Guanine nucleotide exchange on heterotrimeric Gi3 protein controls autophagic sequestration in HT-29 cells.

Recent results have shown that autophagic sequestration in the human colon cancer cell line HT-29 is controlled by the pertussis toxin-sensitive heterotrimeric Gi3 protein. Here we show that transfection of an antisense oligodeoxynucleotide to the αi3-subunit markedly inhibits autophagic sequestration, whereas transfection of an antisense oligodeoxynucleotide to the αi2-subunit does not change the rate of autophagy in HT-29 cells. Autophagic sequestration was arrested in cells transfected with a mutant of the αi3-subunit (Q204L) that is restricted to the GTP-bound form. In Q204L-expressing cells, 3-methyladenine-sensitive degradation of long lived [14C]valine-labeled proteins was severely impaired and could not be stimulated by nutrient deprivation. Autophagy was also reduced when dissociation of the βγ dimer from the GTP-bound αi3-subunit was impaired in cells transfected with the G203A mutant. In contrast, a high rate of pertussis toxin-sensitive autophagy was observed in cells transfected with an αi3-subunit mutant (S47N) which has an increased guanine nucleotide exchange rate and increased preference for GDP over GTP. Cells that express pertussis toxin-insensitive mutants of either wild-type αi3-subunit (C351S) or S47N αi3-subunit (S47N/C351S) exhibit a high rate of autophagy.

The role of heterotrimeric G proteins (␣␤␥) as signal transducers is well known (1)(2)(3). Their interaction with activated receptors triggers guanine nucleotide exchange on the ␣ subunit. After dissociation from the ␤␥ dimer, the activated ␣-subunit (␣-GTP) acts on the appropriate downstream effector. Likewise the ␤␥ dimer can also act on specific effectors (4). During the past few years the function of heterotrimeric G proteins in membrane transport along the exo/endocytotic pathways has been recognized. This includes endoplasmic reticulum to Golgi (5,6) and intra-Golgi transports (7), constitutive and regulated vesicle budding from the trans-Golgi network (8), control of exocytosis (9,10), formation of endocytic vesicles (11), and endosome fusion (12). In polarized cells different trimeric G proteins were shown to control the sorting of proteins: G s protein governs the apical sorting from the trans-Golgi network (13) and basolateral to apical transcytosis (14) and G i3 protein is involved in the basolateral sorting of proteins (7,13).
Membrane fusion is also required during macroautophagy, a major cellular catabolic pathway whose terminus is the lysosomal compartment (15)(16)(17). Autophagy is a nonselective process which begins with the sequestration of various organelles and fractions of the cytosol into a membrane probably derived from the rough endoplasmic reticulum (18) or from another organelle called a phagophore (19,20) to form closed vacuoles (early autophagosomes). After maturation (see Ref. 20 and references therein for a recent discussion), these vacuoles acquire lysosomal enzymes by either direct fusion with lysosomes (21)(22)(23) or late endosomes/prelysosomes (24). We have previously shown that the impairment of N-linked glycoprotein processing (25)(26)(27) in undifferentiated HT-29 cells is the consequence of a constitutive autophagic pathway (28). Recently, we have shown that autophagic sequestration is controlled by a G i3 protein in this cell line (29). The regulation of autophagic sequestration by GTPases does not seem to be restricted to HT-29 cells since it has been reported that GTP␥S inhibits the de novo formation of autophagic vacuoles in permeabilized rat hepatocytes (30). An important step toward the understanding of the control of autophagy is to identify the state of activation of the G i3 protein (i.e. nucleotide binding) that regulates autophagic sequestration. In the present study using site-directed mutagenesis and stable cell transfections, we demonstrate that autophagic sequestration is switched on when the G i3 protein is bound to GDP. In contrast, autophagic sequestration is inhibited when the G i3 protein is bound to GTP. Autophagy is also reduced when dissociation of the ␤␥ dimer from the GTP-bound ␣ i3 -subunit is inhibited. Using G i3 pertussis toxin (PTX) 1 -resistant mutants we show that the GDP-bound form of G i3 must attach to an intracellular membrane in order to initiate autophagic sequestration.
Construction of Antisense ␣ i2 -and ␣ i3 -Subunit Expression Vectors-The 39 bases of the 5Ј-noncoding region, immediately upstream of and including the ATG translation initiation codon of G␣ i2 (5Ј-GCGTGT-GGGGGCCAGGCCGGGCCGGCGGACGGCAGGATG-3Ј) and G␣ i3 (5Ј-GCGAGCCAGGGCCCGGTCCCCTCTCCGGCCGCCGTCATG-3Ј) were selected for use as antisense probes to take advantage of the diversity of the nucleotide sequence in this region and to provide specificity (48% identity in 5Ј-noncoding region versus Ͼ85% identity in coding region) (32,33). Oligonucleotides of these sequences were synthesized, and the double-stranded DNAs were inserted into the EcoRI cloning site of the polylinker of a pcDNA3 vector. The orientation of inserts was analyzed by DNA sequencing.
Cell Culture and Transfection of HT-29 Cells-HT-29 cells were cultured as described previously (34). Plasmids (20 g) containing the cDNA sequences of either G␣ i3 -S47N, G␣ i3 -Q204L, G␣ i3 -G203A, G␣ i3 -C351S, G␣ i3 -S47N/C351S mutants or G␣ i2 and G␣ i3 antisense oligodeoxynucleotides were introduced into exponentially growing HT-29 cells by the calcium phosphate precipitation method (29). Twenty-four hours after transfection, cells were grown in selective medium containing G418, 400 g/ml, for at least 3 weeks. Resistant cells were cloned by serial dilution, then several clones derived from each transfection were then selected and screened for their level of either ␣ i2 or ␣ i3 protein expression.
Sucrose Density Centrifugation-Membrane extracts were prepared as described previously (29). Membrane protein (1 mg) was extracted for 2 h at 4°C in 500 l of 0.4% Lubrol (w/v) in buffer containing 100 mM NaCl, 10 mM MgCl 2 , 1.25 mM EDTA, 20 mM HEPES (pH 8.0), 2.0 mM ␤-mercaptoethanol, 1.0 mM phenylmethylsulfonate, 50 g/ml leupeptin, and 50 g/ml soybean trypsin inhibitor (37). GTP␥S (final concentration: 1 or 100 M) or extraction buffer was added to 200 l samples of the supernatant of membrane extracts and the mixture was incubated 10 min at room temperature. When [ 35 S]GTP␥S was used, the final concentration of the radioactive nucleotide was 10 nM. Alternatively, samples of membrane extracts were primed with 1 M GTP␥S for 10 min at room temperature and then incubated for an additional 30 min period with 100 M GDP␤S. At the end of the incubation membrane extracts were combined with 35-l aliquots of the marker proteins and loaded onto 5-20% sucrose gradients made with the extraction buffer except that Lubrol was 0.1% (w/v) (37). Gradients were centrifuged at 55,000 rpm for 18 h in a Beckman SW 60 rotor at 4°C (38). Fractions (200 l) were collected and precipitated with trichloroacetic acid. The pellet was resuspended, boiled in SDS Laemmli buffer, and subjected to polyacrylamide gel electrophoresis and immunoblotting as described below. Chemiluminescence was quantitated after exposure of the blots to Kodak X-Omat AR film, and band intensity was recorded by densitometric scanning.
The quantity of ␣ i -subunits in each cell population was evaluated by densitometric analysis of autoradiograms of the nitrocellulose membranes. Stably transfected cell populations used in this study were selected by the following criteria: cells transfected by cDNAs encoding either for the wild-type ␣ i3 -subunit or its different mutants have a 2.8 -3.0-fold overexpression of the ␣ i3 -subunit when compared to untransfected cells. Cells transfected with the plasmid containing antisense oligodeoxynucleotides to ␣ i2 -and ␣ i3 -subunits have a 0.15-fold expression of ␣ i2 -and ␣ i3 -subunits when compared to untransfected cells.

Autophagic Sequestration of [ 3 H]Raffinose-[ 3 H]
Raffinose sequestration was monitored as previously reported (28) using a modification of the method of Seglen et al. (39). Briefly, cells (5 ϫ 10 6 /500 l) were incubated for 15 min at 37°C with 2 Ci of [ 3 H]raffinose, subjected to electroinjection (at ambient temperature), and incubated at 4°C for 30 min, followed by 15 min at 37°C. Cells were then washed twice with phosphate-buffered saline and dispersed in complete medium with 5 mM 3-MA, as indicated. When used, PTX was added 18 h before the electroinjection and was present during the incubation period. Subsequently, at different times, the cells were washed twice with 10% sucrose at 4°C, resuspended in 0.5 ml of 10% sucrose and homogenized by five strokes in a glass/Teflon homogenizer on ice. Immediately after homogenization, 0.5 ml of ice-cold phosphate buffer (100 mM potassium phosphate, 2 mM EDTA and 2 mM dithiothreitol, 100 g/ml bovine serum albumin, 0.01% Tween 20, pH 7.5) was added, and 1 ml of the cell homogenate was layered on top of a 4-ml cushion of buffered metrizamide/sucrose (10% sucrose, 8% metrizamide, 1 mM EDTA, 100 g/ml bovine serum albumin, 0.01% Tween 20, pH 7.5) and centrifuged at 7000 ϫ g for 60 min. The radioactivity associated with the pellet and total homogenate was measured by liquid scintillation counting.
Measurement of the Degradation of Long Lived Proteins-HT-29 cells were incubated for 18 h at 37°C with 0.2 Ci/ml L-[ 14 C]valine. Unincorporated radioisotope was removed by three rinses with phosphatebuffered saline (pH 7.4). Cells were then incubated either in complete medium or in nutrient-free medium (without amino acids and in absence of fetal calf serum) plus 0.1% of bovine serum albumin. Both media were supplemented with 10 mM cold valine. When required, 10 mM 3-MA was added throughout the chase period. After the 1st h of incubation, at which time short lived proteins were being degraded, the medium was replaced with the appropriate fresh medium, and incubation was continued for an additional 4 h. Cells were scraped into 0.5 ml of phosphate-buffered saline and the radiolabeled proteins in the 4-h medium, and cells were precipitated in 10% trichloroacetic acid, 1% phosphotungstic acid (v/v) at 4°C. The precipitated proteins were separated from the soluble radioactivity by centrifugation at 600 ϫ g for 10 min then dissolved in 1 ml of Soluene 350. The rate of protein degra-dation was calculated as acid-soluble radioactivity recovered from both cells and media.

Effect of ␣ i Antisense Oligodeoxynucleotides on Autophagic
Sequestration-Autophagic sequestration has been shown to be stimulated when the ␣ i3 -subunit was overexpressed in HT-29 cells (29). We speculated that a decreased expression of this subunit would suppress the autophagic sequestration. To answer this question, we have transfected HT-29 cells with a plasmid containing an antisense oligodeoxynucleotide of 39 bases of the 5Ј-noncoding region immediately upstream of and including the ATG codon. Fig. 1A shows that after transfection of the ␣ i3 -subunit antisense oligodeoxynucleotide the expression of the ␣ i3 -subunit was reduced by 85% in HT-29 cells (see also Table I). The effect of the ␣ i3 -subunit antisense oligodeoxynucleotide was specific since the expression of the ␣ i2 -subunit was not modified in transfected cells (Fig. 1A). The expression of the ␣ i1 -subunit was not investigated since it has been shown that this subunit is not expressed in intestinal cells (41). Similarly when cells were transfected with an ␣ i2 -subunit antisense oligodeoxynucleotide, only the expression of the ␣ i2subunit was reduced (Fig. 1A). In both cases the expression of the ␣ s -subunit was not modified in transfected cells (Fig. 1A). The autophagic sequestration capacity of cells transfected with antisense oligodeoxynucleotides was measured in two ways (28,40): (i) the rate of sequestration of the cytosolic enzyme lactate dehydrogenase (Table I) and (ii) the sequestration of electroloaded [ 3 H]raffinose (Fig. 1B). The autophagic sequestration capacity of cells transfected with ␣ i2 -subunit antisense oligodeoxynucleotide was similar to that observed in parental cells (Table I and Fig. 1B). In contrast both lactate dehydrogenase and raffinose sequestrations were severely impaired in cells transfected with ␣ i3 -subunit antisense oligodeoxynucleotide (Table I and Fig. 1B). These results confirmed that the G i3 protein is essential for autophagic sequestration. However, these experiments provide no information on the nucleotide requirement for this G protein to either stimulate or abrogate autophagic sequestration.
The GTP-bound Form of the ␣ i3 -Subunit Inhibits the Autophagic Sequestration-Several mutations in the conserved nucleotide binding regions of GTPases including ␣-subunits of trimeric G proteins have been characterized (for a review, see Ref. 42). We have chosen a site-directed mutagenesis strategy to investigate the influence of nucleotide-binding regions to the G i3 protein on autophagic sequestration (see also next paragraph). The first mutant we constructed was the Q204L mutant. This mutation has been shown to stabilize the ␣ i3 -subunit in its GTP-bound form (43).
The GTP binding to ␣-subunits leads to its dissociation from the ␤␥ dimer. Thus to investigate whether or not the dissociation of the ␤␥ dimer from the GTP-bound form of the ␣ i3subunit is important in inhibiting autophagic sequestration, we have transfected HT-29 cells with the G203A ␣ i3 -subunit mutant. This mutant is equivalent to the G203A mutant of the ␣ i1 -subunit (45), the G204A mutant of the ␣ i2 -subunit (46) and to the G226A mutant of the ␣ s -subunit (37). This last mutant cannot dissociate from the ␤␥ complex despite the presence of bound GTP (37,47). We have demonstrated the absence of ␤␥ dimer dissociation from the G203A ␣ i3 -subunit mutant by examination of the sedimentation rate of the G203A ␣ i3 -subunit in sucrose gradients in the presence or absence of GTP␥S (Fig.  3). Wild-type ␣ i3 -subunit sedimented more rapidly in the absence of the activating ligand than in its presence. This change in sedimentation rate is characteristic of the dissociation of ␣-subunits from ␤␥ subunits (37,38). In contrast, GTP␥S did not change the sedimentation rate of the G203A ␣ i3 -subunit mutant. This negative result was not due to the inability of the GTP analog to bind to the G203A ␣ i3 -subunit mutant. When sucrose gradient sedimentation experiments were carried out in the presence of [ 35 S]GTP␥S we detected radioactivity in a position corresponding to trimeric (G203A) G i3 , whereas no radioactivity was present at this position when the wild-type ␣ i3 -subunit was considered (compared panels A and B in Fig. 3). Radioactivity under the peak with the lower sedimentation coefficient could correspond to any GTP-binding proteins, in- cluding different activated ␣-subunits, present in both wildtype ␣ i3 -subunit-overexpressing and G203A-expressing cells. These results indicate that the G203A mutant is unable to undergo the conformational change to dissociate the ␣ i3 -subunit from the ␤␥ dimer as previously demonstrated for the G226A ␣ s -mutant (37,47).
Autophagic sequestration was reduced in G203A-expressing cells when compared to untransfected cells or wild-type ␣ i3subunit-expressing cells (Table I and Fig. 2A). The rate of raffinose sequestration per hour was 4.0 -4.5%, 13-14%, 1.9 -2.1%, and 0.6 -0.8% in untransfected, wild-type ␣ i3 -subunit-, G203A-, and Q204L-expressing cells, respectively. In each cell population studied the rate of sequestration of lactate dehydrogenase (Table I) and ornithine decarboxylase (data not shown), two cytosolic enzymes, was similar to those reported for [ 3 H]raffinose (see also next paragraph). This result demonstrates that autophagic sequestration in HT-29 cells is nonselective toward cytosolic markers as has been previously reported for rat hepatocytes (40). Accordingly the rate of 3-MAsensitive protein degradation was greatly reduced in G203Aexpressing cells and not stimulated by nutrient deprivation (Fig. 2B).
The GDP-bound Form of the ␣ i3 -Subunit Stimulates Autophagic Sequestration-The results reported above suggested that the GDP-bound form of G i3 switches on autophagic sequestration. To confirm that the trimeric form of GDP-bound G i3 stim-ulates the autophagic sequestration, we constructed another ␣ i3 -subunit mutant. This mutant, S47N is equivalent to the S54N mutant of the ␣ s -subunit (48). This last mutant was shown to have an increased guanine nucleotide exchange rate and increased preference for GDP over GTP (48). When membrane preparations were primed with 1 M GTP␥S and then incubated with an excess of GDP␤S, the S47N mutant and wild-type ␣ i3 -subunits behaved differently in sucrose gradients. The wild-type ␣ i3 -subunit sedimented as a monomer (Fig.  4A) indicating that GTP␥S binds almost irreversibly to the ␣ i3 -subunit in the presence of Mg 2ϩ (49). In contrast, the addition of GDP␤S shifted the sedimentation position of the S47N ␣ i3 -subunit mutant to the trimeric G i3 form sedimentation position (Fig. 4B). This result favors the fact that the S47N mutation increased the rate of guanine nucleotide triphosphate dissociation from the ␣ i3 -subunit as previously reported for the S54N mutant of the ␣ s -subunit (48).
The autophagic sequestration in S47N mutant ␣ i3 -subunitexpressing cells was very close to that observed in wild-type ␣ i3 -subunit-expressing cells (Table I and Fig. 5A). Lactate dehydrogenase and [ 3 H]raffinose sequestrations were inhibited in S47N expressing cells by 3-MA treatment ( Fig. 5A; Table I). Similarly to that reported for ␣ i3 -subunit-overexpressing cells (29), autophagy was inhibited by PTX treatment in S47Nexpressing cells (see Fig. 6 and Table I). In S47N-expressing cells, protein degradation was close to that observed in wild-  in panels A and  B). Each sample was then applied to linear 5-20% sucrose gradients with marker proteins. After trichloroacetic acid precipitation, fractions were analyzed by electrophoresis in SDS gels and immunoblotted with anti-␣ i3 antibody as detailed under "Experimental Procedures." Results were analyzed as described in the legend to Fig. 3. Arrows indicate the position of the marker proteins used: 1, cytochrome c; 2, carbonic anhydrase; 3, bovine serum albumin.
type ␣ i3 -subunit overexpressing cells when measured either in complete medium or in nutrient-free medium. Here again the 3-MA-sensitive degradation of radiolabeled proteins was stimulated by nutrient deprivation (Fig. 5B).
Membrane Coupling Is Required in Order for the GDP Occupied ␣ i3 -Subunit to Switch on Autophagic Sequestration-The results reported above suggest that when the ␣ i3 -subunit is occupied by GDP it stimulates autophagic sequestration. However, we have reported that ADP-ribosylation of G i3 is a strong inhibitor of autophagy in HT-29 cells (29). ADP-ribosylation of G i by PTX, which is most efficient when the ␣ i -subunit is coupled to ␤␥ (50), blocks the interaction of the ␣-subunits of G i -proteins with their receptors and consequently prevents the receptor-mediated exchange of GDP for GTP (51).
In order to further study the roles of ADP-ribosylation and GDP binding to the ␣ i3 -subunit, we have constructed two mutants of the ␣ i3 -subunit insensitive to PTX modification (52). After transfection, clones having a 3-fold overexpression of the ␣ i3 -subunits were selected (Table I), and the expression of the subunits was monitored before and after PTX treatment by high resolution urea SDS-PAGE (35). As expected, the migration of proteins from both C351S-and S47N/C351S-expressing mutants was unaffected by PTX treatment (Fig. 6, upper panel). C351S-and S47N/C351S-expressing mutants have a rate of autophagic sequestration similar to that of cells overexpressing the wild-type ␣ i3 -subunit and S47N ␣ i3 -subunit mutant (Fig. 6, lower panels; see also Figs. 2 and 5 and Table I). PTX treatment strongly inhibits autophagic sequestration in untransfected cells, overexpressing wild-type ␣ i3 -subunit and in S47N-expressing cells (Fig. 6, lower panels, and see Ref. 29), whereas autophagic sequestration was unaffected by PTX treatment in C351S-and C351S/S47N-expressing cells (Fig. 6, lower panels, and Table I) but inhibited by 3-MA (Table I). These results suggest that 3-MA and PTX act on different targets to inhibit autophagic sequestration and confirm that G i3 is the only PTXsensitive G protein involved in the control of autophagic sequestration in HT-29 cells. DISCUSSION Our recent results have shown that the autophagic pathway is controlled at the sequestration step by a G i3 protein (29). Data presented here extend our understanding of this new function for trimeric G proteins. Although it cannot be com-pletely excluded that the G i3 protein controls other catabolic pathways (e.g. microautophagy, direct uptake of cytosolic fractions by lysosomes) in HT-29 cells, several experiments indicate that G i3 governs the macroautophagic pathway and more probably the sequestration step: (i) in the different cell populations studied the sequestration of cytosolic markers was greatly reduced by 3-MA which is commonly recognized as an inhibitor of the macroautophagic sequestration step (44). (ii) 3-MA-sensitive degradation of long lived proteins is stimulated by nutrient-deprivation, a condition known to enhance the macroautophagic pathway (53,54), exclusively in cells that express either the wild-type ␣ i3 -subunit (Fig. 2B), the S47N mutant ␣ i3 -subunit (Fig. 5B), or PTX-insensitive mutant ␣ i3subunits (data not shown). In contrast, the rate of 3-MA-sensitive protein degradation is not modified by nutrient deprivation in cells that express the GTP-bound form of the ␣ i3 -subunit (Fig. 2B). (iii) 3-MA-sensitive lysosomal degradation of N-linked glycoproteins substituted with endoplasmic reticulum-type oligosaccharides is dependent upon the level of expression of the ␣ i3 -subunit (29) and of the guanine nucleotide exchange on the ␣ i3 -subunit (data not shown). Nevertheless, in order to unequivocally correlate G i3 protein expression and activation to the macroautophagic pathway, morphometric studies are in progress to evaluate the fractional volume of autophagic vacuoles in each HT-29 cell population.
Transfection of cells with antisense oligodeoxynucleotides gives strong additional support to the key role of G i3 in the control of the autophagic pathway. We have found that an 85% reduction of the expression of the ␣ i3 -subunit was sufficient to cause 90% inhibition of autophagic sequestration in HT-29 cells (see Table I). A Similar reduction of the expression of the ␣ i2 -subunit has been shown to inhibit the adenyl cyclase response by more than 70% in different cell systems (32,55). The stimulatory effect of G i3 in autophagy is due to intracellular events because, as previously reported for the wild-type ␣ i3subunit (29), all generated mutants of the ␣ i3 -subunit are predominently associated with intracellular membranes. 2 These observations must now be added to the mounting evidence for the intracellular distribution of trimeric G protein subunits in 2 E. Ogier-Denis, C. Bauvy, and P. Codogno, manuscript in preparation.
From the results reported in the present study we can hypothesize a mechanism by which Gi 3 controls the autophagic sequestration. When G i3 is occupied by GDP it can interact with a putative membrane receptor. This interaction would lead to autophagic sequestration. The fact that PTX, which inhibits the interaction of G i proteins with their receptors (51), is a very potent inhibitor of autophagic sequestration (this study) (29) suggests that physical contact between GDP-bound G i3 and its receptor is crucial to elicit autophagic sequestration. This assumption is strenghtened by the fact that cells expressing the PTX-insensitive C351S and S47N/C351S G i3 mutants have a high rate of autophagic sequestration that is not downregulated after toxin treatment. Thus when G i3 is occupied by GDP it can drive autophagy in two ways depending on whether or not G i3 interacts with membrane components. The autophagic capacity of cells expressing the S47N mutant was quantitatively similar to that observed in cells overexpressing the wild-type ␣ i3 -subunit. This finding may suggest that G i3 is preferentially bound to GDP and coupled to intracellular membranes in HT-29 cells. As the dissociation of GDP is the ratelimiting step in the trimeric G protein cycle, these results suggest a deficit in the membrane transmitted activation which induces GDP/GTP exchange on G i3 .
Accordingly when the amount of GTP-bound G i3 is increased, by overexpression of the Q204L mutant, a very low rate of autophagic sequestration was observed. These results are in line with those reported by Kadowaki et al. (30) who have demonstrated that GTP␥S inhibits the autophagic sequestration in rat hepatocytes. Whether or not the inhibition of autophagic sequestration requires the interaction of the GTP-bound G i3 with a membrane component is not known. However, results obtained with G203A transfected cells would not favor this hypothesis. This mutant is equivalent to the G226A mutant of the ␣ s -subunit and is unable to activate adenylyl cyclase in vivo (37) although the mutated subunit has GTP binding properties similar to that of the wild-type ␣ s -subunit (47). The defect in the activation of adenylyl cyclase was shown to be due to impairment of the GTP-bound G226A ␣ s -subunit to undergo a conformational change required for the dissociation of the ␤␥ dimer (37,47). Results describe here show that in cells transfected with the G203A mutant the rate of autophagic sequestration was greatly decreased. However, the incomplete inhibition of autophagic sequestration observed in this cell population could be the consequence of GTPase activity of the mutated subunit (47), which partially restored the pool of GDPbound G i3 protein.
Alternatively the inability of the G203A mutant to dissociate from the ␤␥ dimer raises the question as to the function of this complex in autophagy. Although we have not directly addressed this point, a role of the ␤␥ dimer in the control of autophagy seems unlikely because, when the level of free ␤␥ dimer is reduced by either overexpression of ␣ i2 -subunit (29) or increased by ␣ i2 -subunit antisense transfection (this study), no modification of autophagy was observed. A better understanding of the control of autophagic sequestration by G i3 protein will require the characterization of membrane components interacting with G i3 . However, the function of trimeric G proteins, including G i3 , in membrane remodeling is complex and probably does not exclusively depend upon the interaction with membrane proteins but also requires the recruitment of cytosolic proteins (61,62).