Characterization of a Goα Mutant That Binds Xanthine Nucleotides*

Several GTP binding proteins, including EF-Tu, Ypt1, rab-5, and FtsY, and adenylosuccinate synthetase have been reported to bind xanthine nucleotides when the conserved aspartate residue in the NKXD motif was changed to asparagine. However, the corresponding single Goα mutant protein (D273N) did not bind either xanthine nucleotides or guanine nucleotides. Interestingly, the introduction of a second mutation to generate the Goα subunit D273N/Q205L switched nucleotide binding specificity to xanthine nucleotide. The double mutant protein GoαD273N/Q205L (GoαX) bound xanthine triphosphate, but not guanine triphosphate. Recombinant GoαX (GoαD273N/Q205L) formed heterotrimers with βγ complexes only in the presence of xanthine diphosphate (XDP), and the binding to βγ was inhibited by xanthine triphosphate (XTP). Furthermore, as a result of binding to XTP, the GoαX protein underwent a conformational change similar to that of the activated wild-type Goα. In transfected COS-7 cells, we demonstrate that the interaction between GoαX and βγ occurred only when cell membranes were permeabilized to allow the uptake of xanthine diphosphate. This is the first example of a switch in nucleotide binding specificity from guanine to xanthine nucleotides in a heterotrimeric G protein α subunit.

G proteins transduce receptor-generated signals across the plasma membranes of eukaryotic cells. They are heterotrimeric complexes composed of ␣, ␤, and ␥ subunits. Each of the subunits belongs to a multigene protein family, containing at least 18 distinct ␣, 5 ␤, and 11 ␥ subunits. Hundreds of seventransmembrane receptors activated by a great variety of hormones, neuromediators, and growth factors are coupled to G proteins. Receptor-induced activation of a G protein leads to exchange of GDP for GTP bound to the ␣ subunit. The GTPbound ␣ subunit is released from the ␣␤␥ trimeric complex, and both free ␣ and ␤␥ dimers are capable of modulating activities of effector enzymes and ion channels (1)(2)(3). G protein-mediated signaling is complicated; a single receptor can activate more than one kind of heterotrimer, and both the activated ␣ and the ␤␥ subunits can interact with multiple effectors. For example, the thrombin receptor is known to couple to G 12 , Gi, and Gq family members (4), and physiological responses may be the result of contributions by both ␣ and ␤␥ subunits. Furthermore, cross-talk between these different G protein-regulated pathways makes the networks even more complex.
One way to analyze this complex network is to specifically activate a particular G␣ in vivo to discern its function without interference from other G proteins. As a first step toward this goal, we used site-specific mutagenesis to switch the nucleotide specificity of G␣ from guanine to xanthine nucleotides. In cells, xanothine monophosphate is an intermediate in the biosynthesis of GMP; however, the steady-state concentrations of XDP and XTP are relatively low (5). Thus, by subsequent introduction of XTP, we should be able to specifically activate the mutant protein. The ␣ subunits of heterotrimeric G proteins belong to the GTPase superfamily that also includes factors involved in ribosomal protein synthesis, such as EF-Tu, and a large number of Ras-like small guanine nucleotide binding proteins (6,7). Crystal structures of the ␣ subunits of transducin and Gi have been recently solved (8 -11). Both G␣ structures had nearly identical binding pockets for the guanine nucleotide, which was similar to the guanine nucleotide binding pocket revealed in the crystal structures of Ras (12) and EF-Tu (13,14). One of the conserved features was the interaction between a specific G␣ amino acid residue and the guanine nucleotide ring, i.e. a hydrogen bond from the side chain of a conserved aspartic acid (Asp-268 in transducin) to the N-1 nitrogen and the N 2 amine of the guanine ring (see Fig. 1a). Asp-268 of transducin belongs to a conserved motif (NKXD) found in the GTPase superfamily. It has been shown that the characteristic hydrogen bond formed with the aspartic acid residue determines the specificity of guanine nucleotide binding in other GTP-binding proteins, such as EF-Tu and Ras (15,16). A mutation of aspartate to asparagine at this position in several GTP binding proteins, including EF-Tu (17,18), Ypt1 (19), rab-5 (20,21), and FtsY (22) and adenylosuccinate synthetase (23), leads to active proteins regulated by xanthine nucleotides instead of guanine nucleotides. In this report, we studied the effect of the similar D273N mutation on nucleotide binding specificity of Go␣.

MATERIALS AND METHODS
Mutagenesis and Expression of the Go␣-Myristoylated recombinant mouse GoA␣ was expressed in Escherichia coli. Conditions for growth, induction, and lysis of the Go␣-expressing cells were described previously (24). The D273N mutation was introduced in both wild-type Go␣ and the activated mutant Go␣Q205L by oligonucleotide-directed mutagenesis. The oligonucleotide TTTCTAAACAAGAAAAATTTATTTG-GCGAGAAGATTAAGAAGTC was annealed to uracil-containing single-stranded DNA from the plasmids pGo␣ and pGo␣Q205L. The resulting vectors were designated as pGo␣D273N and pGo␣X.
Expression and Purification of His 6 -tagged Go␣-We subcloned wild type and mutant Go␣ cDNAs into the E. coli expression vector pET-15b (Novagen), which added a peptide of 20 amino acids MGSS(H 6 )-SSGLVPRGSH containing the His 6 tag and a thrombin site upstream of the amino terminus of Go␣. These clones were used to transform the E. coli strain BL21(DE3), and proteins were expressed. After harvesting the culture, cell extracts were resuspended in the binding buffer (5 mM imidazole, 0.5 M NaCl, 160 mM Tris-HCl, pH 7.9, 1 mM ␤Me). Binding to the Ni 2ϩ -NTA resin was according to the protocol provided by Novagen.
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  The His 6 -tagged protein was eluted with a gradient of imidazole concentration (5-500 mM). The Go␣ and various mutant proteins eluted at about 250 mM imidazole. Proteins were then transferred to TED buffer (20 mM Tris-HCl, pH 8.0, 1 mM EDTA, 1 mM DTT) with 0.1 mM MgCl 2 and 0.1 mM nucleotide diphosphate (GDP or XDP as appropriate) by gel filtration. Purified proteins were stored in 50% glycerol at Ϫ70°C.
Synthesis of XTP␥S-XTP␥S was synthesized from XDP and ATP␥S with nucleotide diphosphate kinase (NDK) as described previously (25). To produce 35  Proteolysis with Trypsin-Approximately 0.1 g of purified recombinant Go␣ was preincubated with nucleotide at room temperature for 30 min in the TED buffer. 10 ng of trypsin was then added to the mixture, and the reaction was terminated after 10 min by addition of an equal volume of 2 ϫ SDS-PAGE sample buffer and heating for 3 min at 100°C. The proteolytic pattern was subsequently analyzed by Western blot using antibodies against Go␣.
ADP-ribosylation by Pertussis Toxin-Pertussis toxin-catalyzed ADP-ribosylation was performed as described (24). Briefly, 0.1 g of recombinant Go␣ was mixed with 0.1 g of purified retinal ␤␥ subunit complex in the presence of the appropriate nucleotide and incubated for 10 min at room temperature before addition of the reaction mixture (final concentration of 20 mM Tris-HCl, pH 8.0, 1 mM EDTA, 2 mM MgCl 2 , 2 mM DTT, 0.5 M [ 32 P]NAD (20,000 cpm/pmol), and 10 g/ml pertussis toxin (List Biologicals)). Reactions were incubated for 30 min at room temperature and terminated by the addition of 5 ϫ SDS-PAGE sample buffer. Samples were resolved on SDS-PAGE. Gels were stained with Coomassie Blue, dried, and exposed to x-ray film.
Cell Culture and Transfection-COS-7 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. 1 ϫ 10 5 cells/well were seeded in 12-well plates 1 day before transfection. All transfection assays contained a total amount of 1 g of DNA; the plasmid pCIS encoding ␤-galactosidase was used to maintain a constant amount of DNA. To each well, 1 g of DNA was mixed with 5 l of lipofectamine (Life Technologies, Inc.) in 0.5 ml of Opti-MEM (Life Technologies, Inc.), and five h later, 0.5 ml of 20% fetal calf serum in Dulbecco's modified Eagle's medium was added to the cells. After 48 h, cells were assayed for inositol phosphate levels as described previously (26,27).
Permeabilization of COS-7 Cell Membranes-Transfected COS-7 cells were washed twice with phosphate-buffered saline and incubated in 200 l of permeabilization solution consisting of 115 mM KCl, 15 mM NaCl, 0.5 mM MgCl 2 , 20 mM Hepes-NaOH, pH 7, 1 mM EGTA, 100 M ATP, 0.37 mM CaCl 2 (to give a free Ca 2ϩ concentration of 100 nM), and 200 units/ml ␣-toxin with or without 0.1 mM XDP for 10 min at 37°C. Then 2 l of 1 M LiCl was added before the inositol phosphate assay.

RESULTS
To change the binding specificity of Go␣ from guanine nucleotides to xanthine nucleotides, we replaced Asp-273 by an asparagine residue, which was expected on the basis of structural analysis to coordinate with xanthine instead of guanine (Fig. 1b). This mutation was introduced into both the wild-type Go␣ subunit and the GTPase-deficient Go␣ mutant (Q205L). We chose Go␣ because myristoylated Go␣ can be expressed in E. coli, and it has been shown that many of the characteristics of the recombinant Go␣ protein are similar to those of the protein isolated from brain. To further characterize the function of XTP-bound Go␣ mutants, we purified these proteins in the form of non-myristoylated His 6 -tagged Go␣ by affinity chromatography on a Ni 2ϩ -NTA column. It has been shown that the non-myristoylated form of Go␣ has identical nucleotide binding properties compared with the myristoylated form, and it also forms trimeric complexes with ␤␥ subunits although the affinity to ␤␥ is much less than the myristoylated form (44).
Binding of GTP␥S and XTP␥S-The nucleotide binding of Go␣, Go␣D273N, and Go␣X (Go␣D273N/Q205L) was assayed with [ 35 S]GTP␥S and [ 35 S]XTP␥S. In E. coli crude extracts, Go␣ reached maximum binding of GTP␥S in about 30 min (Fig.  2a). As expected, Go␣ showed no affinity for XTP␥S. However, Go␣X revealed a switch in nucleotide specificity. As shown in Fig. 2b, Go␣X had high affinity for XTP␥S but not for GTP␥S. Interestingly, only the double mutant was active while Go␣D273N did not bind either GTP␥S or XTP␥S (data not shown). Go␣ binds GTP␥S very tightly in the presence of 1 M Mg 2ϩ (28,29). Both Go␣ (Fig. 2c) and Go␣X (Fig. 2d) did not exchange bound [ 35 S]NTP␥S when excess non-radioactive nucleotides were subsequently added.
The purified His 6 -tagged proteins in general retained the properties of the untagged myristoylated ␣ subunits. However, we detected some differences in nucleotide binding. His 6tagged Go␣ or Go␣X bound GTP␥S or XTP␥S, respectively, but the binding was less stable than with the untagged myristoylated protein. In the case of His 6 -tagged Go␣, the bound GTP␥S could be exchanged after excess non-radioactive GTP␥S was added (Fig. 2c). Similar behavior was observed in the XTP␥S binding of pure His 6 -tagged Go␣X, which also showed distinct nucleotide exchange after non-radioactive XDP or XTP were added to the binding reaction (Fig. 2d). The decrease in nucleotide affinity was apparently the result of the presence of the His 6 -tag. Although the nucleotide binding of His 6 -tagged proteins was less stable, the specificity of binding was clearly maintained, and the mutant bound the xanthine nucleotides rather than the guanine nucleotides. As expected, the purified single mutant Go␣D273N did not show any nucleotide binding activity (data not shown).
Activation Conformational Change as Assessed by Limited Proteolysis-Guanine nucleotides protect G protein ␣ subunits, including Go␣, from complete proteolytic degradation (30 -32). The pattern of fragments derived from partial tryptic digestion can be used as an indicator of the conformation of the protein.
In the presence of GDP, Go␣ is hydrolyzed by trypsin resulting in two products, a stable 25-kDa and an unstable 17-kDa peptide. Binding of non-hydrolyzable analogs of GTP can induce an active conformation of the Go␣ subunit, which is resistant to proteolytic degradation, and protects a stable 37-kDa polypeptide from further degradation. In the case of the activated mutant Go␣Q205L, GTP can also protect the remaining 37-kDa polypeptide from complete proteolytic digestion by trypsin because Go␣Q205L lacks GTPase activity. Fig. 3a shows that XTP protects Go␣X from proteolysis by trypsin (lanes 4 and 5), whereas in the control experiment, GTP␥S protected wild-type Go␣ (lane 8). This experiment indicates that Go␣X binds XTP without hydrolyzing it. After binding to XTP, Go␣X must have assumed a conformation similar to that of GTP␥S-bound wild-type Go␣. In this experiment, wild-type Go␣ needed only 1 M GTP␥S to prevent complete proteolysis. Similarly, Go␣X was sufficiently protected in the presence of 1 M XTP. It is noteworthy that GTP␥S, but not GTP, was also able to protect Go␣X from complete tryptic digestion although this protection required GTP␥S concentrations above 100 M (lanes 1, 2, and 3). Thus, Go␣X has a much lower affinity for GTP␥S than for XTP. We did not detect any of GTP␥S binding activity of Go␣X in our nucleotide binding assay because the highest concentrations of [ 35 S]GTP␥S used in the reaction were micromolar. Consistent with the results of the nucleotide binding experiments, the single mutant Go␣D273N was not protected by any nucleotides including GTP, GTP␥S, and XTP up to millimolar concentrations (data not shown).
Pertussis Toxin-induced ADP-ribosylation-The interaction of Go␣ with the ␤␥ complex can be assayed by ADP-ribosylation of the ␣ subunit induced by pertussis toxin (PTX) because ADP-ribosylation requires the formation of the heterotrimeric complex (33,34). Modification (by ADP-ribosylation) of recombinant Go␣ catalyzed by PTX is the same in the presence of GTP or GDP because of the GTPase activity of Go␣. However, GTP␥S strongly inhibits the modification since Go␣ cannot hydrolyze GTP␥S. GTP␥S binding thus promotes the dissociation of the trimeric ␣␤␥ complex and prevents the ADP-ribosylation of the Go␣ subunit. The activated Go␣Q205L mutant lacks GTPase activity, and the effect of GTP on ADP-ribosylation is similar to that of GTP␥S on the wild-type Go␣. Therefore, PTX labeling can be used not only to examine ␤␥ binding but also GTPase activity. Fig. 3b shows that purified Go␣ was ADP-ribosylated by pertussis toxin (lane 7), and the labeling was strongly inhibited by GTP␥S (lane 6). In contrast, Go␣X was modified by pertussis toxin only in the presence of XDP (lane 4) but not with GDP (lane 5), and as expected, the reaction was strongly inhibited by XTP (lane 2), whereas GTP had no effect (lane 3). Therefore, only XDP-bound Go␣X can form trimeric complexes with ␤␥, and binding of XTP induces dissociation of the trimeric complex. As a control, we did not detect any ADP-ribosylation of Go␣X when GTP␥S, GTP, or XTP alone was present (data not shown). Consistent with the results of trypsin digestion, this experiment indicated that XTP was not hydrolyzed by Go␣X. The quantitation of [ 32 P]ADPribose incorporation revealed that the labeling of Go␣X was proportional to the amount of ␤␥ used and reached a maximum at a Go␣X:␤␥ ratio of 1:1, similar to wild-type Go␣ (data not shown). Interestingly, high concentrations (over 100 M) of GTP␥S also inhibited the ADP-ribosylation of Go␣X (Fig. 3b,  lane 1), offering further evidence that Go␣X was able to bind GTP␥S with low affinity. As expected, Go␣D273N did not interact with ␤␥ and was not modified by pertussis toxin in the presence of either GDP or XDP (data not shown).
XDP-dependent ␤␥ Interaction in Transfected COS-7 Cells-In transfected COS-7 cells, ␤ 1 ␥ 2 is able to activate PLC␤ 2 , and the activation of PLC␤ 2 can be inhibited by cotransfection with Go␣ because of competition for ␤␥ (35). We cotransfected COS-7 cells with PLC␤2, ␤1, ␥2, and Go␣D273N or Go␣X and found that both Go␣ mutants did not inhibit PLC␤ 2 activity, whereas wild-type Go␣ did. This experiment indicates that both mutants do not bind ␤␥ in COS-7 cells and is consistent with the in vitro experiments on PTX-induced ADP-ribosylation. Go␣X bound ␤␥ only in the presence of XDP, and because XDP concentration is negligible inside the cell, the interaction did not occur. To deliver XDP into cells, we tried to permeabilize COS-7 cells by several methods including digitonin treatment, electroporation, and ␣-toxin (36). We found that only ␣-toxin gave us consistent results and had no effect on the PLC␤2 activities stimulated by ␤␥. After incubating cells with ␣-toxin in the presence of XDP, we found that Go␣X inhibited PLC␤2 activity, whereas Go␣D273N was not affected by XDP (Fig. 4). In the control experiments, we found that adding GDP or GTP to the permeabilization buffer had no effect on the PLC␤2 activity of cells transfected with the Go␣ mutants (data not shown). This experiment shows that the Go␣ mutants behave similarly in vitro and in cultured cells; Go␣X binds ␤␥ only when exogenous XDP is available. DISCUSSION We engineered a mutant of Go␣ that switched nucleotide binding activity from guanine nucleotides to xanthine nucleotides. The mutation (D273N) was at a conserved residue of the NKXD motif that appears in all GTPase superfamily proteins. Crystal structures of transducin and Gi showed that this aspartic acid residue participated in hydrogen bonding to the guanine ring (Fig. 1a). The proposed interaction between the mutagenized Asn and the xanthine ring is shown in Fig.  1b in which the hydrogen bond is "flipped" when compared with wild-type G␣. Similar single Asp3 Asn mutations have been made in other GTP binding proteins, including EF-Tu (17,18), Ypt1 (19), rab-5 (20,21), and FtsY (22), and E. coli adenylosuccinate synthetase (23), resulting in active proteins regulated by xanthine nucleotides instead of guanine nucleotides. However, the similar D119N mutant of H-Ras induced transformation of NIH-3T3 cells with efficiency indistinguishable from wild-type H-Ras (16,37). Although the mutant D119N Ras exhibited decreased affinity for GTP and increased affinity for XTP (by 2 to 3 orders of magnitude), the high intracellular concentration of GTP (millimolar) probably ensures that the protein is still bound to the guanine nucle-otides in the cell. Interestingly, we found the corresponding D273N mutation in Go␣ did not result in binding of either GTP␥S or XTP␥S, whereas the D273N/Q205L double mutant, Go␣X, switched nucleotide binding ability. When examining the crystal structure of transducin, it is not clear why the Gln3 Leu mutation (position 200 in transducin ␣), which is at the opposite side of the nucleotide binding pocket from the Asp3 Asn mutation (position 268 in transducin ␣), rescued the xanthine nucleotide binding of Go␣D273N. It is interesting to note that Go␣X binds GTP␥S at concentrations higher than 100 M. In our nucleotide binding experiments, we could not observe this binding because the affinity was weak, requiring concentrations higher than 1 M [ 35 S]GTP␥S, which was the highest concentration that we could use. The P-S bond of the ␥ phosphate in GTP␥S is longer than the P-O bond in GTP, which not only prevents nucleotide hydrolysis when binding to G protein ␣ subunits, it also results in qualitatively different interactions and different affinities.
In vitro experiments using limited trypsin digestion and PTX-induced ADP-ribosylation showed that Go␣X retained the characteristic properties of wild-type Go␣ in the presence of XDP or XTP. In addition, our data confirm the assumption that diphosphate nucleotides are required for the interaction of G protein ␣ subunits with ␤␥ subunits. XTP-bound Go␣X assumed a trypsin-resistant conformation similar to that of the activated wild-type Go␣ and stimulated ␤␥ dissociation from the trimeric complex, suggesting that Go␣X can be activated by XTP. In transfected COS-7 cells, PLC␤2 is activated by G protein ␤␥ subunits, and the activity is inhibited when cotransfecting with Go␣ because of the competition for ␤␥. To study ␤␥ binding of the mutant Go␣X in vivo, we looked for inhibition of PLC␤2 activity as an indication of ␤␥ binding. We found that Go␣X did not affect ␤␥-stimulated PLC␤2 activity because of the absence of XDP. To turn on ␤␥ binding, we used ␣-toxin to make cell membranes permeable to XDP, and indeed under these conditions, Go␣X attenuated PLC␤2 activity. G proteinderived ␤␥ subunits are shown to be able to bind many proteins other than G␣, and may be involved in many signal transduction pathways. We demonstrated that XDP can be delivered into cells and Go␣X may be used as a ␤␥ quencher that can make the cellular ␤␥ pool unavailable to other ␤␥ effectors. The ability to turn on and off ␤␥ in vivo could be useful to better understand the physiological function of ␤␥.
Go␣ is one of the G protein ␣ subunits whose functions are not well understood although there is some evidence supporting a role in the regulation of calcium channels (38 -42). Since ␤␥ subunits are also proposed as regulators of calcium channels (43), it is difficult to differentiate the activities of Go␣ and ␤␥ in some situations when activated receptors release both Go␣ and ␤␥ subunits. This is one of the problems that the Go␣X mutant might be used to address. The channel may be activated directly by adding XTP without releasing free ␤␥ in cells that have been transfected with cDNA expressing the mutant protein. Cross-talk between the different G protein-mediated signaling pathways has been well demonstrated. Activating Go␣X directly and instantly by XTP would avoid the interference of other pathways and help us to differentiate individual pathways. Introducing this mutation into other G protein ␣ subunits may be used to study their functions as well.