dulbecco's modified Eagle's medium (DMEM), human neuroectodermic cells SK-N-MC in DMEM/F12 Glutamax, and human neuroblastoma SH-SY5Y in RPMI; all media were supplemented with 10% fetal bovine serum, Glutamax-I, and penicillin/streptomycin. To select for Ret9 stably transfected clones in SK-N-MC, complete medium was supplemented with 0.4 mg/ml of geneticin (G418; Invitrogen). G418-resistant clones were expanded and assayed for Ret expression. For experiments, clone 3 (SK-S3) was selected for moderate Ret expression. Cells were plated at 70% confluence in 60-mm plates. HEK293T and N2a cells were transfected with JetPei reagent (Bridge Bioscience, Palo Alto, CA) for 16 h according to the manufacturer's instructions, and the transfection efficiency was ϳ50% for both cell lines. In N2a cells, the GFR␣1: Gas1 plasmid ratio was 1:2. Before treatment, cells were incubated in serum-free medium for 6 h (HEK293T and SK-S3 cells) or 2-4 h (N2a cells). HEK293T cells were stimulated with 100 ng/ml of SHH (R&D Systems, Minneapolis, MN) or 50 ng/ml of GDNF (Peprotech, Rocky Hill, NJ) for 30 min. N2a cells were stimulated with 25 ng/ml of GDNF for 15 min, except when otherwise stated. SK-S3 and SH-SY5Y were stimulated with 100 ng/ml of GDNF for 15 min.
Chemical Cross-linking-Chemical cross-linking was performed essentially as described (30). HEK293T cells were transfected as above and serum deprived for 6 h. The 125 I-GDNF ligand (500 nM; Amersham Biosciences) was incubated with cells (30 min) and removed before cross-linking with ethyl-dimethylaminopropyl-carbodiimide supplemented with sulfo-NHS (Pierce) according to the manufacturer's instructions.
Cell Lysis, Immunoprecipitation, and Immunoblotting-Cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, protease inhibitor mixture (Roche Applied Science) and phosphatase inhibitors (Calbiochem Phosphatase Inhibitor Mixture Set II), and 1 mM NaO 4 Va) for 10 min on ice and for an additional 30 min rotating at 4°C. Lysates were cleared by centrifugation. Immunoprecipitation was done using antibodies and 10 l of protein G-Sepharose beads (GE Healthcare) 4°C, overnight. The immunoprecipitates were washed three times with lysis buffer and solubilized in sample buffer. Anti-Gas1 has been described (8). Antibodies against Ret, Src, GFR␣1, FRS2, ERK-2, thyroglobulin-stimulating hormone receptor (TSHR), and actin were from Santa Cruz Biotechnology, phosphotyrosine (4G10), Shc, and phospho-Shc were from Upstate Biotechnologies; Akt, phospho-Akt, and cAMP-response element-binding protein (CREB) were from Cell Signaling. TR was from Zymed Laboratories Inc., Myc, (9E10) was from BD Biosciences, and phospho-ERK was from Sigma. Blots were developed by chemiluminescence (Super Signal West Dura; Pierce or ECL Advance, GE Healthcare) and quantified using NHI Image software. Fold induction was calculated by ratio of phosphorylated form to total protein, and the ratio in basal condition was set to 1.0. Statistical analysis was performed using Student's t-test or one-way analysis of variance followed by Tukey's posthoc test for the significance of differences.
Lentiviral Infection-HEK293 cells were cotransfected with three packaging plasmids and the vector encoding rat antisense Gas1 (pRRLsin.PPT. hCMV.Gas1AS) or GFP (PRRLsin.PPT.hCMV.GFPWpre) as described (31). After 36 -60 h, viral particles were collected and concentrated by ultracentrifugation. Virus concentration was estimated by measuring the amount of p24 protein (PerkinElmer). SH-SY5Y was transduced 12 h after plating using 5 g of p24/10 6 cells. Medium was changed 8 h after virus addition, and cells were allowed to express the mRNA of interest for at least 48 h before experiments. Typical infection efficiency was ϳ90% as assessed using viral delivery of GFP.
Flotation Gradient and Detergent-resistant Membranes-For analysis of detergent-insoluble complexes in flotation gradients, 1.5 ϫ 10 6 cells were cooled on ice, washed with phosphate-buffered saline, and lysed in 350 l of TNE buffer (50 mM Tris⅐HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA) with 0.3% Triton X-100 (Calbiochem) as described (32). Cells were extracted (20 min, on ice) and the extract subsequently brought to 35% (v/v) Optiprep (Reactiva). The lysate (250 l) was overlayered sequentially with 3.5 ml of 30% (v/v) Optiprep and 200 l of TNE with detergent in SW60 tubes. After centrifugation (4 h, 170,000 ϫ g, 4°C), six 600-l fractions (1 to 6) were collected from the gradient (top to bottom) and precipitated with trichloroacetic acid (all steps performed in a 4°C cold room, on ice). Normalized protein amounts for each fraction were analyzed by SDS-PAGE and immunoblotting. As controls for the gradients, we used anti-Src (present in the raft-associated fraction one) and anti-transferrin receptor (present in the nonraftassociated fraction six). Detergent-resistant membrane fractions were isolated as described (13).
Flow Cytometry Analysis of Propidium Iodide Incorporation-SH-SY5Y cells were plated and infected after 12 h in culture. At 24 h postinfection, medium was replaced with serum-free medium; at 48 h, cells were treated with GDNF until 72 h postinfection. Cells were collected, prepared for cytometry using the Coulter DNA-Prep Reagents kit (Beckman-Coulter), and analyzed in a Coulter Epics XL-MCL Flow Cytometer (Beckman-Coulter). Ten thousand events were counted on the scatter gate. Expo32 was used as statistical analysis software. Experiments were repeated two times in quadruplicates and analyzed by oneway analysis of variance followed by Bonferroni test for the significance of differences.

GAS1 Is Related to the GDNF Family
Receptors ␣-Multiple alignment of the GAS1 protein primary sequence from several species identified two cysteine-rich repeats, residues 48 -147 and 166 -243 in the human sequence, in which the relative positions of the cysteines were highly conserved (Fig. 1). Searches with the global profile HMMer (hidden Markov model) of the cysteine-rich repeat region of GAS1 proteins found GFR␣ with an E-value of 0.049 (sptrembl_id:Q6UXV0, residues 131-210). Reciprocally, the profile of the GFR␣ repeat domain detected Gas1 proteins with an E-value of 0.55 (swissprot_id:GAS1_HUMAN, residues 166 -243). None of these profile HMMer searches retrieved any other related sequences. The human sequence sptrembl_id:Q6UXV0, GRAL (33), was automatically identified as a new GFR␣ member by Pfam (www.sanger.ac.uk/Software/Pfam/). In addition, secondary structure prediction for the Gas1 protein ( Fig. 1) showed good agreement with the crystal structure of GFR␣1 domain 3 (27).
We used sequence alignment of the repeated domain common to the GFR␣ and the GAS1 proteins for phylogenetic analysis. Taking into account both the short length and the high divergence of the alignment, low clade confidence values at deep branches of the tree are predicted. Nonetheless, the topology was consistent and sufficient to establish a relation among the GFR␣ domains ( Fig. 2A). In addition to the sequence similarity, we found an overall resemblance in the domain architecture of GFR␣ and GAS1 proteins (Fig. 2B). We confirmed a signal peptide located at the N terminus of both GAS1 and GFR␣1 and a GPI anchor at the C terminus of GFR␣ (11,12) and GAS1 proteins (2). Mammalian GRAL has a C-terminal transmembrane domain.
To determine whether fold recognition analysis generated consistent results, we submitted the two GAS1 cysteine-rich domains (swiss-  prot_id: GAS1_HUMAN, residues 48 -147 and 166 -243) as a query to a fold-assignment metaserver. The structure of the GFR␣1 protein cysteine-rich domain 3 (Protein Data Bank code 1Q8D) was detected as the optimal template. The disulfide bridges are conserved, and the overall fold is maintained (Fig. 3). Scores were low, as predicted, because of the short length of the domain and the low level of sequence similarity among the proteins. The high degree of sequence divergence can be explained by the fact that the overall fold of the template is sustained by disulfide bridges. This type of protein domain keeps the overall fold as a consequence of spatial restrictions imposed by its conserved disulfide bridges (34). Considering the E-values of the profile HMMer searches, the common domain architecture, the reliability of the secondary structure prediction, and the fold assignment results, we conclude that the cysteine-rich repeat domains in the Gas1 protein are similar to the GFR␣ family repeats. Together, the data suggest that Gas1 is related to the GFR␣.
Gas1 Interacts with Ret in a Ligand-independent Manner-Because the Gas1 repeats showed a fold similar to that of GFR␣1, we hypothesized that Gas1 interacts with Ret, the GFR␣1 binding partner. To study this possibility, we analyzed Gas1 binding to Ret in HEK293T cells, which are not growth arrested by Gas1 and do not express endogenous Gas1 or Ret proteins. We generated a RetMyc construct in which the C-terminal intracellular domain was replaced by six Myc epitopes. Coimmunoprecipitation experiments following Gas1 and RetMyc expression in HEK293T cells showed that RetMyc was precipitated by a Gas1 antibody and Gas1 was precipitated by an anti-c-Myc antibody (Fig. 4A). To assure that the Gas1-RetMyc interaction was not an artifact due to overexpression of a GPI-anchored protein, we generated a Gas1 construct, Gas1-TM, without the GPI anchor. This construct maintains the signal peptide and the two cysteine-rich domains, but the C-terminal amino acids, including the putative GPI modification sequence, were replaced by the transmembrane domain and 12 cytosolic amino acids of the human LDL receptor (29) (Fig. 4B, upper  panel). The Gas1-TM construct was expressed correctly and was recognized by the Gas1 antibody (not shown). HEK293T cells were transfected with RetMyc and Gas1-TM and analyzed by immunoprecipitation. The c-Myc antibody precipitated Gas1-TM efficiently (Fig. 4B, lower panel), confirming that GPI modification is not decisive for interaction with Ret.
Immunoprecipitations were done after cell culture in the presence of serum; we thus cannot rule out the possibility that the interaction is influenced by neurotrophic factors in the serum. To determine whether the interaction is ligand dependent, as is the case for Ret binding to GFR␣1, coimmunoprecipitations were repeated after serum deprivation, with or without known ligands. GFR␣1 coimmunoprecipitated with RetMyc only after addition of its preferred ligand, GDNF (Fig. 4C), as described (13). Conversely, Gas1 coimmunoprecipitated with Ret-Myc both alone and in the presence of its reported ligand, SHH (6), and with or without GDNF ( Fig. 4C and not shown), indicating that Gas1 interacts with RetMyc in a ligand-independent manner.
Gas1 is widely expressed in the nervous system during mouse embryonic development (4) at times when Ret expression has also been reported, e.g. in brain at embryonic day 14 (E14) (35). To demonstrate that the Gas1-Ret interaction can occur in vivo, we used brain extract from E14 mice for immunoprecipitation with an anti-Ret antibody. Immunoprecipitated Gas1 was detected by Western blot (Fig. 4D, upper  panel). To confirm this interaction, Gas1 was immunoprecipitated from E14 mouse brain extract and Ret was detected by Western blot (Fig. 4D,  lower panel). These data indicate that Gas1 could interact with Ret in developing mouse brain.
Gas1 Does Not Bind to GDNF-As the Gas1-Ret interaction is ligand independent, we tested whether Gas1 interferes with the phosphorylation-competent complex formed by GDNF, GFR␣1, and Ret in crosslinking experiments with radiolabeled GDNF in transfected HEK293T cells. GFR␣1 bound 125 I-GDNF, forming a complex of ϳ85 kDa (Fig. 5, lane 3) as described (11,12), whereas Gas1 did not bind to 125 I-GDNF (lane 2). In cells transfected with RetMyc alone or together with Gas1, we observed no binding to 125 I-GDNF (lanes 4, 5). Cotransfection of GFR␣1 and RetMyc resulted in the formation of two complexes, at ϳ85 and 200 kDa, corresponding to the 125 I-GDNF-GFR␣1 and 125 I-GDNF-GFR␣1-RetMyc complexes, respectively (lane 6). Cotransfection of Gas1 with GFR␣1 and RetMyc showed the same banding pattern as for GFR␣1 and RetMyc alone, with no change in the level of the complex at 200 kDa (lane 7). The results indicate that neither Gas1 alone nor the Gas1-RetMyc complex is able to bind 125 I-GDNF and that Gas1 does not interfere with GDNF-GFR␣1 binding to Ret.
Gas1 Modifies Ret Downstream Signaling-Because Gas1 did not affect GFR␣1-Ret receptor complex formation, we analyzed whether Gas1 influenced Ret autophosphorylation and activation. We used the N2a neuroblastoma cell line, which expresses Ret and very low GFR␣1 levels (11). Gas1 expression in serum-deprived N2a cells did not result in Ret tyrosine phosphorylation, although Gas1 coimmunoprecipitated with Ret (Fig. 6A); cells transfected with GFR␣1 and stimulated with GDNF showed robust Ret phosphorylation (Fig. 6A) as described (11). To determine whether Gas1 is involved in the regulation of Ret downstream signaling, we transfected N2a cells with GFR␣1 with or without Gas1. We analyzed the phosphorylation of Akt and ERK, two major effectors downstream of Ret, after serum deprivation and GDNF stimulation. To permit comparison after GDNF stimulation, we verified similar GFR␣1 expression in all samples in each experiment. Coexpression with Gas1 did not modify GFR␣1-GDNF-induced Ret autophosphorylation (Fig. 6B). The presence of Gas1 severely decreased GDNFinduced Akt phosphorylation from 2.9-to 1.4-fold compared with control cells ( p Ͻ0.01, n ϭ 4) (Fig. 6B), whereas ERK phosphorylation increased from 2.0-to 4.5-fold ( p Ͻ0.01, n ϭ 4). Time course analysis of Ret, Akt, and ERK phosphorylation showed that in the presence of Gas1, Akt phosphorylation was impaired at all times analyzed up to 2 h, whereas ERK phosphorylation took place earlier and lasted longer (not shown).
We analyzed the influence of endogenously expressed Gas1 on Ret signaling using the SH-SY5Y human neuroblastoma cell line, which expresses Ret and GFR␣1 and responds to GDNF (36). Gas1 was expressed at very low levels in the presence of serum, whereas Gas1 mRNA was strongly induced after 16 h of serum deprivation (Fig. 6C). This correlated with increased Gas1 protein levels detected by immunoprecipitation (Fig. 6C). We used this cell model to analyze Ret signaling with or without Gas1. SH-SY5Y cells were serum deprived for 3 h (absence of Gas1) or for 24 h (presence of Gas1). Cells were subsequently stimulated with GDNF, and phosphorylation of Ret, Akt, and ERK was analyzed. Concurring with the results in transfected N2a cells (Fig. 6, A and B), in Gas1-expressing cells Ret phosphorylation was independent of Gas1 expression (Fig. 6D), whereas Akt phosphorylation decreased appreciably from 2.0-to 0.9-fold ( p Ͻ0.05, n ϭ 3) and ERK phosphorylation increased from 1.9-to 3.2-fold ( p Ͻ0.01, n ϭ 3) (Fig.  6D). As serum deprivation can affect other pathways, however, these  data do not allow us to conclude that Gas1 is uniquely responsible for inhibition of Akt activation and increased ERK activation after GDNF stimulation.
To confirm the influence of Gas1 on Ret signaling, we analyzed GDNF-induced Ret, Akt, and ERK phosphorylation in SH-SY5Y cells in which Gas1 induction after serum deprivation was knocked down by lentivirus-mediated expression of an antisense Gas1 mRNA. Ret phosphorylation was similar in GFP-infected cells (Gas1-expressing cells) and in cells infected with antisense Gas1 (Gas1-knockdown cells) (Fig.  6E). Gas1-expressing cells (GFP-infected) did not activate Akt after GDNF stimulation (1.3-fold), whereas cells infected with antisense Gas1 (knockdown) were able to do so (2.2-fold, p Ͻ0.001, n ϭ 4) (Fig. 6E). ERK phosphorylation was similar both in Gas1-expressing cells and in Gas1 knockdown cells, 3.2-and 3.3-fold, respectively (Fig. 6E). These results indicate that endogenously induced Gas1 cannot activate Ret but modifies Ret downstream signaling induced by GDNF-GFR␣1 so that Akt phosphorylation is reduced. The relevance of Gas1 in ERK phosphorylation is nonetheless unclear. Overexpressed Gas1 mediates an increase in ERK phosphorylation in response to GDNF. ERK is also strongly activated after 24 h of serum deprivation and GDNF stimula-tion as compared with cells serum deprived for 3 h. This increase was not reversed by antisense Gas1 mRNA, however, suggesting a minor role for Gas1 in ERK activation. Alternatively, serum deprivation may generate sufficient oxidative stress to strongly activate ERK after GDNF stimulation, also in the Gas1 knockdown cells. Taken together, our results show that Gas1 is a regulator of Ret downstream signaling, specifically blocking Akt activation.
Gas1 Recruits Ret to Lipid Rafts-The location of phosphorylated Ret in the cell membrane is an important determinant for the activation of distinct downstream signaling pathways (14). GFR␣1 is constitutively located inside lipid rafts through its GPI anchor, whereas in unstimulated cells Ret is found mainly outside rafts. After GDNF stimulation, GFR␣1-GDNF complex formation induces transient Ret recruitment to lipid rafts (13) and activated Ret is in equilibrium between detergentresistant membranes and the soluble membrane fraction (14). To determine Gas1 location in the plasma membrane, we used N2a cells and SK-S3, a clone derived from the human neuroectodermic SK-N-MC cell line. SK-N-MC cells express endogenous GFR␣1 and GFR␣2 (37), and the SK-S3 clone stably expresses Ret9, the short isoform of Ret (38). We transiently transfected N2a and SK-S3 cells with Gas1 and sepa- rated cell extracts in flotation gradients. Gas1 co-fractionated with a raft-associated marker in the first fraction (Fig. 7A). We used SK-S3 cells to establish whether the effect of Gas1 on Ret signaling is due to modification of Ret location in the membrane. In cells transfected with vector alone, a small fraction (8% of total protein) of Ret was found in lipid rafts (Fig. 7B, upper left, fraction 1), which increased (13%) in the presence of GDNF (Fig. 7B, lower left) as described (13,14). In cells transfected with Gas1, we observed substantial Ret recruitment to lipid rafts ( fraction 1) in the absence of GDNF (16%) (Fig. 7B, upper right), which increased further after GDNF addition (21%) (Fig. 7B, lower  right). This indicates that prior to cell stimulation with GDNF, Gas1 increases the Ret fraction in lipid rafts, which may modify signaling events downstream of Ret.
To analyze how Gas1 affects the membrane distribution of phosphorylated Ret, we isolated detergent-resistant membranes and soluble fractions from transfected N2a cells. When Gas1 was cotransfected with GFR␣1, the level of phosphorylated Ret increased in the detergentresistant fraction both in unstimulated and GDNF-stimulated cells (Fig.  7C, upper). The reverse was the case in the soluble fraction, and Ret phosphorylation was higher in the absence of Gas1 (Fig. 7C, lower). This confirms the ability of Gas1 to recruit Ret to lipid rafts and shows that activated Ret is increased in this membrane microdomain in the presence of Gas1.
Gas1 Mediates the Recruitment of Activated Shc and ERK2 to Lipid Rafts-Ret can activate the PI3K/Akt pathway through various adaptors (15). To explain Gas1-mediated impairment of Akt activation, we analyzed activation of the adaptor FRS2, which transduces Ret signaling in rafts and is proposed to activate the PI3K/Akt pathway (39). We also analyzed the Shc adaptor, which transduces Ret signaling outside rafts (14) and activates the PI3K/Akt pathway (40). We immunoprecipitated FRS2 from N2a cell extracts and analyzed tyrosine phosphorylation by immunoblotting. GDNF stimulation induced a strong tyrosine-phosphorylated band at 75 kDa (3.5-fold), corresponding to activated FRS2. In the presence of Gas1 and GDNF, the level of activated FRS2 was higher (5.2-fold, p Ͻ0.05, n ϭ 3) (Fig. 8A, left), in agreement with the Ret increase in rafts. Gas1 thus increases GDNF-induced FRS2 activation, which concurs with the fact that Gas1 is changing the raft/non-raft equilibrium of Ret. It cannot, however, explain the reduction in Akt phosphorylation.
In cells transfected with GFR␣1 and stimulated with GDNF, Shc phosphorylation increased 1.8-fold (Fig. 8A, right) as described (40). Strikingly, in cells transfected with GFR␣1 and Gas1, Shc was strongly activated, 3.5-fold ( p Ͻ0.01, n ϭ 3) in the absence of GDNF. Addition of GDNF resulted in rapid Shc dephosphorylation to basal levels (1.2-fold) (Fig. 8A, right). The strong, Gas1-mediated Shc activation in the absence of GDNF did not result in activation of Akt or ERK, as shown in Fig. 6B. GDNF addition thus reversed Shc phosphorylation, suggesting that Shc is involved in the mechanism that blocks Akt activation in the presence of Gas1.
Gas1 recruits non-phosphorylated Ret to lipid rafts (see Fig. 7), which may contribute to the modified function of adaptor proteins. We thus analyzed recruitment to lipid rafts of Shc, as well as of ERK, a major downstream effector that is implicated in negative loops that regulate adaptor signaling (41,42). In the absence of Gas1 and prior to GDNF stimulation, we detected no Shc or phospho-ERK recruitment to lipid rafts (Fig. 8B, upper left). In unstimulated cells transfected with Gas1, we observed strong Shc recruitment and a single band of activated ERK in lipid rafts (Fig. 8B, upper right). An ERK2-specific antibody identified the single phospho-ERK band as ERK2 (Fig. 8B). After GDNF stimulation in the presence of Gas1, Shc and phosphorylated ERK2 recruitment to lipid rafts was not maintained (Fig. 8B, lower right). In summary, Gas1 not only recruits Ret to lipid rafts in a ligand-independent manner but also recruits Shc and phosphorylated ERK2. This modifies the protein environment of Ret, resulting in a Gas1-specific pathway downstream of Ret that blocks Akt activation.
Gas1 Blocks GDNF-induced Survival-Because Gas1 blocks GDNFinduced Akt phosphorylation, we studied the effect of endogenous Gas1 expression on GDNF-mediated cell survival after serum deprivation. We quantified cell viability after knock down of endogenous Gas1 expression induced by serum deprivation, using a lentivirus expressing antisense Gas1 mRNA (LV-ASgas1). We used a lentiviral vector coding for GFP (LV-GFP) as a negative control and to monitor equivalence of infectivity between experiments as determined by fluorescence-activated cell sorter. Infection with lentiviral particles had no effect on SH-SY5Y neuroblastoma cell viability compared with uninfected cells ( Fig. 9 and not shown). In LV-GFP-infected cells, serum deprivationinduced cell death was only slightly modified by GDNF treatment (100 ng/ml) (Fig. 9). Knock down of endogenously induced Gas1 in LV-AS-gas1-infected cells improved cell survival only slightly. Treatment with the same GDNF concentration in Gas1 knockdown cells, however, significantly rescued neuroblastoma cells, reducing the percentage of cell death by half (Fig. 9). These data suggest that endogenous Gas1 expression counteracts the GDNF-activated survival pathway by providing a scenario in which Gas1 and Ret interaction has clear physiological consequences.

DISCUSSION
Sequence alignments, secondary structure predictions, and threading analyses yielded a template suitable for building a feasible model of GAS1. The GFR␣1 structure is maintained by cysteine bridges (27), and although the conservation in the alignment was low, Gas1 proteins conserve these critical residues. The E-values obtained in HMMer searches showed a high degree of reliability, and based on the biological evidence presented here, we conclude that Gas1 is related to the GDNF family receptors ␣. The phylogenetic analysis did not provide high confidence branches; this is expected due to the short and highly divergent nature of the sequences, but tree topology was consistent when analyses were independent. This suggests that Gas1 was separated from GFR␣ very early in metazoan evolution and developed behavior different from that of GFR␣. The ability of Gas1 to interact with Ret in the absence of ligand reflects this early divergence, suggesting that Gas1 could have functions or partners different from those described for the GFR␣.
Our data suggest that Gas1 is a regulator of Ret signaling. The temporal expression patterns of these two proteins at defined developmental stages suggests a physiological function for the Gas1-Ret interaction. Gas1 and Ret are both expressed in midbrain, hindbrain, and spinal cord at E10 -10.5 (4,35). At E14 -14.5, Gas1 and Ret are present in hindbrain, spinal cord, dorsal root ganglia, lung bronchioles, and gonads, although coexpression has not been analyzed in the different cell types in these tissues (4,35). Another putative site for a Gas1-Ret interaction is the mouse embryonic retina. Ret is expressed strongly at E13.5 in neural retina and in the retinal pigmented epithelium, a layer in which Gas1 transcripts have been detected from E12.5 to E14.5 (43). Gas1-deficient mice show transdifferentiation of After immunoprecipitation with an anti-FRS2 antibody, FRS2 phosphorylation (left panel) was analyzed by immunoblotting with a phosphotyrosine-specific antibody. Phosphorylation of Shc (right panel) was analyzed by immunoblotting using a specific anti-phospho-Shc antibody. B, flotation gradients prepared from transfected N2a cells, serum deprived and GDNF stimulated, were analyzed by immunoblotting. In cells transfected with GFR␣1 alone, recruitment of nonphosphorylated ERK2 was detected neither before nor after GNDF stimulation (upper and lower left panels). Cells transfected with Gas1 and GFR␣1 showed Shc and activated ERK2 recruitment to lipid rafts prior to GDNF stimulation (upper right). After GDNF stimulation, the pattern of raft recruitment was similar to that in GFR␣1-only-transfected cells (lower right). the retinal pigmented epithelium to neural retina, (43), which may indicate a Gas1 association to differentiation. Taken together, the expression data and the present results suggest that Gas1 may have an influence on Ret signaling during development. Through the tyrosine kinase Ret, the GDNF family ligands are able to activate many different signaling pathways (15). Akt and ERK are two main effectors of Ret signaling, and both proteins are usually activated together. Because Akt is associated with survival and proliferation (37), whereas ERK is implicated in differentiation and neurite outgrowth (37,44), there must be a coordinating mechanism that permits them to mediate the different biological activities of Ret. Ectopic Gas1 expression in N2a cells suggests that Gas1 negatively regulates GDNF-induced Akt signaling whereas ERK activation is increased. Transient knock down of endogenous Gas1 in SH-SY5Y cells confirmed the ability of Gas1 to block GDNF-induced Akt phosphorylation. The role of endogenous Gas1 as a positive regulator of ERK activation has not been completely resolved, however, because induced levels of activated ERK, which are resistant to Gas1 knock down, could be due to activation of parallel signaling pathways during serum deprivation.
In the absence of Gas1, Ret is located mainly outside rafts; following GDNF stimulation, Ret is recruited to the rafts and is tyrosine phosphorylated. Activated Ret is in equilibrium between rafts and soluble membrane, where Ret associates with the adaptors FRS2 and Shc, respectively (14). Ret downstream signaling from outside rafts seems to be less efficient, according to results using the GFR␣1 (13) and Gas1 TM constructs (not shown). FRS2 and Shc both activate Akt and MAPK, eventually resulting in survival/differentiation. These two pathways are activated by the formation of different, independent complexes. One complex involves Grb2-Gab1/2 and leads to recruitment of p85, which activates the PI3K/Akt pathway; the other, mediated by Grb2-Sos, leads to Ras/MAPK activation (39,40).
Gas1 is located in lipid rafts and recruits Ret to these microdomains, resulting in a higher level of activated FRS2. A consequent increase in the activation of Akt and ERK would be predicted. Remarkably, the Akt pathway was inhibited whereas the ERK pathway was not. To explain these modifications, it is important to consider the Gas1-induced changes that occur before GDNF stimulation; these are Ret recruitment to rafts, which promotes Ret interaction with FRS2, and the recruitment of activated Shc and ERK2. The biological significance of these latter recruitments is presently unknown, but they can be mechanistically linked to the negative regulation of the PI3K/Akt pathway.
Signals through Ret activate PI3K/Akt through many different adaptors, which must be blocked by Gas1 to inhibit the PI3K/Akt pathway. One candidate mediator of this inhibition is the scaffold adaptor protein Gab1/2, which links the adaptors and the PI3K regulatory subunit p85 (15). Gab1/2 can be inhibited or activated by ERK via phosphorylation at different threonine residues. Phosphorylation of Gab1 at Thr-477 enhances recruitment, whereas phosphorylation at a still unidentified threonine residue correlates with inhibition of PI3K activation (42). We observed ERK2 inhibition of Gab1/2 in the presence of Gas1 before stimulation; this could agree with existing data (39,40), as the block in PI3K activation via Grb2-Gab1/2-p85 does not impair Grb2-Sos-Ras formation and activation of MAPK. In addition to Gab1/2, another protein(s) must participate in mediating the Gas1 block of Akt activation. Among these, we might hypothesize a protein phosphatase that rapidly inactivates Shc after GDNF stimulation in the presence of Gas1. The nature of this phosphatase, as well as of other mediators of Gas1-induced Akt inactivation, is presently unknown.
Gas1 is proapoptotic during development (4) and participates in excitotoxic neuronal death in adult brain (8). Here we have shown that GDNF-induced Akt phosphorylation is blocked by Gas1. GDNF-induced neuron survival and proliferation is dependent on PI3K/Akt activation and not on the Ras/ERK pathway (40,37). Suppression of the Akt phosphorylation blockade by antisense Gas1 knock down concurs with the improved survival effect of GDNF in serum-deprived SH-SY5Y cells. Our data show that Gas1 also modifies GDNF-induced ERK activation. ERK phosphorylation is associated to differentiation (37,44), and it is tempting to speculate that Gas1 could function as a switch between proliferation and differentiation during neuron development. In the adult brain, ERK activation has also been associated with a proapoptotic function, because in vivo inhibition of ERK activation after ischemia promotes cell survival (45). Emerging data relating ERK activation with neuron degeneration and apoptosis were recently reviewed (46). A putative dual mechanism for the proapoptotic function of Gas1 in adult brain could thus be linked to Akt blockage as well as to ERK activation.
Gas1 regulates Ret signaling in a ligand-independent manner, and our data show that SHH does not affect the Ret-Gas1 interaction. Recently, however, SHH was described as repressing GDNF-induced differentiation and migration of neural crest cells (47). Gas1 interacts with SHH (6) and with Ret; for this reason, an interaction involving SHH and Ret cannot be excluded. Gas1 blocks the cell cycle in fibroblasts (1), a cell type that does not express Ret, indicating that other transmembrane receptor(s) could also mediate the biological effects of Gas1 in different cell systems. GFR␣1 is known to interact with p140 N-CAM (30), for instance, suggesting that cell adhesion molecules (CAM) could be potential partners for Gas1 biological effects. Nevertheless, because of the early divergence between GFR␣ and Gas1 during evolution, this putative interaction between cell adhesion molecules and Gas1 must be demonstrated experimentally. The interaction between Gas1 and Ret provides a new framework to understand the functions described for Gas1 during development and neurodegeneration.