Conformational states of the small G protein Arf-1 in complex with the guanine nucleotide exchange factor ARNO-Sec7.

Arf1 is a small G protein involved in vesicular trafficking, and although it is only distantly related to Ras, it adopts a similar three-dimensional structure. In the present work, we study Arf1 bound to GDP and GTP and its interactions with one of its guanosine nucleotide exchange factors, ARNO-Sec7. The (31)P NMR spectra of Arf1.GDP.Mg(2+) and Arf1.GTP.Mg(2+) share the general features typical for all small G proteins studied so far. Especially, the beta-phosphate resonances of the bound nucleotide are shifted strongly downfield compared with the resonance positions of the free magnesium complexes of GDP and GTP. However, no evidence for an equilibrium between two conformational states of Arf1.GDP.Mg(2+) or Arf1.GTP.Mg(2+) could be observed as it was described earlier for Ras and Ran. Glu(156) of ARNO-Sec7 has been suggested to play as "glutamic acid finger" an important role in the nucleotide exchange mechanism. In the millimolar concentration range used in the NMR experiments, wild type ARNO-Sec7 and ARNO-Sec7(E156D) do weakly interact with Arf1.GDP.Mg(2+) but do not form a strong complex with magnesium-free Arf1.GDP. Only wild type ARNO-Sec7 competes weakly with GDP on Arf1.GDP.Mg(2+) and leads to a release of GDP when added to the solution. The catalytically inactive mutants ARNO-Sec7(E156A) and ARNO-Sec7(E156K) induce a release of magnesium from Arf1.GDP.Mg(2+) but do not promote GDP release. In addition, ARNO-Sec7 does not interact or only very weakly interacts with the GTP-bound form of Arf1, opposite to the observation made earlier for Ran, where the nucleotide exchange factor RCC1 forms a complex with Ran.GTP.Mg(2+) and is able to displace the bound GTP.

Similar to Ras and other small G proteins, Arf1 switches between the GTP-bound "on" state, where it transmits signals to its effectors in the signaling pathway, and the GDP-bound "off" state. A characteristic feature of the nucleoside triphosphate form is the coordination of the ␥-phosphate group of GTP with the Mg 2ϩ ion and the amide group of Thr 48 from switch I in Arf1 (7). Simultaneously, the hydroxyl group of Thr 48 is coordinated to the metal ion. This bond pattern seems to be essential for stabilizing the correct conformation of the effector loop for effector recognition. The molecular details of the GDP-GTP switching mechanism are largely conserved among all small G proteins as well as the general features of the activation-deactivation cycle. GDP is replaced by GTP, which is present in higher concentration in the cytoplasm, by guanine nucleotide exchange factors (GEFs). 1 The specific GEFs for Arf1, which can be subdivided into four different subfamilies containing the Gea/Gnom/GBF family, the Sec7/BIG family, the ARNO/cytohesin/GRP family, and the EFA6 family (8), share a common domain (Sec7) of about 200 residues, which contains the catalytic activity. The Sec7 domain is also active on a truncated form of Arf1 that lacks the first N-terminal helix of 17 residues ([⌬17]Arf1). It demonstrates that the Sec7 domain of ARNO (ARNO-Sec7) interacts with the core domain of Arf1 and not with the myristate or the N-terminal amphipathic helix (9). Recently, a glutamic finger hypothesis in the catalysis of GDP-GTP exchange by the guanine nucleotide exchange factor ARNO-Sec7 has been formulated (10). Indeed, a specific glutamate residue, Glu 156 , located in the hydrophilic loop of ARNO-Sec7 helps to destabilize the GDP bound to Arf1 by displacing the Mg 2ϩ and by repulsing the GDP ␤-phosphate. The conservative mutation E156D as well as the charge reversal mutation E156K reduce the exchange activity of ARNO-Sec7 on Arf1 by several orders of magnitude. In addition, ARNO-Sec7(E156K) forms a complex with the Mg 2ϩ -free form of [⌬17]Arf1⅐GDP without inducing the release of GDP. The published crystal structure of a complex between nucleotide-free [⌬17]Arf1 and the Sec7 domain of Gea2 supports the hypothesis of a glutamic acid finger (7). A shift in the position of the switch 1 region in Arf1 exposes its active site to Glu 156 of the Sec7 domain. This results in steric and electrostatic repulsion of the ␤-phosphate and the Mg 2ϩ ion on Arf1, promoting nucleotide dissociation (7).
The molecular pathway of Arf that leads to GDP dissociation most probably involves several steps: docking, conformational change at the switch/Sec7 interface, and Mg/GDP release. These steps can be inferred from biochemical studies (e.g. brefeldin A (11), membrane lipids (9)) but have not been trapped by crystallography. Phosphorus NMR was shown to be an alternative to characterize different steps of GTP-induced conformations in effector recognition (15)(16)(17) as well as of GDP dissociation (16) of small G proteins. At a millimolar concentration of protein and nucleotide used, one can detect weak ternary complexes between a small G protein, nucleotide, and exchange factor. These complexes represent critical intermediates in the nucleotide exchange pathways but can hardly be isolated, since they tend to disappear to the advantage of more stable binary complexes during purification. Thus, if one considers a simple scheme for the exchange reaction: Arf⅐GDP ϩ Sec7^ArfGDP-Sec7^Arf-Sec7 ϩ GDP, phosphorus NMR is an excellent method to be able to distinguish in principle the phosphorus ␣ and ␤ signals at each step and notably when the nucleotide encounters the catalytic machinery of the exchange factor, here the Glu finger, before being expelled.
We set out to study the interaction of ARNO-Sec7 with [⌬17]Arf1 by phosphorus NMR using the wild type ARNO-Sec7 protein as well as the mutated forms E156D, E156K, and E156A. In all cases, the data suggest that stable ternary complexes of Arf1⅐GDP⅐Sec7 exist in solution in both the Mg 2ϩ -free and the Mg 2ϩ -bound form, depending on the ion concentration.

Protein Expression and Purification-[⌬17]
Arf1, a truncated form of Arf1, lacking the first 17 N-terminal residues, and ARNO-Sec7 (residues 50 -252 of ARNO) were expressed in Escherichia coli and purified as described (5,10,12). After purification, [⌬17]Arf1 is complexed at 60 -70% to GDP and at 30 -40% to GTP as determined by tryptophan fluorescence. To shift the equilibrium toward the GDP-bound form or the GTP-bound form, the protein was incubated with an excess of GDP or GTP, respectively, in the presence of 2 mM EDTA during 15 min at 298 K with stirring. Four mM of MgCl 2 was added, and the sample was loaded on a desalting column, NAP-5, and eluted in 20 mM Tris, pH 7.5, 1 mM MgCl 2 , 1 mM dithiothreitol in order to eliminate the free nucleotide.
NMR Spectroscopy-31 P NMR experiments were performed on a Bruker DRX-500 NMR spectrometer working at a phosphorus resonance frequency of 202 MHz. The 31 P NMR spectra were indirectly referenced to sodium 2,2-dimethyl-2-silapentane-5-sulfonate according to Ref. 13. A ⌶-value of 0.4048073561 was used, which corresponds to 85% external phosphoric acid contained in a spherical bulb. Unless noted otherwise, phosphorus spectra were recorded at a temperature of 278 K with a total spectral width of 40 ppm. For one-dimensional 31 P NMR spectra, 1024 -12,288 free induction decays were accumulated after excitation with a 65°(90°) pulse using a repetition time of 3-6 s. A total of 32,768 time domain data points were recorded and transformed to 32,768 real data points, corresponding to a digital resolution of 0.25 Hz/point. All data were processed on a Silicon graphics O2 work station using the software package UXNMR (Bruker, Karlsruhe, Germany) for data processing. Phosphorus NMR spectra used for peak integration and fitting were filtered by an exponential window function causing no significant line broadening.
Data Analysis-Quantitative analysis of the NMR spectra was carried out with a fitting and peak integration method, using the program Origin 6.0 from Microcal Software (Northampton, MA). First, all spectra were equally processed and scaled up or down depending on the number of scans, and a base line was determined by fitting the peak free areas with a constant. From the nonlinear fit of the more simple spectra with Lorentzian lines, the chemical shifts of the Arf1⅐GDP and Arf1⅐GTP phosphate groups were determined as accurately as possible under different free Mg 2ϩ ion concentrations (see Table I). These values were then used as constants in the analysis of the more complicated spectra (complexes with ARNO-Sec7), where a severe overlap of lines occurred. When fitting these kinds of spectra with a multi-Lorentzian model, we tried to reproduce the measured spectrum as closely as possible and making sure at the same time that consistent line widths were used for the single components. As an example, with a growing fraction of [⌬17]Arf1⅐GDP bound to ARNO-Sec7, the line width increases because of a higher total correlation time of the larger complex. This approach led to satisfying results in most cases. The biggest errors resulting from this method were caused by the slight deviation of the measured signals from a strict Lorentzian line shape (in the case of separated signals, a simple integration was carried out as well in order to verify the results and to estimate the error) and especially by the bad signal to noise ratio, a problem that was not avoidable even with data acquisition times of several hours. Fig. 1, a  and  Usually proteins with a single high affinity site for guanine nucleotides show only one resonance line for each phosphorus atom of a phosphate group if they occur in only one conformational state (Table I). Because only one set of resonance lines is observable for [⌬17]Arf1⅐GTP and [⌬17]Arf1⅐GDP, we concluded that Arf1⅐GDP⅐Mg 2ϩ and Arf1⅐GTP⅐Mg 2ϩ exist predominantly in a single (NMR-distinguishable) conformational state, in contrast to other small G proteins such as Ran or Ras. This is true for the whole temperature range from 277 to 303 K studied, where no indication for a second conformational state could be obtained from NMR spectroscopy.

P NMR Spectra of the Arf1⅐Nucleotide Complexes-
Sec7 domains are believed to expel the bound GDP from Arf1 in part by destabilizing the bound Mg 2ϩ . Therefore, it was important to record the phosphorus NMR spectrum of isolated [⌬17]Arf1⅐GDP in the absence of Mg 2ϩ . Fig. 1c shows the phosphorus NMR spectrum of [⌬17]Arf1⅐GDP after complexation of Mg 2ϩ with 2 mM EDTA (leaving ϳ1 M free Mg 2ϩ in solution). Two new resonances appear at Ϫ10.56 and Ϫ4.54 ppm corresponding to the ␣and ␤-phosphate group of bound GDP in Mg 2ϩ -free [⌬17]Arf1⅐GDP (Fig. 1c). The removal of Mg 2ϩ is not perfect, since the resonance lines assigned previously to [⌬17]Arf1⅐GDP⅐Mg 2ϩ are weakened but still visible. Note that the signal of the ␤-phosphate resonance of [⌬17]Arf1⅐GTP⅐ Mg 2ϩ (which is present as a contaminating species here) is not influenced much by the addition of EDTA, suggesting that the magnesium ion is not removed from the complex as to be expected from the higher affinity of Mg 2ϩ for Arf1⅐GTP than for Arf1⅐GDP. 31 P NMR Spectra of Wild Type ARNO-Sec7 Complexed with Arf1⅐GDP⅐Mg 2ϩ or Arf1⅐GDP- Fig. 2 shows the 31 P-NMR spectra of wild type ARNO-Sec7 titrated with increasing concentrations of Arf1⅐GDP⅐Mg 2ϩ . As was to be expected, ARNO-Sec7(wt) was not detected by 31 P NMR, and the 31 P signals of the ␣and ␤-phosphate grew in proportion to the increasing fraction of Arf1⅐GDP⅐Mg 2ϩ added. No larger shifts of the resonance lines of bound GDP are observed in the presence of ARNO-Sec7 (Table I), which would allow the unequivocal as-signment of the resonances of the putative Arf1⅐GDP⅐Mg 2ϩ -ARNO-Sec7 complex. With the molecular mass of 22,413 Da of ARNO-Sec7, the T 2 relaxation rate of bound GDP should increase in the rigid body approximation by a factor of 2.1, which represents a rather small difference in line width for distinguishing between free Arf1⅐GDP⅐Mg 2ϩ and Arf1⅐GDP⅐Mg 2ϩ complexed with ARNO-Sec7. The total line width at half height of the ␣-phosphate and the ␤-phosphate lines increased from 60 to 86 Hz, respectively, when the Arf1 to ARNO-Sec7 concentration was 0.84 (Fig. 2d). This means that there is an interaction between the two proteins under our conditions. The detailed analysis of the spectra reveals that at least the ␤-phos- At a ratio of 0.84 and 1.7 of Arf1⅐GDP⅐Mg 2ϩ to ARNO-Sec7, the signal of the ␤-phosphate resonance of a small amount of free GDP (partly complexed with Mg 2ϩ ) becomes visible at Ϫ5.78 ppm. In addition, the concentration of free inorganic phosphate increases in the sample. This is only seen in the case of the wild-type ARNO-Sec7 and with none of the mutants. A number of explanations seem to be reasonable to account for the free P i . 1) A contaminating phosphatase is present in the sample (Arf or Sec7 domain), which would hydrolyze GDP as soon as it dissociates from Arf1 upon the action of the nucleotide exchange factor. This would in fact imply the increase of GMP as a separate phosphorus line in our NMR spectra that cannot be seen. 2) Arf1⅐GDP⅐Mg 2ϩ exists partly as Arf1⅐GDP⅐P i Mg 2ϩ in solution, and only the catalytically active Sec7-wt domain can expel the Mg 2ϩ , the P i , and the GDP. Again, even here we would expect a separate additional phosphorus signal for protein-bound P i , which could be severely exchange-broadened beyond detection. A quantitative analysis of the data leads to the conclusion that GDP and P i are released from Arf1 by ARNO-Sec7, since the increase of its signal is not completely explained by the free P i and GDP contained in the Arf1 samples added. Fig. 3a shows a spectrum of Arf1⅐GDP⅐Mg 2ϩ in complex with ARNO-Sec7. After complexing of Mg 2ϩ ions by the addition of EDTA, the signals originating from Arf1⅐GDP⅐Mg 2ϩ complexed with ARNO-Sec7 disappear and are replaced by new signals at Ϫ10.60 and Ϫ4.54 ppm at positions close to those of magnesium-free Arf1⅐GDP (Fig. 4a). The resonance positions do not deviate significantly from those observed for Arf1⅐GDP in the absence of the exchange factor (Table I). However, the resonance lines are significantly broadened, which is indicative for an interaction of Arf1⅐GDP with ARNO-Sec7. A simulation of the data leads to a relaxation time of 2.61 ms for the ␤-phosphate resonance (i.e. the relaxation time decreases by a factor of 1.7 in the presence of ARNO-Sec7). 31 P NMR Spectra of Arf1⅐GDP⅐Mg 2ϩ Complexed with Mutants of ARNO-Sec7-A similar behavior is observed for ARNO-Sec7(E156D), which still contains a negatively charged side chain at the position of the glutamate finger but shortened by one methylene unit (Fig. 3b). Again, the NMR signals are indicative for a superposition of resonance lines of free [⌬17]Arf1⅐GDP⅐Mg 2ϩ and [⌬17]Arf1⅐GDP⅐Mg 2ϩ -ARNO-Sec7(E156D). The resonances of bound Mg 2ϩ ⅐GDP do not shift much but become broadened and clearly inhomogeneous. A broader component of the ARNO-Sec7 complex is especially well visible for the ␤-phosphate resonance, which shows a clear downfield shifted shoulder at Ϫ3.16 ppm. As discussed above, the main difference between wild type ARNO-Sec7 and the E156D mutant is actually at the level of free GDP and P i , which suggests that the mutant does not dissociate GDP from Arf1.
Replacing the glutamate in ARNO-Sec7 by the uncharged amino acid alanine leads to more dramatic changes in the spectrum; a pair of new lines is created at positions similar but not identical to those of magnesium-free Arf1⅐GDP, which indicates that in the complex with ARNO-Sec7(E156A), part of the bound Mg 2ϩ ions are released (Fig. 3c). The addition of ARNO-Sec7(E156A) does not result in an inhomogeneity of ␤-phosphate of line [⌬17]Arf1⅐GDP⅐Mg 2ϩ , which was identified as the complex of [⌬17]Arf1⅐GDP⅐Mg 2ϩ with ARNO-Sec7 in the case of the wild type protein. However, the lines corresponding to the [⌬17]Arf1⅐GDP can only arise from its complex with the ARNO-Sec-7 mutant, since they can only be observed in the presence of this mutant and are significantly shifted and broadened (T 2 of the ␣-phosphate and ␤-phosphate resonances increases to 2.71 ms for both). Qualitatively, the line broadening effect can be seen when comparing the virtually unperturbed lines of [⌬17]Arf1⅐GTP⅐Mg 2ϩ with the lines of bound GDP in the presence of ARNO-Sec7. As a consequence of the E156A mutation, we can hypothesize that residue 156 can expel the Mg 2ϩ through its first CH 2 group of the side chain.
An interesting mutation at position 156 can be created by substituting the negatively charged glutamate by a positively charged lysine. As in the case of the alanine mutant, a complex with magnesium-free Arf1⅐GDP can be detected. In addition, strong signals of free GDP (partly complexed with Mg 2ϩ ) at Ϫ5.67 and Ϫ10.35 ppm are visible in the spectrum. However, these signals originate from free GDP already present in the [⌬17]Arf1⅐GDP⅐Mg 2ϩ sample before adding the exchange factor. As seen in the Ala case, additional signals appear compared with wild type ARNO-Sec7, which can be assigned to the magnesium-free complex of [⌬17]Arf1⅐GDP⅐ARNO-Sec7(E156K) (Fig. 3d). The positively charged side chain lysine seems to weaken the Mg 2ϩ ion bound to the Arf protein by electrostatic interaction leading to magnesium-free complex.
Since the interaction with ARNO-Sec7 influences the Mg 2ϩ binding to [⌬17]Arf1⅐GDP, a set of measurements was performed in the presence of 2 mM EDTA that should establish a concentration of free magnesium of ϳ1 M (Fig. 4). As to be expected at low magnesium concentration, the equilibrium between Arf1⅐GDP and Arf1⅐GDP⅐Mg 2ϩ is shifted to the magne- sium-free complex of Arf1⅐GDP. The resonances of ␣and ␤-phosphate groups of bound, magnesium-free GDP are observable at Ϫ10.60 and Ϫ4.54 ppm. The complexes with wild type and ARNO-Sec7 mutants are characterized by small shift changes (Table I) and a clear line broadening. Note that the used concentrations of EDTA are not sufficient to remove significant amounts of magnesium from Arf1⅐GTP⅐Mg 2ϩ , since its signals are not perturbed. In the spectrum of Arf1⅐GDP in the presence of ARNO-Sec7(E156K), strong signals of free GDP are visible at Ϫ5.67 and Ϫ10.35 ppm. However, this free GDP was already present as impurity in the Arf1 solution used in this case (see above). High free magnesium shifts the equilibrium between Arf1⅐GDP and Arf1⅐GDP⅐Mg 2ϩ also in the presence of wild type ARNO-Sec7 and the ARNO mutants E156D and E156A completely to the magnesium-bound population (data not shown). An exception is the mutant E156K, where even the presence of 9 mM free magnesium is not sufficient to shift the equilibrium of the Arf1-ARNO complex completely to the magnesium-containing state. GTP at Ϫ9.58, Ϫ15.23, and Ϫ4.91 ppm (Fig. 1). In the presence of wild type ARNO-Sec7, no shift changes or line broadenings are observable, indicating that under the concentrations used, the population of the complex with ARNO-Sec7 is very low. Consistently, no GTP is released and observable as free GTP in the spectra. Further evidence is provided by a detailed examination of the corresponding 1 H NMR spectra, which can be deconvoluted into the single component 1 H NMR spectra of Arf1⅐GTP⅐Mg 2ϩ and ARNO-Sec7(wt) (data not shown) without showing a sign of the formation of the ternary complex. nucleotide exchange factor. It was suggested to mimic the ternary complex that precedes the dissociation of GDP and the formation of the stable and binary nucleotide-free complex (10). The phosphorus NMR data show that this abortive complex is indeed magnesium-free and that it forms also with the ARNO-Sec7(E156A). In contrast, the wild-type and the mutant protein (E156D) do not form this complex at higher levels of magnesium; however, phosphorus NMR indicates a weak but NMRdetectable interaction with Arf1⅐GDP⅐Mg 2ϩ .
Interaction of Arf1⅐GTP⅐Mg 2ϩ with ARNO-Sec7-Under similar experimental conditions where in the Ran system the interaction of Ran⅐GTP⅐Mg 2ϩ with RCC1 exhibited a substantial amount of the bound GTP released from the protein (16), we could not detect a release of GTP from Arf1⅐GTP⅐Mg 2ϩ by ARNO-Sec7. In addition, we see no effect on the phosphorus line width of Arf1⅐GTP⅐Mg 2ϩ upon the addition of ARNO-Sec7(wt). In line with these observations, the 1 H spectrum of a solution containing Arf1⅐GTP⅐Mg 2ϩ and ARNO-Sec7(wt) can simply be reconstructed as the sum of the 1 H spectra of the isolated proteins. This indicates again that no specific interaction between the two components is detectable even at the high protein concentrations used (i.e. the affinity for ARNO-sect is very low).
Conclusions-The phosphorus NMR data show that the negative charge of the glutamate residue at position 156 is very important for the interaction of ARNO-Sec7 with Arf1⅐GDP⅐ Mg 2ϩ complexes. At intermediate magnesium concentrations, ARNO-Sec7(wt) and ARNO-Sec7(E156D) interact weakly with Arf1⅐GDP⅐Mg 2ϩ but are not able to form a stable complex with magnesium-free Arf1⅐GDP. After replacing the negative charge at position 156 by introduction of an uncharged (Ala) or positively charged residue (Lys), ARNO-Sec7 shifts the equilibrium to the magnesium-free state and forms a stable complex with Arf1⅐GDP without releasing GDP. To our surprise, the alanine mutant ARNO-Sec7(E156A) was even more effective in displacing the magnesium ion from the catalytic site than the lysine mutant ARNO-Sec7(E156K) with its long, positive charged side chain. The proposal that Glu 156 acts as a "glutamic finger" (10) interacting with the Mg 2ϩ ion and the ␤-phosphate group of bound GDP is supported by our phosphorus NMR data as well as the published crystal structure of the complex (7). ARNO-Sec7 preferentially binds to Arf1⅐GDP⅐Mg 2ϩ and directly displaces GDP (or GDP⅐Mg 2ϩ ) but has a low affinity to the metalfree Arf1 complex, since the negative charge of Glu 156 is here not shielded by the Mg 2ϩ ion. Replacing Glu 156 by Asp reduces the replacement of GDP by the complex formation, since the negative charged side chain is further away from the negatively charged ␤-phosphate group of bound GDP. Replacing Glu 156 by a nonpolar residue (Ala) or a positively charged residue (Lys) reduces the affinity to Arf1⅐GDP⅐Mg 2ϩ and thus leads to a preferential binding to Arf1⅐GDP.
In conclusion, we can hypothesize a dual role for the Glu residue in Sec7. As a first step, the Mg 2ϩ ion has to be removed. This can be achieved through a CH 2 group and is mimicked by the Ala mutation. An even larger stabilization of the Mg 2ϩ -free complex can be reached by a substitution through the positively charged amino acid Lys. As a second step, the GDP is removed by the carboxylate ion of Glu 156 .