On the Significance of Toc-GTPase Homodimers*

Precursor protein translocation across the outer chloroplast membrane depends on the action of the Toc complex, containing GTPases as recognizing receptor components. The G domains of the GTPases are known to dimerize. In the dimeric conformation an arginine contacts the phosphate moieties of bound nucleotide in trans. Kinetic studies suggested that the arginine in itself does not act as an arginine finger of a reciprocal GTPase-activating protein (GAP). Here we investigate the specific function of the residue in two GTPase homologues. Arginine to alanine replacement variants have significantly reduced affinities for dimerization compared with wild-type GTPases. The amino acid exchange does not impact on the overall fold and nucleotide binding, as seen in the monomeric x-ray crystallographic structure of the Arabidopsis Toc33 arginine-alanine replacement variant at 2.0Å. We probed the catalytic center with the transition state analogue GDP/AlFx using NMR and analytical ultracentrifugation. AlFx binding depends on the arginine, suggesting the residue can play a role in catalysis despite the non-GAP nature of the homodimer. Two non-exclusive functional models are discussed: 1) the coGAP hypothesis, in which an additional factor activates the GTPase in homodimeric form; and 2) the switch hypothesis, in which a protein, presumably the large Toc159 GTPase, exchanges with one of the homodimeric subunits, leading to activation.

The vast majority of GTPases serve as molecular switches that regulate various signaling and transport processes within the cell. GTPases bind and hydrolyze GTP and the nucleotide is recognized by five loops of specific function, called G1-G5 loop (1). Typically, GTPases have only low intrinsic GTPase rates and rely on auxiliary proteins such as GTPase activating pro-teins (GAPs) 3 and guanosine nucleotide exchange factors (2). Regulation of hydrolytic activity can in various ways also be achieved through dimerization of the GTPase. For the different studied cases of dimeric GTPases (3)(4)(5)(6)(7)(8), differences exist with respect to interaction mode or the function of dimerization.
It is not entirely clear how dimerization and hydrolytic activity are linked (16,18,23,25). This is surprising as the dimerization interface not only involves a number of Tocspecific insertions (17,25) but also several G loops that bind the nucleotide. An arginine, contacting nucleotide in trans in dimeric GTPase complexes, has been assigned a function in dimer formation (15,23), in nucleotide recognition (16), and in catalysis (16,17,23). To decipher the specific role of this residue, we studied the GTPases psToc34 and atToc33 as well as arginine to alanine replacement variants psToc34 R133A and atToc33 R130A . We discuss the physiological role of Toc GTPases Toc33/34 homodimers, which are in abundance in the Toc complex (28 -30).
Recombinant proteins were purified using nickel affinity chromatography (GE Healthcare) in 50 mM Tris buffered at pH 7.4 containing 500 mM NaCl, 10 mM imidazole, 5 mM MgCl 2 , 10% glycerol, and 0.7 mM ␤-mercaptoethanol as running buffer; elution buffer additionally contained 500 mM imidazole. For crystallization the protein was further purified by gel filtration using a Superdex 75 prep grade 26/60 column (GE Healthcare) with 20 mM HEPES buffered at pH 7.4 containing 150 mM KCl, 3 mM MgCl 2 , and 0.7 mM ␤-mercaptoethanol as running buffer. For analytical ultracentrifugation, psToc34 and psToc34 R133A were further purified after nickel purification by size exclusion chromatography using a Superdex 75 26/60 column and 20 mM Tris buffer at pH 8.5 containing 100 mM NaCl and 3 mM MgCl 2 . For nucleotide exchange of psToc34, buffer exchange after nickel affinity purification with 20 mM Tris at pH 8.5, containing 100 mM NaCl and 3 mM MgCl 2 , was performed on a PD10 column (GE Healthcare). The protein was incubated with 2 mM GMPPNP and 100 units of alkaline phosphatase (New England Biolabs, Frankfurt, Germany) for 10 h at 15°C. A subsequent purification step by nickel affinity chromatography and size exclusion chromatography was performed to remove alkaline phosphatase.
Crystallography-atToc33 R133A was crystallized at a concentration of 0.8 mM using sitting drop vapor diffusion and a 2-l drop size at 19°C. Crystals were obtained within 3 days in 20% polyethylene glycol 3350 with 0.2 M NH 4 Cl, subsequently frozen in liquid nitrogen, and stored using mother liquor containing an additional 20% glycerol as cryoprotectant. Data were collected at European Synchrotron Radiation Facility, Grenoble, France, on beamline ID14-4 at a wavelength of 0.933 Å on an ADSC Quantum-q4 CCD imaging device.
Data were integrated and scaled with the HKL software (31). Data reduction, Free-R assignment, and all further data manipulation were carried out with the CCP4 suite of programs (32). The structure was determined by molecular replacement using the program MOLREP (33) with atToc33 as a search model (PDB code 3BB3 (25)). Iterative model building and refinement were carried out with the programs "coot" (34) and REFMAC5 (35) cycled with ARP (36) Data collection and refinement statistics are summarized in Table 1.
NMR Spectroscopy-19 F NMR spectra were measured on a DRX300 spectrometer (Bruker, Rheinstetten, Germany) operating at 270 MHz. The spectra were acquired at 25°C using protein at a concentration of 0.5 mM in 20 mM Tris buffered at pH 7.0 containing 75 mM NaCl, 3 mM MgCl 2 , 10 mM NaF, 1 mM AlCl 3 , and 10% D 2 O added before acquisition. The spectra were referenced to external trifluoracetate. A 90°pulse was used with a repetition rate of 2 s. 4096 free induction decays were summed up. The spectra were processed with TOPSPIN (Bruker, Germany).
Biochemical and Biophysical Assays-For analytical ultracentrifugation, nucleotide load of the protein sample was controlled by reverse phase HPLC as described (25). A preparation of GDP-loaded GTPase was split, and to one half of the preparation 10 mM NaF and 1 mM AlCl 3 were added. Both samples were incubated overnight at 4°C. The final protein concentration for analytical ultracentrifugation on a Beckman Optima XL-A ultracentrifuge equipped with absorbance optics and an An60 Ti rotor (Beckman Coulter, Fullerton, CA) was adjusted to 45 M. Sedimentation velocity runs were carried out at 20°C at 40,000 rpm using the size exclusion buffer as described above as reference. Buffer density (1.00314 ml g Ϫ1 ), buffer viscosity (1.002 millipascals⅐s) as well as the partial specific volume of psToc34 based on amino acid sequence (v ϭ 0.7410 ml g Ϫ1 ) were calculated using the program SEDNTERP, Version 1.05. 4 Sedimentation coefficients were determined from the c(s) distribution using the program SEDFIT (37,38), normalized for water and 20°C.
Dimerization behavior of nickel affinity purified atToc33 and atToc33 R130A was analyzed by size exclusion chromatography using a Superdex75 16/100 (GE Healthcare) equilibrated with 20 mM Tris buffered at pH 8.5 containing 75 mM NaCl and 3 mM MgCl 2 . For molecular weight determination, 100 l of nickel affinity-purified protein at a concentration of ϳ0.6 mM was loaded onto a Superdex75 HR 10/300 gel filtration column (GE Healthcare) equilibrated with 20 mM Tris buffered at pH 7 containing 75 mM NaCl and 3 mM MgCl 2 . Alternatively, the buffer contained 10 mM NaF and 1 mM AlCl 3 . For in-line detection, a Mini Dawn light scattering instrument (Wyatt Technology, Santa Barbara, CA) and a refractory index detector (WGE Dr. Bures, Dallgow, Germany) were used. Data were evaluated using the AstraV software (Wyatt Technology).
GTPase activities of atToc33, atToc33 R130A , psToc34, and psToc34 R133A were determined by a HPLC-based hydrolysis assay as described previously (25). Protein was used at a concentration of ϳ0.

Homodimerization and Nucleotide Recognition of Small
Toc-GTPases-To understand the function of dimerization of the two homologous GTPases atToc33 and psToc34, we first analyzed the previously published three-dimensional structures of psToc34 (17,25). Five G loops bind the nucleotide, as in other GTPases (1), but in addition, the G loops G2, G3, and G4 are involved in dimerization (gray, blue, and green in Fig. 1A). Furthermore, the conserved box loop (CB loop) contributes to dimerization ((25) red in Fig. 1A). This sequence motif is unique to Toc and the so called Aig-like GTPases (39). Also at the dimer interface is helix ␣5 (green in Fig. 1A).
We investigated how nucleotide recognition and dimerization are linked, as the G loops are located at the interface. We note a change in function of the G4 loop which has lost properties of nucleotide recognition to gain properties in dimerization. Small Toc GTPases have a conserved histidine in the G4 loop (sequence motif THAQ) not present in the G4 loop of canonical small GTPases like the GTPase Ras p21 (sequence motif NKXD (17,40,41); Fig. 1B). In psToc34, psHis-163 of the G4 loop is in hydrogen-bonding distance to psTyr-132 in the CB loop of the homodimerization partner in trans (Fig. 1C). psHis-163 further makes a -stacking interaction with the guanine ring of the base. The Toc34 G4 loop is deprived of a central aspartate residue that in small GTPases normally would specify the nucleotide through direct interaction with N1 and the 2-amino group of the guanine base ( Fig. 1, B-D). Consequently, in psToc34 other interactions are responsible for nucleotide recognition; the G5, glutamate contacts guanosine N1 directly and the guanosine 2-amino group via one bridging water (WAT, Fig. 1C). The lack of specificity explains the reported hydrolysis of Xanthosine-5Ј-triphosphate (42,43).
The Structural Impact of R130A Exchange on Nucleotide Recognition-The CB loop not only provides psTyr-132 but, more importantly, the two adjacent arginines, psArg-128 and psArg-133, for dimerization. Replacement of any of these arginines with alanine leads to abrogation of dimerization (15)(16)(17)23). A structural study carried out at 3.2 Å resolution on atToc33 R130A reported a monomeric structure of the GTPase (PDB code 2J3E (16)). Interestingly, the guanosine moiety was fitted in an unusual conformation in which the guanidine group was turned by 150° (16). Given the analysis presented above on incomplete nucleotide recognition by the G4/G5 loops, one might suppose that this unusual nucleotide conformation could occur, even though it would be in disagreement with other GTPase structures.
The earlier data prompted us to initiate a crystallographic study with the aim to collect atomic resolution data on atToc33 R130A . The expression construct encoded amino acids 1-251 rather than amino acids 1-256 as in the previous study (16). Furthermore, crystallization conditions were slightly different; we used 20% polyethylene glycol 3350, 0.2 M NH 4 Cl, pH 7.4, instead of 30% polyethylene glycol 4000, 0.2 M ammonium acetate, 0.05 M sodium acetate, pH 4.8 with added NADH (16). The crystals diffracted to a Bragg spacing below 2 Å using synchrotron light (Table 1). We determined the structure by molecular replacement using the model of native atToc33 (PDB code 3BB3 (25)). Space group symmetry and molecular packing are identical to the earlier study on atToc33 R130A (16) but different from wild-type atToc33. At position 130, no side chain electron density is seen, consistent with the argininealanine exchange. All three structures are observed in the GDPbound form and are highly similar (with respective root mean square deviation values of 0.58 Å over 231 C ␣ positions and of 1.1 Å over 233 C ␣ positions for comparison of atToc33 R130A with native atToc33 (PDB code 3BB3 (25) and with the earlier structure of atToc33 R130A (PDB code 2J3E (16)). Importantly, structural changes reported to occur in atToc33 R130A (16) are not confirmed.
Initial refinement without nucleotide resulted in clear negative F obs Ϫ F calc density for the nucleotide, as shown in Fig. 2A. Compared with the earlier report (16), the syn conformation of the glycosidic bond in the GDP molecule is not confirmed ( Fig.  2A, white nucleotide). Instead, the common anti-conformation is observed ( Fig. 2A, black nucleotide). Thus, although it was previously suggested that a change in nucleotide conformation might have occurred either by lack of dimerization or as a result of the atR130A exchange (16), comparison with the wild-type structure and the high resolution structure of atToc33 R130A presented here rule out this possibility, demonstrating that nucleotide binding is unaffected.
Effect of Arginine to Alanine Exchange on the Homodimerization of Toc34-In the structure of dimeric psToc34, an arginine (Arg-133, the equivalent of atArg-130) contacts the ␤and ␥-phosphate groups in trans. This interaction is suggestive of a function as arginine finger often found in GAP-GTPase interactions ( Fig. 1C) (44). This has led to the proposal that the psToc34 homodimer could be a self-activating GAP complex (16,17). Furthermore, this interaction has been described to be pivotal for dimerization (15,16,23).
We established the dimerization properties of atToc33 and of atToc33 R130A using size exclusion chromatography (Fig. 2B). As the chromatogram shows, both proteins dimerize. However, the dimerization behavior of atToc33 R130A is impaired. The difference in migration behavior of both, monomeric and dimeric species, can be explained by a presumed difference in dimerization behavior. A fast dimerization equilibrium in atToc33 would lead to a decreased apparent size of the dimer; in turn, it would increase the apparent size of the monomer. This is supported by static light scattering data, given below. Although dimeric and monomeric species do not base-line separate for atToc33, atToc33 R130A is different and shows baseline separation.
Effect of Arginine to Alanine Exchange on the Hydrolysis Rate of Toc34-Previously, it was shown in multiple turnover assays that arginine-alanine replacement impacts on GTP hydrolysis. atToc33 R130A shows a minor reduction in hydrolysis rate compared with atToc33 (15,16), and psToc34 R133A shows a loss of hydrolytic activity (23).
We investigated the effect of an arginine to alanine replacement on GTP hydrolysis using an HPLC-based single turnover assay for determination of enzymatic rates (45). The assay was carried out at higher concentrations than the multiple turnover experiments, allowing for dimer formation of wild-type proteins (see "Experimental Procedures" for details). atToc33 R130A showed 0.6-fold hydrolytic activity compared with wild-type protein (k cat ϭ (2.3 Ϯ 0.4) ϫ 10 Ϫ5 s Ϫ1 for atToc33 R130A and k cat ϭ (4.4 Ϯ 0.7) ϫ 10 Ϫ5 ) ϫ 10 Ϫ5 s Ϫ1 for wild-type atToc33).
Similarly, psToc34 R133A showed 0.3-fold hydrolytic activity of the wild-type GTPase (k cat ϭ (1.9 Ϯ 0.7) ϫ 10 Ϫ5 s Ϫ1 for psToc34 R133A and k cat ϭ (8.4 Ϯ 0.7) ϫ 10 Ϫ5 s Ϫ1 for wild-type  . C, the G4 and G5 loops of psToc34 interact with the nucleotide. The 2-amido group of the nucleotide is only in indirect contact with the protein via a water molecule (WAT). Thus, guanosine and xanthosine nucleotides cannot be distinguished. psHis-163 makes a hydrogen-bonding contact with psTyr132Ј of the CB loop of the interacting homodimerization partner. Also shown is psArg-133Ј, interacting with phosphate moieties of the dimerization partner. D, a similar representation as in C for the Ras p21 protein (GMMPNP-bound state, PDB code 5p21, (54)). The conserved Asp-119 in the G4 loop recognizes GTP specifically by interacting with N1 and the 2-amino group. protein). On one hand this establishes that psToc34 R133A possesses hydrolytic activity; on the other hand it demonstrates that the arginine-alanine exchange is only of minor influence on GTP hydrolysis in either GTPase. Apparent differences in determined hydrolysis rates with earlier reports (23) are likely explained by the different experimental setup of single and multiple turnover measurements.
Binding of AlF x to psToc34-Because exchange of atArg-130/ psArg-133 with alanine has only limited influence on the GTP hydrolysis rate, we tested whether the respective arginines can act at all as arginine fingers, employing aluminum fluoride as a probe. Aluminum fluoride exists in an equilibrium of different species in solution and is, thus, abbreviated here as AlF x . AlF x can act as a transition state mimicry of phosphoryl transfer reactions (46) and has been shown to directly bind to the G␣ proteins (47) that contain an intrinsic domain for stimulation of catalysis. Intrinsic stimulatory domains are absent in small GTPases such as Ras or Toc34, and they instead require a GAP for activation. For example AIF x binding to Ras GDP depends on the presence of the RasGAP proteins (48). GTPase⅐GAP⅐AlF x complexes show AlF x binding in the active site in place of the ␥-phosphate and require GDP-loaded GTPase subunits. The GAP arginine finger is often present as binding partner. If psArg-133 acts as an arginine finger in the psToc34 dimer, AlF x binding to psToc34 can be expected. AlF x binding to psToc34 was tested using 19 F NMR. A buffer solution containing AlCl 3 and NaF shows peaks at Ϫ77.0 ppm and at Ϫ41.7 ppm, corresponding to AlF x and free F Ϫ (Fig. 3A). After the addition of psToc34 GDP , a peak shifted by Ϫ24.5 ppm from the resonance signal of free F Ϫ is observed at Ϫ66.2 ppm (Fig. 3D). The chemical shift, varying between Ϫ20 ppm and Ϫ22.4 ppm in previous studies (47,49,50), is indicative of AlF x binding to nucleotide binding proteins and has been described before for psToc34 (17). To verify the specificity of the interaction in the catalytic center, psToc34 loaded with non-hydrolyzable GTP analogue GMPPNP was used. Because the binding site is occupied by the ␥-phosphate of GMPPNP, specific binding of AlF x to the ␥-phosphate site can be excluded. Indeed, no binding of AlF x is detected when the GMPPNP-loaded GTPase is investigated (Fig. 3E).
Effect of AlF x on psToc34 Dimerization-We next tested the stability of the psToc34⅐GDP⅐AlF x complex. Size exclusion chromatography of psToc34 GDP in the presence of AlF x in the buffer indicated stabilization of the dimer, as evidenced by a shift to a higher molecular weight species (data not shown). To quantify AlF x -induced oligomerization, analytical ultracentrifugation was employed. Two samples were compared, distinguished by the presence of AlF x . Without AlF x treatment, psToc34 is present in monomer-dimer equilibrium (Fig. 4A (23,  25)). With AlF x treatment, psToc34 was exclusively dimeric (Fig. 4A). Thus, the addition of AlF x leads to stabilization of the psToc34 homodimer as reported for classical GTPase-GAP interactions like Ras-RasGAP (48).
When analytical ultracentrifugation was repeated with psToc34 R133A , no dimeric species was observed regardless of AlF x treatment (Fig. 4B). This demonstrates that the effects seen before with wild-type psToc34 are specific and require the presence of psArg-133, in line with the NMR data (Fig. 3).
Influence of AlF x on the Dimerization of atToc33-We then assayed the effect of AlF x binding on the dimerization behavior of atToc33. Because the protein exhibits a lower K a for dimerization, analytical ultracentrifugation is impractical due to the high protein concentrations that would be required. Instead, we employed a setup where size exclusion chromatography was coupled with static light scattering and a refractive index detector to determine absolute molecular weights. This method does not require use of internal standards (51).
Similar to previous reports (15)(16)(17)23), two molecular species were observed for wild-type atToc33 in the absence of AlF x (Fig. 5A). The analysis of static light scattering gave a signal yielding a molecular mass of 63 kDa for protein fractions in the first peak, which fits well with the value of 60 kDa for an atToc33 dimer. However, protein fractions of the second peak displayed a molecular mass of 45 kDa; the tail of this second peak was fitted with a molecular mass of 36 kDa, likely to represent the monomer. The 45-kDa species likely results from a dynamic equilibrium between dimeric and monomeric species. Thus, atToc33 exists as a fast equilibrium between the two states (compare Fig. 2B). This is consistent with data on psToc34 (23). When atToc33 R130A was investigated, a single peak fitted to a molecular mass of 30 -31 kDa was obtained, corresponding to monomeric protein (Fig. 5B).
When AlF x was present in the buffer (Fig. 5C), atToc33 shifted to a higher molecular mass species. Light scattering data were fitted to a molecular mass of 66 kDa, corresponding to the molecular mass of dimeric atToc33 (60 kDa). Thus, stabilization of the dimeric species occurs with atToc33 upon the addition of AlF x . However the addition of AlF x does not affect atToc33 R130A , which remains monomeric in the presence of AlF x (Fig. 5D). The fit of the light scattering data yielded a molecular mass of 25 kDa.
The dimerization behavior of atToc33 (Fig. 5A) and of psToc34 (Fig. 4A) are similar. The dimeric state of both proteins is stabilized by AlF x (Figs. 4A and 5, A and C). With the arginine-alanine exchange mutants, it can be shown that binding is specific, since stabilization of the dimer requires the presence of atArg-130/psArg-133 (Figs. 4B and 5, B and D).

DISCUSSION
Dimerization of Toc GTPases is generally assumed to be a feature of the assembly of the Toc apparatus (24). GTPase dimerization is recurrent, and the Toc GTPases are not exceptional in this respect. Documented examples of dimeric GTPases are, for instance, the SRP GTPases FtsY, Ffh (4,5), and FlhF (3), the GTPases of the dynamin type, e.g. hGBP (7), the  19 F NMR spectrum of buffer containing AlCl 3 and NaF shows two peaks (Ϫ77.0 ppm and Ϫ41.7 ppm) that have been assigned to free F Ϫ and to AlF x . B and C, the addition of nucleotides (GDP, GMPPNP) shows no changes in the spectrum compared with A. D upon addition of psToc34 GDP to buffer containing AlCl 3 and NaF, an additional peak is observed at Ϫ66.2 ppm. E upon addition of psToc34 GMPPNP to buffer containing AlCl 3 and NaF, no changes are observed in the spectrum compared with A. GTPase MnmE involved in tRNA modification (8), and the metal binding GTPase HypB (6). However, the dimerization interface is different between these GTPases, and so is the functional significance of dimerization of these various GTPases.
The isolated G domains of psToc34 and atToc33 both dimerize, but they differ with respect to their dimerization properties (16,25). The K d of the atToc33 dimer is about 1 order of magnitude higher than that for psToc34. Both K d values are in the submillimolar range (25,27). Although these figures, determined for the isolated GTPase, seem high, dimerization may still occur in the physiological context on the membrane or within the Toc complex via elevated local concentrations. Interaction may be helped by the C-terminal membrane anchor, not present in the proteins analyzed here.
The dimerization interface itself is preserved between different Toc34 GTPases (16) and involves the CB motif (25) as well as the G4/G5 loops, with G4 performing a dual role in nucleotide recognition and dimerization (Fig. 1C). The CB motif carries the arginine that contacts the nucleotide in trans, the function of which is controversially discussed with respect to dimerization, nucleotide binding, and catalysis (15-17, 23, 25). The function of this arginine, thus, requires further clarification, as it is the key to elucidate the task of the Toc34 homodimer.
The role of psArg-133/atArg-130 in dimerization was previously investigated using a variety of techniques, including native PAGE, analytical ultracentrifugation, and size exclusion (15,16,23). It was shown that atToc33 R130A and psToc34 R133A are unable to dimerize. In contrast, we show that atToc33 R130A forms dimers at high protein concentrations using size exclusion chromatography (Fig. 2B). This suggests that Arg-130 in atToc33 is a key, but not the sole player in homodimerization.
A function of psArg-133/atArg-130 in nucleotide recognition was suggested on the basis of the previous 3.2 Å structure of atToc33 R130A (16) in a monomeric state with an unusual nucleotide conformation. However, the limited resolution and the lack of a wild-type reference GTPase structure did not allow concluding whether the amino acid exchange directly affected the structure or whether the effect was indirect and caused by the lack of dimerization. Based on the 2-Å resolution structure presented here, an altered nucleotide conformation can be excluded. This is also evident from superposition with the now available monomeric wild-type atToc33 structure (25).
Finally, participation of psArg-133 in catalysis was previously proposed on the basis of the crystal structure of psToc34 that showed this residue in a conformation similar to the classic GAP arginine finger (17). The transition state mimic GDP/AlF x binds to the psToc34 homodimer but not to the psArg-133 mutant or to the GMP-PNP-loaded GTPase (Fig. 3). This implies that AlF x indeed acts as a transition state mimicry, demonstrating the arginine is in an appropriate conformation to act during catalysis. AlF x binding also stabilizes the homodimers in psToc34 (Fig. 4A) and in atToc33 (Fig. 5C). AlF x -induced dimerization suggests a composite binding site formed by both dimerization partners involving psArg-133/atArg-130.
Summing up from our studies on psArg-133/atArg-130 and the literature, the following evidences doubt or directly contradict that small Toc GTPases form self-associating GAP-complexes (17). (i) Although it would be expected that a GAP complex favors the GTP state, it is observed that GMPPNP and GDP-loaded states of the GTPase both dimerize with similar efficiency (17,25). (ii) Despite the stabilization of G loops in the dimerization interface, switch I retains some flexibility, which would be unexpected for a GAP complex (25). (iii) Despite the presence of the stabilizing arginine, reminiscent of the classic GAP arginine finger, the catalytic center is incomplete, as no residue for the positioning of the catalytic water is present (25); instead the catalytic site is accessible for solvent by a short tunnel. (iv) Kinetic data argue against the formation of a GAP-like complex; the acceleration of hydrolysis in GTPase⅐GAP complexes is typically in the order of 2-5 magnitudes (44). However, no significant catalytic activation is observed by dimer formation (23,25). This is supported by data from argininealanine exchange proteins which show that replacement of psArg-133/atArg-130 has only a minor effect on the hydrolysis rates of psToc34/atToc33.
Thus, small Toc GTPases represent a paradox since psArg-133/atArg-130 seems poised properly to act as an arginine finger, similar to that of a GAP, but hydrolysis rates are not accelerated. Acceleration of hydrolysis in Ras-like GTPases and their respective GAPs is mainly due to the positioning of a catalytic residue to polarize a water for hydrolytic attack (52). For instance, replacement of this crucial residue in a Ras:RasGAP system leads to abrogation of hydrolysis even when an arginine finger interaction is present (53). Our structural analysis of the Toc34 homodimer shows that despite the presence of psArg-133 or atArg-130 the catalytic machinery remains incomplete, and an essential catalytic residue for the positioning of the catalytic water is required, explaining the minor effect of dimerization on hydrolysis rates (25).
We have previously shown that the nucleotide binding pocket is accessible in the psToc34 GMPPNP homodimer and suggested a catalytic residue could be inserted into the catalytic center. This would functionally define the homodimer as a coGAP complex that requires both the homodimeric interaction and a third protein (coGAP hypothesis) (25). In addition, the third protein may also be required to stabilize and organize the catalytic center, then giving it its true GTPase/GAP type character. psArg-133/atArg-130, thus, would fulfill the role of an arginine finger in catalysis only if the coGAP is present.
In a second proposal, the homodimer has to dissociate to become functional in the physiological context (switch hypothesis). Catalytic data suggest that the monomeric species requires interaction with another protein for activation. The large GTPase Toc159 is an obvious candidate for this interaction. Toc159 can supply an arginine, similar to what is seen in the homodimer but probably with slightly different geometry (25). The heterodimer is, thus, not only asymmetric but also different from the homodimer.
We conclude there may well be a physiological role for the small Toc33/34 GTPase homodimer, complementing the postulated heterodimer of small and large Toc GTPases. This is further supported by the stoichiometry of the Toc complex, where small GTPase subunits are in molar excess over large GTPase subunits. Hence, two differentially regulated events in the Toc-mediated chloroplast protein import cycle would exist.