Tyrosine Phosphorylation of Tub and Its Association with Src Homology 2 Domain-containing Proteins Implicate Tub in Intracellular Signaling by Insulin*

A mutation in the tub gene leads to maturity-onset obesity, insulin resistance, and progressive retinal and cochlear degeneration in mice. tub is a member of a growing family of genes that encode proteins of unknown function that are remarkably conserved across species. The absence of obvious transmembrane domain(s) or signal sequence peptide motif(s) suggests that Tub is an intracellular protein. Additional sequence analysis revealed the presence of putative tyrosine phosphorylation motifs and Src homology 2 (SH2)-binding sites. Here we demonstrate that in CHO-IR cells, transfected Tub is phosphorylated on tyrosine in response to insulin and insulin-like growth factor-1 and that in PC12 cells, insulin but not EGF induced tyrosine phosphorylation of endogenous Tub.In vitro, Tub is phosphorylated by purified insulin receptor kinase as well as by Abl and JAK 2 but not by epidermal growth factor receptor and Src kinases. Furthermore, upon tyrosine phosphorylation, Tub associated selectively with the SH2 domains of Abl, Lck, and the C-terminal SH2 domain of phospholipase Cγ and insulin enhanced the association of Tub with endogenous phospholipase Cγ in CHO-IR cells. These data suggest that Tub may function as an adaptor protein linking the insulin receptor, and possibly other protein-tyrosine kinases, to SH2-containing proteins.

Tubby, an autosomal recessive syndrome in mice, is characterized by maturity-onset obesity, insulin resistance, infertility, and progressive cochlear and retinal degeneration (1)(2)(3). The tubby phenotype resembles human sensory/obesity syndromes such as Alstrom's and Bardet/Biedl that are characterized by combined cochlear hair cell and retinal photoreceptor loss accompanied by infertility and moderate late-onset obesity (4 -6). To date no candidate human genes for these syndromes have been identified. Positional cloning of the tub locus in mice lead to the identification of Tub, which encodes a novel protein of unknown function (7,8). Tub and Tub-like proteins have been found in rice, maize, Arabidopsis, Caenorhabditis elegans, Drosophila, and mammals (7,8). Although Tub is expressed in the key hypothalamic nuclei that have been implicated in the central control of energy homeostasis, it is not known whether Tub plays a regulatory role in the central mechanisms that control body weight. Alternatively, the weight gain observed in tubby mice may be due to a loss of cells in the hypothalamic nuclei that regulate body weight homeostasis accompanied by gradual loss of function as it occurs in the retina and cochlea of these animals.
The most striking feature of the Tub family of proteins is a highly conserved C terminus. Specifically, the last 250 amino acids are 55-95% identical across species, suggesting that this domain may perform a very ancient and basic function. In contrast, the N-terminal portion of Tub and Tub-like proteins are poorly conserved. The mutation in tubby mice is a G-to-T transversion in the 3Ј coding region of the tub gene that abolishes a donor splice site and results in the generation of a larger transcript containing the unspliced intron (7,8). Interestingly, this mutation disrupts the highly conserved C terminus and is predicted to generate a Tub mutant protein in which the last 44 amino acids are replaced with 20 amino acids not found in the wild type protein.
To gain insight into the role of Tub we searched for motifs that might be relevant for function (see Fig. 1). Tub encodes a highly hydrophilic protein containing putative tyrosine phosphorylation and SH2 1 -docking sites. No membrane-spanning domain(s) or signal sequence peptide motif(s) were identified. Phosphotyrosine-mediated interactions with SH2-containing signaling proteins have been shown to play an important role in intracellular signal transduction (reviewed in Refs. 9 -11).
Insulin action in the central nervous system has been implicated in the regulation of energy intake and expenditure (reviewed in Ref. 12). Administration of insulin directly into the central nervous system leads to a reduction in food intake and body weight in rats (reviewed in Ref. 13). The insulin receptor, a member of a large family of transmembrane protein-tyrosine kinases (14), is widely distributed in the central nervous system (15), and its hypothalamic expression matches that of the leptin receptor (16,17) and Tub (7). Binding of insulin to its cell surface receptor initiates a cascade of events that lead to the phosphorylation of downstream targets, including the insulin receptor substrate 1 (IRS-1), a member of a growing family of adaptor proteins that link upstream kinases to downstream signaling pathways (18). Tyrosine phosphorylated IRS-1 binds to various SH2-containing signaling molecules, including p85, the regulatory subunit of phosphatidylinositol 3-kinase, Grb2, SHP2, and Nck, linking the insulin receptor to a myriad of signaling pathways that have been implicated in mediating the cellular responses to insulin including growth and metabolic signals (reviewed in 18). By analogy to IRS-1, we postulated that Tub may function downstream of the insulin receptor. Consistent with this model, we demonstrate that Tub can serve as a substrate for the insulin receptor protein-tyrosine kinase both in vivo and in vitro. Furthermore, in addition to phosphorylation by the insulin receptor, Tub is phosphorylated in vitro by Abl and JAK2 but not by EGFR and Src proteintyrosine kinases. We also show insulin-enhanced association of Tub with SH2 domains, suggesting a functional role for Tub as an adaptor protein, linking the insulin receptor to downstream signaling cascades.

EXPERIMENTAL PROCEDURES
Cell Culture-CHO cells stably expressing the wild type human insulin receptor (CHO-IR) (19) were cultured in RPMI supplemented with 10% fetal calf serum and 400 g/ml G418 (Life Technologies, Inc.). The CHO-IR cells were a generous gift of Dr. Morris White (Joslin Institute, Harvard Medical School). PC12 cells (ATCC) were cultured in collagen-coated plates in Dulbecco's modified Eagle's medium supplemented with 5% fetal calf serum and 10% horse serum.
Antibodies-The anti-Tub serum was raised against full-length His 6tagged wild type Tub (murine) and was kindly provided by Dr. Rene Devos (Roche-Gent Research Institute). The anti-HA mouse monoclonal antibody (clone 12CA5) was purchased from Roche Molecular Biochemicals. The anti-phosphotyrosine mouse monoclonal antibody conjugated to horseradish peroxidase (RC20H) is from Transduction Laboratories (Lexington, KY), whereas the mouse monoclonal 4G10 was acquired from Upstate Biotechnology (Lake Placid, NY). The anti-PLC␥1 antibodies used were mouse monoclonal clone 10 from Transduction Laboratories, and the rabbit polyclonal (530) was from Santa Cruz Biotechnology (Santa Cruz, CA).
Plasmids and cDNAs-Mouse wild type Tub cDNA was obtained by polymerase chain reaction from mouse brain total cDNA. The coding region was verified by sequencing. To enhance the translation efficiency of the cDNA, site directed mutagenesis was used to place a "Kozak" translation initiation consensus motif at the ATG of the cDNA (ggatC-CACCATG). In addition to this mutation, for ease of subcloning, an EcoRI site was introduced just 3Ј of the cDNA termination codon. To aid in the detection of Tub protein during transfections, WT-Tub was introduced as 5Ј-BamHI Ͼ EcoRI-3Ј restriction fragments into the pN10 3X flu (HA) transient expression vector.
GST Fusion Proteins-The N-terminal and C-terminal SH2 domains of p85 (phosphatidylinositol 3-kinase regulatory subunit) and Abl-SH2 were kindly provided by Dr. Lewis C. Cantley (Department of Cell Biology, Harvard Medical School) and Dr. Richard Van Etten (Center for Blood Research, Harvard Medical School), respectively. All others were obtained from commercial sources (Upstate Biotechnology).
Transfection, Immunoprecipitation/GST-SH2 Precipitation, and Immunoblotting-CHO-IR cells were split 1:5 onto 100-mm plates 24 h prior to transfection. We used the LipofectAMINE (Life Technologies, Inc.) method to transfect CHO-IR cells with either empty expression vector (pN10) or the expression vector containing HA-tagged Tub (8 g/100 mm plate). The medium was replaced 24 h post-transfection with serum-free RPMI, and the cells were maintained in serum-free medium for at least 16 h prior to the addition of growth factors. PC12 cells were split 1:5 (from an 80% confluent dish) onto collagen-coated 100-mm plates. Two days later the cells were placed in low serum Dulbecco's modified Eagle's medium (0.05% fetal calf serum, 0.1% horse serum) for at least 16 h prior to the addition of growth factors.
In Vitro Kinase Assay-Purified rat IRS-1 (Upstate Biotechnology, Lake Placid, NY), GST, and GST-Tub (1 g/reaction) were incubated or not with 2 l of the following purified kinases: baculovirus-expressed insulin receptor kinase domain (bIRK), p60 c-Src (Src), Abl, EGFR, or JAK2 (Calbiochem, San Diego, CA), as indicated. The reactions were started by the addition of 10 l of 2ϫ kinase buffer (100 mM Hepes, pH 7.5, 200 M ATP, 300 mM NaCl, 10 mM MnCl 2 , 4 mM dithiothreitol, 100 M sodium vanadate, 10 Ci of [␥-32 P]ATP) in a final volume of 20 l, incubated for 10 min at room temperature and stopped by the addition of gel loading buffer. The samples were resolved by 10% SDS-polyacrylamide gel electrophoresis and transferred onto nitrocellulose. The phosphorylated proteins were revealed by autoradiography of the membrane.

Putative Tyrosine Phosphorylation Sites on Tub-Although
Tub does not share any regions or domains of homology with other known proteins, closer inspection of the amino acid sequence revealed 10 potential tyrosine phosphorylation sites on Tub as indicated in Fig. 1 and 438 are conserved in all Tub family members from Arabidopsis to man. The sequence surrounding tyrosine 464 (YIVM) fits the consensus motif YXXM for phosphorylation by insulin and IGF-1 receptors (9,18). Further analysis of Tub amino acid sequence by a program being developed in Dr. Lewis Cantley's laboratory that searches for sites likely to be phosphorylated by protein kinases predicted that, in addition to tyrosine 464, tyrosines 343 (YDNG), 371 (YETN), and 438 (YVLN) may serve as potential sites of phosphorylation by protein-tyrosine kinases.
Tub Undergoes Tyrosine Phosphorylation in Response to Insulin and IGF-1 in CHO-IR Cells-To investigate whether Tub becomes phosphorylated on tyrosine residues in response to insulin or IGF-1, we transfected HA-tagged wild type Tub (HA-Tub) into CHO-IR cells, which do not express endogenous Tub. Following treatment with either insulin or IGF-1, WCLs, anti-Tub, or anti-phosphotyrosine (␣-Tyr(P)), immunoprecipitates were analyzed for the presence of Tub and phosphotyrosine-containing proteins by Western blot (WB). Analysis of the WCL by ␣-Tyr(P) immunoblot demonstrated tyrosine phosphorylation of several proteins in response to insulin and IGF-1 (Fig. 2, top panel). An ␣-Tyr(P) immunoblot of ␣-Tub immunoprecipitates revealed that Tub was phosphorylated following insulin and IGF-1 stimulation (Fig. 2, middle panel). The lower level of HA-Tub tyrosine phosphorylation in response to IGF-1 reflects the lower number of IGF-1 receptor in these cells (22). Anti-Tub immunoblotting of the WCL demonstrated equivalent amounts of Tub expressed under the different conditions (Fig.  2, bottom panel).
To further characterize the effect of insulin on Tub phosphorylation we examined the dose dependence and kinetics of response. As shown in Fig. 3A (top panel), transfected Tub became tyrosine phosphorylated when CHO-IR cells were treated with 10 nM insulin. This phosphorylation peaked at 50 nM. In this experiment, phosphorylation of bands corresponding to IR or IRS-1, in the WCL, peaked at 10 and 50 nM insulin, respectively (Fig. 3A, lower panel). Phosphorylation of Tub occurred as early as 1 min after insulin addition (Fig. 3B) and was maintained for up to 60 min. These results indicate that insulin induces rapid and sustained tyrosine phosphorylation of Tub at physiological doses.
Endogenous Tub Becomes Tyrosine Phosphorylated in Response to Insulin but Not EGF in PC12 Cells-In the mouse, Tub is expressed only in neuronal tissues (23). 2 Therefore, we examined a panel of cell lines by Western blot to determine the best expressor of Tub. All neuronal cell lines examined express Tub at low levels (PC12, GT1, and SHY-5Y); among them PC12 cells presented the highest level of expression (data not shown). Tub expression was not detected in any cell lines of non-neuronal origin (data not shown). Therefore we selected PC12 cells to examine whether endogenous Tub becomes phosphorylated on tyrosine in response to growth factors.
Undifferentiated PC12 cells, following serum starvation, were treated with either EGF (100 ng/ml) or insulin (100 nM) and lysed, and the WCLs or anti-Tub immunoprecipitates were analyzed for the presence of phosphotyrosine-containing proteins and Tub by WB. Anti-Tyr(P) immunoblot of the WCL showed that a number of proteins, including the EGFR itself, became tyrosine phosphorylated in response to EGF (Fig. 4, top panel, lane 2). Less phosphorylation was observed in the WCL in response to insulin (Fig. 4, top panel, lane 3), which is due to the lower number of insulin receptors in PC12 cells as compared with EGFR. Nevertheless, Tub became phosphorylated in response to insulin but not to EGF in these cells as shown (Fig. 4, middle panel, compare lanes 3 and 2). Equivalent amounts of Tub were detected in all three anti-Tub immunoprecipitates (Fig. 4, lower panel).
Tub Is an in Vitro Substrate for Insulin Receptor, Abl, and JAK2 Kinases but not for Src and EGFR Kinases-To determine whether Tub is a direct substrate for the insulin receptor protein-tyrosine kinase, we performed in vitro kinase assays. As a positive control we used purified rat IRS-1, which has been previously shown to be directly phosphorylated by the insulin receptor both in vitro and in vivo (24,25). IRS-1, GST, or GST-Tub were incubated in the absence or presence of purified bIRK. As shown in Fig. 5A, only IRS-1 and GST-Tub (lanes 6 and 8) but not GST (lane 7) were phosphorylated when incubated with bIRK. Autophosphorylation of bIRK was also observed (lanes [5][6][7][8]. No phosphorylation was detected in the reactions in which bIRK was not added (lanes 1-4). This result 2 T. Chickering and R. Kapeller, unpublished results. indicates that Tub, like IRS-1, can serve as a substrate for the insulin receptor protein-tyrosine kinase in vitro .
To evaluate the specificity of Tub phosphorylation, we tested whether Tub could be phosphorylated in vitro by other proteintyrosine kinases as well. As shown in Fig. 5B, GST-Tub became phosphorylated when incubated with purified Abl and Jak2 (lanes 8 and 16, respectively) but not when incubated with Src or EGFR (lanes 4 and 12, respectively). Both Src and Abl phosphorylated IRS-1 (lanes 2 and 6), whereas little or no phosphorylation of IRS-1 occurred in the presence of EGFR and Jak2 (lanes 10 and 14). These results suggest that Tub phosphorylation is specific and regulated by the recognition of tyrosine-containing sequences on Tub by a subset of PTKs.
Tub Associates with SH2-containing Proteins-Because WT-Tub is tyrosine phosphorylated in response to insulin, we examined whether this phosphorylation could induce the association of Tub with SH2 domains. Lysates from control and insulin-treated CHO-IR cells transiently expressing HA-Tub were incubated with various GST-SH2 fusion proteins immobilized onto glutathione-Sepharose beads. Bound HA-Tub was detected by anti-HA immunoblot. As shown in Fig. 6A, HA-Tub associated strongly with the SH2 domains of Abl (lanes 17 and 18), Lck (lanes 15 and 16), as well as the C-terminal SH2 domain of PLC␥ (PLC␥-CSH2, lanes 9 and 10). Insulin treatment enhanced this interaction. In contrast, Src-SH2 (lanes 19 and 20) and PLC␥-N-terminal SH2 domain (PLC␥-NSH2, lanes 11 and 12) associated only weakly with HA-Tub, but some insulin-dependent binding was still observed. Even less HA-Tub was detected in p85-CSH2 (lanes 6 and 7) and GRB2-SH2 (lanes 13 and 14) precipitates, and in p85-NSH2 precipitates HA-Tub was undectectable (lanes 5 and 6). No HA-Tub was detected in GST precipitates (lanes 3 and 4). An anti-Tyr(P) immunoblot of the same membrane revealed that all the SH2 domains tested were able to precipitate phosphotyrosine-containing proteins, confirming that they were all functional (data not shown). Ponceau S staining of the nitrocellulose membrane furthermore shows that all the SH2 domains were present at comparable levels (data not shown).
To examine whether the interaction between Tub and the various SH2 domains was mediated through the SH2 phosphotyrosine binding pocket, we incubated GST-PLC␥-CSH2 with 20 mM phenylphosphate prior to the addition of HA-Tub-containing lysates. Phenylphosphate binds to the phosphotyrosine binding pocket of SH2 domains and blocks their association with phosphoproteins. As shown in Fig. 6B, phenylphosphate blocked the binding of GST-PLC␥-CSH2 to phosphorylated HA- Tub (top panel, compare lanes 7 and 8 with lanes 5 and 6), confirming that this is a phosphotyrosine-dependent interaction.
To further investigate the ability of insulin to induce the association of Tub with SH2-containing signaling proteins, we examined whether transiently expressed HA-Tub interacted with endogenous PLC␥ in CHO-IR cells. We chose PLC␥ because it is expressed at reasonable levels in CHO-IR cells, whereas Lck, Src, and Abl are not. As depicted in Fig. 6C, HA-Tub was detected in anti-PLC␥ immunoprecipitates, and insulin enhanced the association of Tub with PLC␥. These results suggest that Tub can associate with endogenous SH2containing signaling proteins in an insulin-dependent manner stimulation and may serve as an adaptor protein linking the insulin receptor to intracellular signaling cascades.  lanes 2 and 6), GST (lanes 3 and 7), or GST-Tub (lanes 4 and 8) were incubated with (lanes 5-8) or without (lanes 1-4) bIRK in kinase buffer. The reactions were resolved by SDS-polyacrylamide gel electrophoresis, and an autoradiograph is shown. B, Tub is phosphorylated in vitro by Abl and JAK2 but not by Src and EGFR. Src (lanes 1-4), Abl (lanes 5-8), EGFR (lanes 9 -12), and JAK2 (lanes 13-16) were incubated with IRS-1 (lanes 2, 6, 10, and 14), GST (lanes 3, 7, 11, and 15) and GST-Tub (lanes 4, 8, 12, and 16) in kinase buffer, and the samples were resolved as above. An autoradiograph is shown.

DISCUSSION
Tub and Tub-like proteins, because of their highly conserved C termini, the wide distribution across species, and the phenotypical alterations observed in tubby mice, are thought to play an important and basic role in the tissues where they are expressed (7,8,26). Because of the retinal and cochlear degeneration observed in tubby mice, apoptosis has been suggested as the mechanism underlying the tubby phenotype. To date, however, no other programmed cell death-induced alterations have been observed in tubby mice that could explain the obesity phenotype, although that possibility cannot be ruled out. In addition to tubby, mutation in the Tub-like protein 1 (TULP1) in humans has been implicated in RP14, an autosomal recessive form of retinitis pigmentosa (27)(28)(29).
As a first step to gain more insight into the biological function of Tub, we explored the possibility that Tub is an intracellular signaling protein. Here we show that insulin as well as IGF-1 can induce tyrosine phosphorylation of transiently expressed HA-Tub in CHO-IR cells (Fig. 2), with a dose response curve and kinetics (Fig. 3) that match that of other insulin receptor substrates such as IRS-1 and CBL (24, 30, 31). However, the EC 50 for Tub phosphorylation is higher (about 50 nM) than that observed for some other insulin receptor substrates, suggesting that in CHO-IR cells phosphorylation may be mediated in part by the IGF-1 receptor. In this experiment, however, phosphorylation of a band in the WCL that corresponds to IRS-1 also peaked at 50 nM insulin, although the IR was maximally phosphorylated at 10 nM insulin. These results suggest that Tub may serve as a substrate for both insulin and IGF-1 receptors.
In PC12 cells, endogenous Tub is tyrosine phosphorylated in response to insulin but not to EGF (Fig. 4). The phosphorylation of endogenous Tub by insulin in PC12 cells is low, however, and that may be due to the low levels of expression of both Tub and insulin receptor in these cells. Interestingly, although Tub does not possess any of the known domains that mediate recruitment to either the plasma membrane (i.e. PH domain, lipid acceptor sites) or to the receptor itself (i.e. SH2 or PTB domains), we have evidence that a fraction of Tub localizes at the plasma membrane in PC12 cells and that the remainder is localized in the nucleus. 3 In CHO-IR cells Tub also localizes to the plasma membrane and nucleus, and a point mutation in the C-terminal domain of Tub abolishes the membrane localization as well as insulin-induced Tub phosphorylation (data not shown), indicating that only the plasma membrane fraction is tyrosine phosphorylated in response to insulin. We are currently investigating the mechanism by which Tub is recruited to the plasma membrane.
Further evidence that Tub serves as a direct substrate for the catalytic domain of the insulin receptor (bIRK) was obtained in an in vitro kinase assay (Fig. 5A). Tub is furthermore phosphorylated in vitro by some other purified PTKs, such as Abl and Jak2, but not by Src and EGFR (Fig. 5B). The lack of Tub tyrosine phosphorylation by purified EGFR kinase agrees with the data obtained using PC12 cells stimulated with EGF. (Fig. 4). These results indicate that PTK-mediated Tub phosphorylation is specific and determined by the recognition of tyrosine-containing acceptor sites on Tub by a subset of PTKs. The data, however, also suggest that like IRS-1/IRS2 (reviewed in Ref. 18) and Cbl (reviewed in Ref. 32), Tub may be a target for several PTKs. The specificity of Tub as a substrate for PTKs is likely to be dictated by its restricted expression pattern and its co-expression with PTKs (see below).
Transiently expressed Tub mutant protein, predicted to arise from the tubby mutation, was not phosphorylated in response to insulin in CHO-IR cells (data not shown). Interestingly, the Tub mutant protein lacks three of the putative tyrosine phosphorylation sites of Tub (tyrosines 464, 481, and 483). However, when we probed brain lysates from tubby mice with the anti-Tub antibody, we were unable to detect the Tub mutant protein, 4 suggesting that the tubby mutation generates an unstable protein or that it is not translated from the larger tub transcript present in tubby mice (7). Furthermore, in parallel with this study, we have generated tub knockout mice (tubϪ/Ϫ) and shown that the phenotype of these mice, including increased weight gain, hyperinsulinemia, and photoreceptor loss, is indistinguishable from that of tubby mice, indicating that the tubby phenotype is due to loss of function of the Tub protein and not to altered function because of the difference in the C-terminal domain of the Tub mutant protein. 4 It is well documented that SH2 domains recognize phosphotyrosine-containing motifs in a sequence-specific manner (10,34) and that this recognition dictates the nature of the signaling complexes that are then formed in response to the activa- tion of cellular protein-tyrosine kinases. Our results demonstrate that Tub indeed binds selectively and with different affinities to a subset of SH2 domains, including PLC␥-CSH2, Lck-SH2, Abl-SH2, and Src-SH2 (Fig. 6A). Preincubation of the PLC␥-CSH2 domain with phenylphosphate abolished its association with phospho-HA-Tub (Fig. 6B), indicating that the association of Tub with SH2 domains is mediated through the binding of the phosphotyrosine-binding pocket in the SH2 domains to specific phosphotyrosine-containing sequence(s) on Tub. Surprisingly, however, Tub did not interact in an insulindependent fashion with the SH2 domains of p85, although it contains a site in its C terminus, DY 464 IVM, that is predicted to bind to p85-SH2 domains (34,35). Interestingly, Tyr 464 is not conserved among all Tub family members, it is replaced by phenylalanine in TULP2 (26). In addition we have replaced the last three C-terminal tyrosines (464, 481, and 483) with phenylalanine. This mutant behaves identical to wild type Tub with regard to insulin-induced tyrosine phosphorylation, association to SH2 domains, and subcellular localization (data not shown). This result indicates that the last three C-terminal tyrosines are neither sites for phosphorylation by the insulin/ IGF-1 receptor or SH2 docking sites. We are currently investigating the sites on Tub that are phosphorylated on tyrosine and that mediate the interaction with SH2 domains.
In summary, these results suggest that Tub is a downstream target for the insulin and/or IGF-1 receptor and may function as a link to a diverse array of downstream signaling pathways. IRS-1, IRS-2, and Cbl, for example, are known substrates for the insulin receptor protein-tyrosine kinase and have been shown to associate with subsets of SH2-signaling proteins upon tyrosine phosphorylation in response to insulin (18,31,36). These substrates work as adaptor molecules and play a key role in insulin signaling by linking the insulin receptor to phosphatidylinositol 3-kinase and the Ras/MAPK pathway. Tub can potentially play a similar role linking the IR to a different subset of SH2-containing signaling proteins, possibly coupling the insulin receptor to PLC␥-mediated phosphatidylinositol turnover and Src-like kinases. In mice, Tub is expressed only in neuronal-derived tissues, including the hypothalamus, retina, olfactory epithelium, dorsal root ganglia, and adrenal medulla (23). 2 We postulate that in these cells, Tub plays an IRS-1/2like role in insulin/IGF-1 signaling, by linking the activated protein-tyrosine kinase receptor to downstream signaling pathways. In addition, Tub, like IRS-1, may also serve as a substrate for JAK2, which has been implicated in mediating leptin receptor signaling (37). Interestingly, we and others have observed that the expression of Tub in the central nervous system correlates with that of the insulin and IGF-1 receptors (7, 15, 23, 38), 2 although co-localization in the same neurons has yet to be demonstrated. This is the first study proposing a functional role for the obesity gene Tub as an intracellular adaptor protein. Our current model proposes that in tubby mice the central insulin signal is deregulated. Insulin action in the central nervous system has been implicated in appetite suppression and energy homeostasis, perhaps through its ability to modulate the expression of peptides such as NPY, a key regulator of energy intake (12, 39 -41). Recently, Guan et al. (42) have shown that the regulation of NPY and pro-opiomelanocortin mRNA expression in the arcuate nucleus is impaired in tubby mice. Therefore, one could postulate that disruption of Tub function impairs central insulin-mediated regulation of energy homeostasis resulting in weight gain and insulin resistance, in the same manner that disrupting IRS-1 or IRS-2 results in systemic defects of insulin action (33,43,44). Alternatively, Tub may play a key role in the transmission of signal(s) generated by the insulin and/or IGF-1 receptor that promote neuronal survival, and obesity may result from the death of cells in the key hypothalamic nuclei that control body weight. We are currently conducting in vivo studies to distinguish between these two possibilities.