Research Article

G-Protein Signaling Through Tubby Proteins

See allHide authors and affiliations

Science  15 Jun 2001:
Vol. 292, Issue 5524, pp. 2041-2050
DOI: 10.1126/science.1061233


Dysfunction of the tubby protein results in maturity-onset obesity in mice. Tubby has been implicated as a transcription regulator, but details of the molecular mechanism underlying its function remain unclear. Here we show that tubby functions in signal transduction from heterotrimeric GTP-binding protein (G protein)–coupled receptors. Tubby localizes to the plasma membrane by binding phosphatidylinositol 4,5-bisphosphate through its carboxyl terminal “tubby domain.” X-ray crystallography reveals the atomic-level basis of this interaction and implicates tubby domains as phosphorylated-phosphatidyl- inositol binding factors. Receptor-mediated activation of G protein αq(Gαq) releases tubby from the plasma membrane through the action of phospholipase C–β, triggering translocation of tubby to the cell nucleus. The localization of tubby-like protein 3 (TULP3) is similarly regulated. These data suggest that tubby proteins function as membrane-bound transcription regulators that translocate to the nucleus in response to phosphoinositide hydrolysis, providing a direct link between G-protein signaling and the regulation of gene expression.

Obesity has become a severe worldwide epidemic that may soon displace malnutrition as the most significant single factor affecting human health (1,2). Obesity is a cause of or contributing element in a number of systemic diseases leading to increased risk of mortality. While obesity among the young is a problem of considerable magnitude, adult-onset obesity is a leading cause of decreased life expectancy and is a primary risk factor for type II diabetes, heart disease, and hypertension (3, 4).

The tubby strain of obese mice (5) provides one of the few defined models for adult-onset obesity. Thetubby gene, which is highly expressed in the paraventricular nucleus of the hypothalamus and several other brain regions, was identified by isolating the genetic locus that transmits this autosomal recessive obesity syndrome (6, 7).Tubby mice have a naturally occurring splice site mutation at the junction of the 3′ coding exon. Targeted deletion of thetubby gene results in a phenotype identical to that of the naturally occurring mutant, indicating that the tubby obesity syndrome indeed arises from a loss of function (8).

The tubby protein is a member of a homologous family with four members (tubby and TULPs 1 to 3) encoded in the human genome (9–11) and with others present in various multicellular organisms (12). These proteins feature a characteristic “tubby domain” of about 260 amino acids at the COOH-terminus that forms a unique helix-filled barrel structure; this COOH-terminal domain binds avidly to double-stranded DNA. Mutation of the tubby-like protein 1 gene TULP1 is the genetic origin of human retinitis pigmentosa type 14 (RP-14) (12–14). Mapping mutations from RP-14 patients onto the tubby COOH-terminal domain structure outlines a long, positively charged groove implicated in DNA binding (15). Tubby proteins also include NH2-terminal regions that, in their primary sequence, resemble activation domains from known transcription factors (15). Although the NH2-terminal domain of tubby can activate transcription of reporter constructs when fused to GAL4, potential targets of transcriptional regulation by tubby have not yet been identified. Furthermore, little is known about signaling pathways regulating the function of tubby.

Genetic and physiological studies have revealed a host of signaling molecules that play critical roles in the neuroendocrine regulation of body weight. These molecules include an abundance of neuropeptides and neurotransmitters, and their G protein–coupled receptors (GPCRs) expressed in the hypothalamus (16). GPCRs are among the primary cellular sensors implicated in systemic energy homeostasis. Nonetheless, intracellular components of these GPCR signaling systems specifically relevant to function in weight regulation have not yet been identified.

Here we show that tubby, a molecule for which loss of function leads to obesity, serves as a downstream effector of GPCRs that signal through the Gq subclass of Gα proteins. Tubby is anchored to the plasma membrane through binding phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P2]. Gαq, but neither other Gα proteins nor Gβγ subunits, releases tubby from the plasma membrane through phospholipase C–β (PLC-β)–mediated hydrolysis of PtdIns(4,5)P2, which results in tubby translocation to the nucleus. Efficient nuclear translocation can be induced by activation of Gαq-coupled receptors, including the serotonin receptor 5HT2c. These data suggest that tubby—and other tubby-like proteins—may provide a direct link between GPCR signaling and the regulation of gene expression.

Tubby associates with the plasma membrane, but also contains a functional nuclear localization signal. Confocal laser scanning microscopy of Neuro-2A cells transfected with green fluorescent protein (GFP)–tubby fusion proteins revealed that tubby initially localizes to the plasma membrane. However, beginning at about 36 hours after transfection, fluorescence starts to accumulate within the cell nucleus without diminishing at the plasma membrane (Fig. 1A) (17). The accumulation of fluorescence in the nucleus occurs almost exclusively in highly expressing cells, as judged by the intensity of fluorescence. This implied that tubby might translocate to the nucleus when specific membrane binding sites for tubby (see below) become saturated. We confirmed the dual localization to the plasma membrane and the nucleus by cell fractionation using both native tubby-expressing Neuro-2A cells and transfected human embryonic kidney (HEK)–293T cells (Fig. 1B).

Figure 1

Tubby has dual localization properties, directing it to both the plasma membrane and the nucleus. Nuclear localization is directed by the NH2-terminal domain, and plasma membrane localization is directed by the COOH-terminal “tubby domain.” (A) Confocal microscopy of GFP-tubby fusion protein Neuro-2A cell transfectants. The full-length protein (left) first exhibits a primarily membrane localization pattern, but then begins to show increasing nuclear staining. The COOH-terminal domain (amino acids 248 to 505) GFP fusion protein (middle) exhibits a pattern that is completely plasma membrane localized, and is excluded from the nucleus. The NH2-terminal domain–GFP fusion (amino acids 1 to 247) (right) presents a marked inverted pattern in which the protein localization is entirely nuclear. This behavior is reminiscent of other transcription factors involved in signaling, like NF-κB and SREBP, which are held outside of the nucleus until particular signaling events trigger their nuclear translocation. (B) (Top left) Naı̈ve Neuro-2A cells were fractionated, and the presence of endogenous tubby protein in the nuclear, cytoplasmic, and plasma membrane fractions was analyzed by SDS-PAGE protein immunoblotting (T19-Tub antibody, Santa Cruz). Tubby protein, apparent in three splice forms (68), was found prominently in both nuclear and membrane fractions. (Top right) An identical gel, developed with antibodies against the SP1 transcription factor, which has an exclusively nuclear localization. This control experiment demonstrates the purity of the subcellular fractionation. (Bottom) Tubby-transfected 293T cells were fractionated in a similar way and detected with anti-tubby antibody (Alpha Diagnostics, San Antonio, Texas; polyclonal antibody Tub11A).

To determine the contribution of the NH2- and COOH-terminal domains to the localization of tubby, we made GFP fusion constructs corresponding to each of these individual domains (18). The COOH-terminal domain directed localization exclusively to the plasma membrane, whereas the NH2-terminal domain directed localization exclusively to the nucleus (Fig. 1A). We mapped the functional nuclear localization signal (NLS) of tubby to the sequence K39KKR within the NH2-terminal domain (19) [Web fig. 1 (20)]. The dual localization of tubby is reminiscent of that of transcription factors like SREBP, NF-κB, SMADS, STATs, and N-FAT, which transit into the nucleus in response to particular cell signaling events (21–25). Therefore, the dual localization of tubby protein suggests that tubby may also function in a signaling role.

Tubby binds the plasma membrane through specific interaction with select phosphorylated inositol phospholipids. We found that tubby binds to select phospholipids in part through a fortuitous observation. While performing yeast two-hybrid experiments with a tubby COOH-terminal domain “bait” and a system for readout by Ras-pathway activation upon “bait” localization to the membrane (the CytoTrap system from Stratagene) (26), we found tubby autoactivation. This suggested that the tubby COOH-terminal domain itself binds to the plasma membrane, even in yeast. Owing to the limited similarity of the protein components of yeast and mammalian cell membranes, we hypothesized that tubby might interact with a common membrane component, such as phospholipid.

To explore this idea, we performed a set of lipid blot experiments in which nitrocellulose membranes spotted with various purified phospholipids (27, 28) were incubated with tubby COOH-terminal domain glutathione-S-transferase (GST) fusion protein. Binding was evaluated by staining of the washed membranes with antibody to GST (anti-GST) (Fig. 2A). These experiments revealed that the tubby COOH-terminal domain does not interact with abundant lipid components such as phosphatidylcholine or the acidic phospholipids phosphatidylserine and phosphatidic acid, but rather interacts with only a subset of inositol head-group phospholipids. The differential binding of tubby to inositol lipids appears to depend primarily on the positions of the phosphorylation sites of the inositol ring. Of the lipids we tested (27), tubby COOH-terminal domain protein bound avidly only to PtdIns(3,4)P2, PtdIns(4,5)P2, and PtdIns(3,4,5)P3. No appreciable binding was observed to any singly phosphorylated PtdIns or to PtdIns(3,5)P2. This suggested that tubby may bind preferentially to phosphatidylinositol lipids phosphorylated at adjacent ring positions, such as PtdIns(3,4)P2 and PtdIns(4,5)P2, but not to lipids whose inositol rings do not have this characteristic structure, such as PtdIns(3,5)P2.

Figure 2

The tubby COOH-terminal domain binds specifically to bidentate-phosphorylated phosphoinositide lipid head groups. (A) Purified GST-tubby COOH-terminal domain fusion protein was incubated with nitrocellulose strips spotted with various phospholipids (27). Bound protein was visualized by immunoblotting. Binding of the wild-type protein (left) is observed only to the phosphoinositides phosphorylated at adjacent positions of the inositol ring. This observation agrees well with the bidentate coordination of K330, observed as a central feature of the protein-phospholipid interface in the cocrystal structure inFig. 3. GST alone exhibits no binding (right). (B) Surface plasmon resonance (SPR) traces of wild-type tubby COOH-terminal domain protein binding to various lipid bilayers coated on BIACORE L1 biosensor chip. Tubby protein bound with high affinity only to the PtdIns(4,5)P2 surface, and not to PC, PtdIns (PI), or PtdIns(4)P (PI4P) surfaces.

To confirm these observations, we performed surface plasmon resonance (SPR) biosensor experiments to evaluate the binding of tubby COOH-terminal domain protein (cleaved from the GST fusion partner) to liposome bilayers containing different phospholipid head groups (29). The results of the SPR studies matched those of the lipid blots. A representative data set (Fig. 2B) demonstrates binding of tubby protein with high affinity to lipids phosphorylated in adjacent ring positions, such as PtdIns(4,5)P2, but not to phosphatidylcholine, phosphatidylinositol, or monophosphorylated inositol lipids such as PtdIns(4)P. The tubby-PtdIns(4,5)P2 complex was very stable, but could be disrupted with a short pulse of 10 mM NaOH, demonstrating that the reaction was reversible (30).

Tubby proteins localize to the plasma membrane in quiescent cells. Whereas many pleckstrin homology (PH) and FYVE domain–containing proteins require activation of phosphoinositide 3-kinase (PI 3-kinase) to initiate plasma membrane association with phosphoinositides phosphorylated on the 3-position (31–33), the plasma membrane localization of tubby proteins is not reduced by wortmannin, a potent PI 3-kinase inhibitor (34). This suggested that tubby might associate with the plasma membrane through binding PtdIns(4,5)P2, which is the most abundant bisphosphorylated inositol lipid in quiescent cells.

A subset of the known PH domains bind specifically to PtdIns(4,5)P2 to mediate protein localization to the plasma membrane. These include the PH domains from phospholipase C (PLC)–γ, spectrin, son-of-sevenless, pleckstrin, and dynamin (35). In addition, cytoskeletal proteins such as profilin and gelsolin are thought to be anchored to the membrane through binding to PtdIns(4,5)P2, and overexpression of the PtdIns(4,5)P2-binding PH domain from PLC-γ induces cell rounding and blebbing (36), which is thought to arise from competition for membrane binding sites. We note similar morphological changes arising in the most highly expressing tubby cells [for example, see Web fig. 6 (20)].

Structural basis of PtdIns(4,5)P2 binding by the tubby COOH-terminal domain. To define the atomic-level basis of the interaction of tubby with phosphoinositide lipids, we infused preformed tubby COOH-terminal domain crystals with soluble lipid head-group analogs and determined cocrystal structures of the complexes formed (37). In each case, the lipid head-group molecules bound at the same site on the tubby protein. Here we report the 1.95 Å resolution crystal structure of the complex formed withl-α-glycerophospho-d-myo-inositol 4,5-bisphosphate (GPMI-P2), an analog of the head group from PtdIns(4,5)P2 (Fig. 3).

Figure 3

Structural basis of PtdIns(4,5)P2 binding by the tubby COOH-terminal domain. (A) Ribbon diagram of the cocrystal structure with GPMI-P2, a soluble analog of the head group from PtdIns(4,5)P2, shown in a van der Waals sphere representation. GPMI-P2 binds at a site on the tubby β barrel nestled against helix 6A. (B) A view down the axis of the central helix, rotated 90° from the view in (A). The position of GPMI-P2 is indicated by the blue arrow. Mutations in TULP1 from RP-14 patients that were mapped to corresponding positions in the tubby structure, which are thought to line the likely DNA binding groove (15), are shown in red. All ribbon-type diagrams were produced with the program SETOR. (69) (C) GRASP (70) surface representation of an orientation similar to that in (A), showing that PtdIns(4,5)P2 binds in a cavity of positive charge. The charge distribution is color-coded with blue for positive (≥30 kT/electron), red for negative (≤−30kT/electron), and white for neutral. (D) Molecular details of the interaction. Three positively charged side chains, those of R332, R363, and K330, coordinate the phosphate groups. Of particular note, K330 coordinates in-between the 4- and 5-phosphates, suggesting a structural basis for the requirement of adjacent inositol phosphate groups to accommodate this bidentate coordination architecture. ExperimentalF oF cdensity is shown in red (1.8σ) and blue (3.0σ) and demarcates the positions of the most electron-dense groups: the phosphates and oxygen atoms. Many of the side chains shown in this binding groove are conserved among the known tubby proteins (see text for details). (E) Phosphatidylinositol lipid “PIP arrays” (Echelon Research) (27)—nitrocellulose membranes spotted with serial dilutions of the phosphatidylinositol and its biologically relevant phosphorylated forms—were used to assay the binding avidity of wild-type tubby COOH-terminal domain and K300A,R332A double-mutant GST fusion proteins. (Left) The wild-type protein binds equally to PtdIns(4,5)P2, PtdIns(3,4)P2, and PtdIns(3,4,5)P3, but not to other forms of phosphatidylinositol. (Right) The mutant protein shows no detectable binding. (F) GFP localization analysis of site-directed mutants of the crystallographically identified phosphoinositide binding site. In early-stage transfections, wild-type, full-length tubby protein exhibits a primarily plasma membrane localization. The panels boxed in blue represent mutations within the binding site, whereas the panels boxed in red represent mutation of positively charged residues outside the edges of the binding site. Alanine mutation of each of the coordinating amino acids within the binding site abolishes the ability of tubby to bind to the plasma membrane. This echoes the in vitro finding that these mutations also abolish the ability of purified tubby protein to bind bidentate PtdInsPn's. Within the binding site, the Y327F and Y343F mutants do not affect plasma membrane localization, consistent with the observation that phenylalanine is an evolutionarily accepted substitution in these positions. The mutations we tested that lie outside of the crystallographically identified binding site do not affect the localization behavior of tubby.

The amino acids that interact with GPMI-P2 are mostly in β-strands 4, 5, and 6 and helix 6A [Fig. 3 and Web fig. 2 (20)]. This site forms a pocket located at one end of the putative DNA binding groove, demarcated by mutations mapped to the structure from RP-14 patients (Fig. 3B). As a central feature of the binding interface, K330 coordinates between the 4- and 5-position phosphates of GPMI-P2, thus providing a structural basis for the preference for ligands that are phosphorylated on adjacent positions of the inositol ring (Fig. 3D). A similar binding mechanism has been observed for GPMI-P2 binding to the clathrin adaptor protein CALM-N (38) and inositol 1,4,5-trisphosphate (IP3) binding to the PLC-δ1 PH domain (31). In both the PLC-δ1 and CALM-N structures, two lysine side chains intervene peripherally between the phosphorylated ring positions of GPMI-P2, whereas in tubby K330 coordinates directly between them. In the tubby structure, the 4-position phosphate is additionally stabilized by a salt bridge to the side chain of R332. The side chain of R363 coordinates the inositol ring at the 3-position. This suggests that phosphoinositides phosphorylated at the 3-position could be additionally anchored through a salt bridge with this residue. Oxygen atoms of both the 4- and 5-position phosphoester bonds hydrogen bond to the side-chain NH2 group of N310. The two free oxygen atoms of the GPMI-P2 phosphatide group are jointly coordinated by the NH2 group of the N348 side chain (Fig. 3D).

The pattern of residue conservation among tubby-like proteins suggests that PtdIns(4,5)P2 binding is likely to be a function common to all members of the protein family [Web fig. 2 (20)]. The central positively charged phosphate-coordinating residues, K330 and R332, are conserved among all known tubby-like proteins. R363 and N310 are conserved in all mammalian TULPs; poor sequence similarity in this region prevents rigorous sequence comparison to TULPs from plants. N348 is also conserved in all TULPs except TULP3 and those from Arabidopsis andDrosophila. In TULP3 and Drosophila TULP, N348 is replaced by a cysteine and lysine, respectively, which likely could also participate in forming hydrogen bonds.

In the F oF c maps (for which the head-group ligand is not included in the model), clear density is observed at 4.0σ for each of the phosphates at positions that can be unambiguously fit to the GPMI-P2 structure (Fig. 3D). Furthermore, 4.0σ density is also observed for the oxygen atoms of the glycerol moiety. Clear density, however, is not observed for the inositol ring. This is similar to the observations for the crystallographic structures of PtdIns(4,5)P2 analogs binding to the clathrin adaptor protein CALM-N, for which the ring also lacked strong density (38). The inositol ring, perhaps owing to its ability to adopt different pucker conformations, appears to be substantially more disordered than other parts of the PtdIns(4,5)P2 molecule.

The structure of the tubby COOH-terminal domain–GPMI-P2 complex permitted identification of the putative PtdInsPn binding site on the tubby protein. To determine whether the crystallographically identified binding site was indeed responsible for binding to inositol phospholipids, we evaluated the binding of both wild-type tubby COOH-terminal domain and the double-mutant COOH-terminal domain GST fusion protein K330A,R332A using nitrocellulose membranes spotted with serial dilutions of di-C16 phosphatidylinositol lipids (27). Wild-type protein bound with apparently equal avidity to PtdIns(3,4)P2, PtdIns(4,5)P2, and PtdIns(3,4,5)P3, whereas the mutant protein failed to bind above background levels (Fig. 3E). In quiescent cells PtdIns(3,4)P2 and PtdIns(3,4,5)P3 should be present at vanishingly low levels, and therefore we attribute membrane binding as most likely arising from interaction with PtdIns(4,5)P2.

To evaluate the contribution of the phosphoinositide binding site to membrane localization, we generated a series of site-directed mutants for use in cell-based experiments. When side-chain mutations within this site were introduced into the full-length tubby GFP fusion, tubby localization to the plasma membrane was abolished, yielding subcellular localization of tubby that was constitutively nuclear (Fig. 3F). However, mutation of positively charged residues outside of the lipid binding site (R290A and R397A) or amino acids within the binding site (Y327F or Y343F) that do not directly coordinate the GPMI-P2 had no effect on tubby localization, which, like wild-type protein, was primarily membrane-bound during early-stage expression (Fig. 3F).

q activation removes tubby from the plasma membrane, enabling translocation to the nucleus. The binding of tubby, through its COOH-terminal domain, to PtdIns(4,5)P2suggested that the localization of tubby might be controlled by the action of PLC-β. The action of PLC-β is often regulated through the Gαq subset of G proteins, which includes Gαq and Gα11. Activation of a Gαq-coupled receptor results in activation of Gαq (to the active form Gαq*), and Gαq* in turn interacts with and activates PLC-β (39). PLC-β can also be activated through interaction with G-protein βγ subunits, which bind to PLC-β at a distinct site.

We therefore tested the effect of Gαq (Fig. 4A) and Gβγ [Web fig. 3B (20)] activation on the subcellular localization of tubby. First, we performed experiments in which GFP-labeled tubby was cotransfected with various G-protein expression plasmids (40). Wild-type Gαq, which is inactive unless stimulated by a Gαq-coupled receptor, had no effect on tubby localization. However, cotransfection with a constitutively active (Gαq*) expression plasmid resulted in the total removal of GFP-tubby fluorescence from the plasma membrane and the complete redistribution of fluorescence to the nucleus. The capacity of Gαq* to induce nuclear translocation was dependent on its own plasma membrane attachment, as revealed by the inability of a C9S,C10S palmitoylation-site double mutant to induce translocation [Web fig. 3A (20)].

Figure 4

Tubby associates with Gαq family proteins, and activation of these G proteins triggers nuclear localization of tubby. (A) A GFP fusion of full-length, wild-type tubby protein was transiently cotransfected into Neuro-2A cells with either wild-type or constitutively active mutant forms (indicated by asterisks) of various G proteins. None of the inactive G proteins affected the localization pattern of tubby. Whereas Gαo* and Gαs* also showed no activity, the constitutively activated Gαq family proteins, Gαq* and Gα11*, efficiently triggered complete translocation of tubby to the cell nucleus. These experiments gave similar results, irrespective of the cell type used for transfection. (B) Specific coprecipitation of Gαq with tubby. A GST-tubby fusion protein was cotransfected with wild-type and active forms of either the Gαo or Gαq proteins. Precipitation with glutathione agarose demonstrates a prominent coassociation of Gαq with tubby, whereas Gαo shows no such association. (C) TULP3 exhibits localization behavior essentially identical to that of tubby. GFP fusion proteins of full-length TULP3 expressed in Neuro-2A cells show both membrane and nuclear localization 48 hours after transfection. The COOH-terminal domain from TULP3 is exclusively membrane bound, and the NH2-terminal domain is found exclusively in the nucleus. Cotransfection with Gαq* induces nuclear translocation of the full-length TULP3 protein, as is also the case for tubby.

The Gαq family member Gα11 functions to induce tubby translocation in a manner identical to that of Gαq. Active Gα11 (Gα11*) induced total nuclear translocation of tubby, whereas wild-type Gα11, in the absence of receptor stimulation, had no effect on tubby localization. We also tested the action of other Gα proteins in their native and constitutively active forms. Gαo, Gαo*, Gαs, and Gαs* had no observable effect on tubby localization. Furthermore, the localization of tubby, and its responsiveness to Gαq*, were not affected by GFP as a fusion partner; identical results were observed by immunofluorescence with FLAG-tagged or GST-tagged tubby, as well as untagged protein. These experiments were also reproduced with essentially identical results in a number of different cell types, including Neuro-2A, NIH-3T3, HEK-293T, and hypothalamic GT-17 cells (34).

Cotransfection with Gβ2γ2 or Gβ2γ2 and PLC-β1 did not induce the nuclear translocation of tubby, suggesting a specific function for Gαq family proteins beyond the activation of PLC-β [Web fig. 3B (20)]. This led us to ask whether tubby might associate in a complex with Gαq. To test this possibility, we performed experiments in which GST-tubby and either untagged Gαq, Gαq*, Gαo, or Gαo* were transiently expressed in 293T cells, and then complexes were isolated from cell lysates with glutathione Sepharose beads (41). Coprecipitated Gα proteins were detected by protein immunoblot analysis. We found that either full-length or COOH-terminal domain GST-tubby proteins coprecipitated Gαq in either the active or inactive state. However, Gαo did not coprecipitate with either tubby construct (Fig. 4B). This suggests that Gαq can form a specific complex with the tubby protein, irrespective of its activation state. It is not clear whether this complex results from direct protein-protein interactions or whether other accessory factors are also required.

To determine whether Gαq-responsive signaling might be a feature common to tubby family proteins, we constructed GFP fusion proteins of TULP3 and its modular domains. Like tubby, the COOH-terminal domain TULP3 fusion protein associated with the plasma membrane, whereas the NH2-terminal domain exhibited nuclear localization (Fig. 4C). Moreover, TULP3 COOH-terminal domain GST fusion protein, when analyzed on phospholipid blots, exhibited binding specificity similar to that of tubby [Web fig. 4 (20)]. As was observed for tubby, cotransfection of GFP-TULP3 with Gαq* or Gα11*, but not with other G proteins, led to nuclear translocation of the fusion protein (Fig. 4C). This suggests the likelihood that other members of the tubby-like protein family also function as transducers of Gαq-coupled receptor signaling.

Activation of Gαq-coupled receptors results in nuclear translocation of tubby.q can be naturally activated by ligand stimulation of Gαq-coupled receptors. We examined whether the activation of Gαq-coupled receptors could induce the nuclear translocation of tubby, as was observed by cotransfection with Gαq*. To investigate whether tubby translocation is specific to Gαq signaling, we tested two receptors stimulated by the same ligand, acetylcholine receptors M1 and M2 (AChR-M1 and -M2). AChR-M1 signals through Gαq and Gα11, whereas AChR-M2 signals through Gαi and Gαo. Neuro-2A cells cotransfected with GFP-tubby and AChR-M1 showed a progressive change in GFP-tubby localization from the membrane to the nucleus upon stimulation with acetylcholine (42). However, no such activation of translocation was observed with AChR-M2 (43) (Fig. 5A), illustrating the specificity to Gαq family signaling.

Figure 5

Activation of tubby translocation by Gαq-coupled receptors. (A) Time course of the localization in Neuro-2A cells doubly transfected with GFP-tubby and the acetylcholine receptors M1 and M2. Only the M1 receptor is Gαq-coupled, whereas M2 is coupled to Gαoand Gαi. Acetylcholine stimulation of the M2 receptors does not affect the predominant membrane localization of tubby. However, stimulation of the M1 receptor leads to cytoplasmic localization of tubby by the 10-min time point, and complete nuclear localization by 120 min. (B) Tubby was transfected into 3T3-fibroblast cells (3T3) and fibroblasts stably transfected with the Gαq-coupled serotonin receptor 5HT2C. Normal fibroblast transfectants demonstrate predominant membrane staining, whereas 5HT2c stable cells show no membrane staining, with tubby localized exclusively in the nucleus. Because 5HT2cis constitutively active, the addition of serotonin is not necessary to trigger this translocation. Complete nuclear translocation is also observed in Neuro-2A cells doubly transfected with GFP-tubby and 5HT2c. (C) Inhibition of AChr-M1–induced tubby nuclear translocation by the specific PLC inhibitor U73122. Cells were transfected with full-length GFP-tubby fusion protein and AChr-M1, and preincubated with either DMSO (no inhibitor) or 1 μM U73122 (PLC inhibitor) or 1 μM U73343 (mock inhibitor). After 10 min of preincubation, 100 μM acetylcholine was added to trigger translocation. At each time point, 800 cells were evaluated for the location of fluorescence (nuclear or nonnuclear), and the results were displayed at time points of 30, 60, 120, and 300 min. The PLC inhibitor U73122 markedly decreased nuclear translocation of tubby protein, whereas mock inhibitor or the DMSO control each exhibited efficient nuclear translocation. (D) Gαq/Gα11-deficient fibroblasts derived from double-knockout mice (47) were used to demonstrate the requirement for Gαq family proteins for tubby nuclear translocation. Cells were transfected with full-length GFP-tubby and either 5HT2c receptor alone, wild-type Gαq alone, or Gαq and 5HT2c receptor together. Only the combination of these two molecules could induce the nuclear translocation of tubby, demonstrating the requirement for 5HT2c activation of Gαq to Gαq* and, conversely, the requirement for Gαq to tranduce the signal from 5HT2c receptor.

Tubby localizes almost completely to the nucleus in Neuro-2A cells transfected with another Gαq-coupled receptor, the serotonin receptor 5HT2c (44). Constitutive nuclear localization is observed with or without the addition of serotonin; however, 5HT2c is often constitutively active in certain cellular environments (45). Similar results were observed in fibroblasts: Cells stably expressing 5HT2c showed constitutive nuclear localization of GFP-tubby, whereas naı̈ve fibroblasts showed tubby localization primarily at the plasma membrane (Fig. 5B). Furthermore, 5HT2c receptor mutants previously shown to interfere with the Gαq-mediated signaling function of the receptor were also tested (45, 46). The ability of these mutant receptors to induce the nuclear translocation of tubby was substantially diminished in all cases [Web fig. 5 (20)], again showing the requirement for Gαq signaling.

To determine whether the tubby translocation induced by these Gαq-coupled receptors is indeed mediated through Gαq activation, we performed cotransfection experiments using a fibroblast cell line derived from Gαq/Gα11 double-knockout mice (47). In these cells, transfection of 5HT2c receptor alone did not induce nuclear translocation of tubby. However, transfection of wild-type Gαq along with 5HT2c receptor enabled the activation of tubby translocation (Fig. 5D). Thus, the activation of Gαq to Gαq* appears to be both necessary and sufficient for induction of tubby nuclear translocation in the cells we tested.

Nuclear translocation of tubby requires activation of PLC-β.q family proteins are known to couple receptor signaling to the activation of PLC-β (39). The observations that tubby is held at the plasma membrane at least in part through binding to PtdIns(4,5)P2 and that Gαqsignaling removes tubby from the plasma membrane suggest that the hydrolysis of PtdIns(4,5)P2 may be a central aspect of the mechanism of tubby translocation. To determine whether the action of PLC-β was necessary for nuclear translocation of tubby, we assayed the effect of including a specific PLC inhibitor in cotransfection experiments with a Gαq-coupled receptor, AChR-M1, and stimulating the receptor with its ligand acetylcholine (43). Activation of a Gαq-coupled receptor, rather than cotransfection with Gαq*, was used to test for translocation inhibition, because the latter method leads to an overwhelming nuclear translocation response for which it is difficult to measure inhibition.

The specific PLC inhibitor U73122 (48) provides a nearly complete blockade of acetylcholine-induced nuclear translocation of tubby–AChR-M1 double transfectants in the presence of acetylcholine (Fig. 5C). The structurally similar molecule U73343, which only weakly inhibits PLC, provided a negative control. U73343 does not appreciably block translocation. This demonstrates the requirement for PLC function in the GPCR-mediated translocation mechanism of tubby.

Although it has been demonstrated that PtdIns(4,5)P2 can function in plasma membrane protein localization through PH domains, it has not yet been shown definitively that enzymatic processing of PtdIns(4,5)P2 can regulate the localization of proteins in response to biological stimuli, although such a mechanism has been proposed (35). We have demonstrated that the plasma membrane localization of tubby does indeed arise from its interaction with PtdIns(4,5)P2, and furthermore that activation of PLC-β through Gαq signaling induces the nuclear translocation of tubby and other TULPs. Tubby proteins therefore represent the first protein family for which dynamic nuclear translocation is clearly regulated by PtdIns(4,5)P2 hydrolysis.

A model for tubby function. On the basis of the findings presented here, we have developed a schematic model for tubby function (Fig. 6). Tubby, bound at the plasma membrane through its association with PtdIns(4,5)P2, is likely to be associated with Gαq, which is also membrane bound. Activation of Gαq by a ligand-complexed GPCR activates Gαq to Gαq*, promoting the activation of PLC-β and the subsequent hydrolysis of PtdIns(4,5)P2. This diminishment of PtdIns(4,5)P2 releases tubby from the membrane, thereby enabling nuclear translocation, potentially leading to the transcriptional regulation of downstream genes. Receptor-induced hydrolysis of PtdIns(4,5)P2 can promote dissociation of a PtdIns(4,5)P2-specific PH domain from the plasma membrane (35). It seems likely, therefore, that the transient reduction of membrane PtdIns(4,5)P2 is sufficient to promote the dissociation of tubby from the plasma membrane.

Figure 6

Schematic model of Gαq signaling through tubby proteins. Tubby associates with Gαq (or Gα11) at the plasma membrane, as implied by immunoprecipitation experiments. Activation of Gαq to Gαq* by interaction with a GPCR-ligand complex leads to the activation of PLC-β. Hydrolysis of multiply phosphorylated inositol lipids in the microenvironment of this complex then leads to the dissociation of tubby from the plasma membrane, enabling its translocation to the nucleus, directed by the two functional nuclear localization sequences in the NH2-terminal domain. It is not yet clear whether IP3 remains bound to tubby after dissociation or whether other biochemical modifications prevent tubby membrane rebinding. Nuclear translocation enables tubby to interact with its DNA targets and function in regulation of gene expression. Whether other nuclear factors are essential to tubby function is not yet clear.

The observed association of Gαq with tubby suggests a mode of function whereby localized hydrolysis of PtdIns(4,5)P2 could be achieved by recruitment of PLC-β to the tubby-Gαq complex. This could enable PtdIns(4,5)P2 hydrolysis within the tubby microenvironment sufficient to enable dissociation of tubby from the plasma membrane without the need for wholesale hydrolysis of PtdIns(4,5)P2over the entire cytoplasmic surface of the plasma membrane. Notably, another mechanism for activation of PLC-β, ectopic expression of G-protein βγ subunits along with PLC-β1, fails to remove tubby from the plasma membrane [Web fig. 3B (20)], demonstrating the importance of association with Gαq. Moreover, activation of PLC-γ through overexpression of various tyrosine kinase receptors, leading to the hydrolysis of PtdIns(4,5)P2, did not induce translocation of tubby [Web fig. 6 (20)], further indicating an essential role for Gαq.

Hydrolysis of PtdIns(4,5)P2 by PLC-β produces the second-messenger molecules IP3 and diacylglycerol. IP3 activates Ca2+-release channels to raise the concentration of intracellular Ca2+, thus activating protein kinase C (PKC). The PKC inhibitor BIM-I does not alter the translocation behavior of tubby in GPCR activation experiments [Web fig. 7A (20)], suggesting that PKC does not play a role in this process. Furthermore, raising intracellular Ca2+concentrations through the use of the ionophore ionomycin, which should result in the activation of PKC, does not induce nuclear translocation of tubby [Web fig. 7B (20)]. Although signaling effects downstream of PLC-β cannot be ruled out, these data show that the most prominent pathway downstream of PLC-β—that mediated by PKC—is not involved in tubby nuclear translocation. Coupled with the observation of a specific binding site for PtdIns(4,5)P2 on tubby, and the requirement for PLC-β action, we think it is most likely that tubby is removed from the plasma membrane by the transient reduction of membrane PtdIns(4,5)P2 induced by PLC-β–mediated hydrolysis.

Several aspects of the molecular mechanism of tubby nuclear translocation remain unclear. What permits tubby to localize to the plasma membrane after its synthesis and avoid translocation to the nucleus before membrane binding? Similarly, what prevents rebinding to the membrane after receptor stimulation? Several plausible mechanisms could explain this behavior. For example, it is possible that the NLS of tubby is masked, like that of NF-κB (21), and some signaling event associated with Gαq activation may result in its unmasking. This could potentially involve a phosphorylation event, or could be induced by a conformational change induced by the structural rearrangement of Gαq upon adopting its active GTP-bound state. A similar unresolved question addresses the mechanism by which tubby dissociates from its complex with Gαq. Coprecipitation experiments indicate that tubby associates both with the active and inactive states of Gαq. Yet tubby translocates to the cell nucleus, whereas Gαq does not. Future studies will be required to resolve these questions. We have demonstrated here that tubby binds phosphoinositides phosphorylated at adjacent ring positions. This raises the question of whether forms of PtdInsPn other than PtdIns(4,5)P2are involved in tubby membrane association, and hence whether PI 3-kinase and PtdIns phosphatases may also play a role in tubby-mediated signaling.

The tubby protein may provide a vehicle for integrating the information content of Gαq signaling and insulin signaling: Tubby can be phosphorylated in response to insulin action (49), and the possibility exists that tubby could elicit different effects in the nucleus depending on its phosphorylation state. We note that two surface-exposed tyrosine residues, Y327 and Y343, are within the lipid binding site, suggesting that tyrosine phosphorylation could potentially release tubby from the membrane by disrupting interactions with inositol phosphates. However, mutation of either of these tyrosines to phenylalanine has no apparent effect on cellular localization. Furthermore, insulin stimulation in a variety of cell lines, including Chinese hamster ovary (CHO)–IR cells, which overexpress insulin receptor, fails to show an effect on the subcellular localization of tubby (34).

Potential implications for the role of tubby in the molecular biology of obesity. The regulation of systemic energy balance is controlled in part by a hypothalamic neural circuit that integrates multiple biological signals (16, 50). Among the most important classes of molecules that transduce these signals are seven–transmembrane helix GPCRs. Gene-targeting experiments have clearly defined a role for several GPCRs and their ligands in the regulation of satiety, energy homeostasis, and body weight (50–56). Here we show that tubby, a molecule for which loss of function leads to obesity, serves as a downstream effector of GPCRs that signal through the Gqsubclass of Gα proteins. Of the hypothalamic GPCRs thought to play a role in energy homeostasis, it has been demonstrated that the 5HT2c, bombesin, melanin concentrating hormone (MCH), melanocortin 4 (MC4), and dopamine D1 receptors are coupled to Gαq (57–60). It is tempting to speculate, therefore, that the role of tubby in energy homeostasis may involve signaling downstream of one or more of these receptors.

Although it is not as yet possible to determine which of these receptors, if any, regulate the nuclear translocation of tubby in the hypothalamus, similarities in the phenotype of tubby mice and 5HT2c receptor knockout mice (56,61, 62) are suggestive of a potential mechanistic connection. Tubby mice manifest an atypical obesity syndrome characterized by late-onset obesity and hyperinsulinemia without substantial dysregulation of glucose levels (5,63). These characteristics are mirrored in the phenotype of 5HT2c knockout mice. These mice begin to demonstrate obesity at 12 weeks of age, bearing a marked resemblance to the time course of obesity onset in tubby mice. Furthermore, both 5HT2c (44) and tubby (6) are most highly expressed in the paraventricular nucleus of the hypothalamus. In humans, activation of the 5HT2c receptor has been associated with the reduction of both body weight and subjective ratings of hunger (64). The ability of different Gαq-coupled receptors to induce the nuclear translocation of tubby suggests that cellular coexpression may provide one level of regulation of tubby function. The possible mechanistic association between tubby, the 5HT2c receptor, and other GPCRs provides an intriguing avenue for future investigation.

In summary, we have shown that tubby domains function in plasma membrane localization through binding to select phosphorylated inositol lipids; we have defined the structural basis of this interaction by x-ray crystallography; and we have shown that activation of Gαq-coupled receptors leads to the nuclear translocation of tubby through a PLC-β–dependent pathway. These observations define a mechanism by which tubby can serve as a critical intracellular messenger of GPCR signaling in energy homeostasis and suggest that other tubby-like proteins are also likely to function as plasma membrane-bound transcription regulators activated through G-protein signaling.

  • * To whom correspondence should be addressed at the Structural Biology Program, Mount Sinai School of Medicine, Room 16-20, 1425 Madison Avenue, New York, NY 10029, USA. E-mail: shapiro{at}


View Abstract

Stay Connected to Science

Navigate This Article