The guanine cap of human guanylate-binding protein 1 is responsible for dimerization and self-activation of GTP hydrolysis

Authors


C. Herrmann, Ruhr-Universität Bochum, Physikalische Chemie I, Universitätsstr. 150, 44780 Bochum, Germany
Fax: +49 2343214785
Tel: +49 2343224173
E-mail: chr.herrmann@rub.de

Abstract

Human guanylate-binding protein 1 (hGBP1) belongs to the superfamily of large, dynamin-related GTPases. The expression of hGBP1 is induced by stimulation with interferons (mainly interferon-γ), and it plays a role in different cellular responses to inflammatory cytokines, e.g. pathogen defence, control of proliferation, and angiogenesis. Although other members of the dynamin superfamily show a diversity of cellular functions, they share a common GTPase mechanism that relies on nucleotide-controlled oligomerization and self-activation of the GTPase. Previous structural studies on hGBP1 have suggested a mechanism of GTPase and GDPase activity that, as a critical step, involves dimerization of the large GTP-binding domains. In this study, we show that the guanine cap of hGBP1 is the key structural element responsible for dimerization, and is thereby essential for self-activation of the GTPase activity. Studies of concentration-dependent GTP hydrolysis showed that mutations of residues in the guanine cap, in particular Arg240 and Arg244, resulted in higher dissociation constants of the dimer, whereas the maximum hydrolytic activity was largely unaffected. Additionally, we identified an intramolecular polar contact (Lys62–Asp255) whose mutation leads to a loss of self-activation capability and controlled oligomer formation. We suggest that this contact structurally couples the guanine cap to the switch regions of the GTPase, translating the structural changes that occur upon nucleotide binding to a change in oligomerization and self-activation.

Structured digital abstract

Abbreviations
GBP

guanylate-binding protein

GppNHp

guanosine 5′-(βγ-imino)-triphosphate

hGBP1

human guanylate-binding protein 1

LG

large GTP-binding

mant

N-methylanthraniloyl

Introduction

A variety of regulation processes in cells depend on regulatory GTPases [1,2]. These include signal transduction, e.g. heterotrimeric G-proteins and members of the Ras superfamily [3,4], regulation of translation, e.g. EF-Tu [5], and vesicle recycling, e.g. dynamin [6,7]. These regulatory GTPases cycle between two different states, which are established by the bound nucleotide and the resulting conformation of the protein [8]. Usually, the GTP-bound state is the active state of the GTPase, which allows interaction with effector molecules, whereas the GDP-bound state only weakly interacts with effectors [9]. By intrinsic or GTPase-activating protein-activated hydrolysis of GTP, resulting in the GDP-bound state, the GTPase is inactivated. Guanine nucleotide exchange factors accelerate the exchange of the bound GDP for GTP, and thereby convey a signal for activation [10].

In contrast to the small GTPases of the Ras superfamily, members of the superfamily of large dynamin-related GTPases are characterized by high intrinsic GTPase activity, relatively low nucleotide affinities, high nucleotide dissociation rates, and self-stimulation of GTPase activity (reviewed in [7]). This self-stimulation is usually coupled to nucleotide-dependent oligomerization and/or lipid binding. Members of the dynamin superfamily have roles in numerous membrane processes, such as budding, fission and organelle division [11,12]. Among these dynamin-related proteins, the guanylate-binding proteins (GBPs) form a family of p67 interferon-inducible GTPases. The best characterized member of this family, human GBP1 (hGBP1), was found to exhibit antiviral and antiangiogenic activity, but its cellular function is not yet entirely understood [13–15]. Recent studies have shown involvement of murine GBPs in defence against intracellular pathogens by recruiting NADPH oxidase to the pathogen vacuole, leading to the production of toxic superoxide, which kills the pathogens [16]. Like other dynamins, hGBP1 oligomerizes in a nucleotide-dependent manner and thereby stimulates GTPase activity [17]. Furthermore, it has been shown that hGBP1 has the unique ability to hydrolyse GTP not only to GDP but also, in two successive steps, to GMP, with GMP being the major product [18–20]. Analysis of the structures obtained from X-ray crystallography suggested that dimer formation is responsible for self-stimulation of GTPase activity [21]. Dimer formation is induced upon GTP binding, and leads to the reorientation of the catalytic Arg48 and Ser73 towards the active site. The ‘arginine finger’ (Arg48) contacts the phosphates to stabilize the negative charges developing in the transition state [21]. The role of Ser73 is the nucleophilic activation of the attacking water molecule. Mutation of these residues leads to the loss of self-stimulation, whereas dimer formation after GTP binding is not impaired [21,22]. In this study, we further address the initial step of self-stimulation, the dimerization step. We investigate the guanine cap’s role in dimer formation, and we elucidate the coupling of nucleotide binding and guanine cap reorientation.

Results and Discussion

The guanine cap is essential for dimerization of hGBP1

After investigation of the known structures of the large GTP-binding (LG) domain of hGBP1 [21], residues that participate in intermolecular interactions in the dimerization interface were selected for mutagenesis. Charged residues were mainly chosen, for reasons explained below. These included residues in switch II (Glu105), residues in the guanine cap (Arg240, Arg244, Arg245, and Asp259), and residues following the guanine-binding motif (Ser186 and Asp192) (Fig. S1). To identify the residues involved in dimerization and self-activation of hydrolysis, we measured GTPase activity as a function of hGBP1 concentration, as described previously [17,22,23]. The mutant proteins were incubated at varying concentrations with a large excess of GTP. At different time points, the reaction mixtures were analysed by HPLC to determine the hydrolysis rate. Oligomerization-dependent self-activation of hGBP1 is indicated by an increase in specific activity with higher protein concentrations, because, at higher concentrations, a larger fraction will be in an oligomeric state and thus GTP hydrolysis will be faster (Fig. 1). As described previously [22–24], data can be analysed with a quadratic binding equation that gives two parameters: Kdimer, the apparent dissociation constant of hGBP1 dimers; and smax, the specific activity at saturating protein concentrations. For wild-type hGBP1, these parameters were Kdimer = 0.03 μm and smax = 0.38 s−1 under low-salt conditions. With addition of 200 mm NaCl to the buffer, the apparent dimer dissociation constant was shifted towards 0.68 μm, whereas nucleotide binding and the maximal activity were hardly changed (Table 1). This rather pronounced change in Kdimer suggests that dimerization is strongly driven by electrostatic interactions. Therefore, we focused mainly on charged residues in the protein–protein interface for our mutagenesis study.

Figure 1.

 Concentration-dependent GTP hydrolysis catalyzed by wild-type hGBP1 and the E105A, R240A and R244A mutants. The initial rates of GTP turnover were normalized to the protein concentration (specific activity) and plotted against the protein concentration. The maximal specific activity (smax) and dimer dissociation constant (Kdimer) were obtained from a quadratic equation modelling GTPase stimulation by dimer formation. The parameters obtained for the different mutants are summarized in Table 1. WT, wild-type.

Table 1.   Parameters of GTP hydrolysis obtained by concentration-dependent hydrolysis at 25 °C (constants as defined in the text). ND, not determined; WT, wild-type.
MutationKdimerm)smax (s−1)ΔΔG (kJ·mol−1)
WT0.030.38
WT + 200 mm NaCl0.680.367.7
K62AND0.07
E105A0.050.281.3
S186A0.030.370.0
D192A0.100.403.0
R240A2.50.2211.0
R244A3.10.3411.5
R245A0.150.204.0
D255AND0.06
D259A0.170.274.3

hGBP1 mutants were analysed in the same manner as the wild-type, and Kdimer and smax values were determined (Fig. 1; Table 1). As with the wild-type, only almost undetectable GDPase activity was observed for all mutants studied, and the GMP/GDP product ratio was similar for the wild-type and all mutants. Mutations of Glu105, Asp192 and Ser186 to alanine had only a minor effect on Kdimer (less than fourfold). In contrast, mutations of residues located in the guanine cap (residues 239–259), in particular Arg240 and Arg244, to alanine yielded a strong increase in Kdimer (Fig. 1). These two mutations, R240A and R244A, led to 75-fold and 120-fold increases in Kdimer, respectively (Table 1). Thus, mutation of these guanine cap arginines decreases the ability of hGBP1 to self-stimulate by weakening the LG dimer interaction as compared with the wild-type (see also size-exclusion chromatography). By using Kdimer to calculate the change in free energy with the Gibbs–Helmholtz equation, it is possible to estimate the contributions of single residue contacts to the dimerization indirectly from the self-activation properties. Whereas the whole dimerization energy derived by this method was ∼ 43 kJ·mol−1, mutations of Arg240 and Arg244 to alanine led to calculated ΔΔG values of 11.0 and 11.5 kJ·mol−1, respectively. Combination of these ΔΔG values showed that these two residues together are responsible for half of the binding energy of the dimer, emphasizing the crucial importance of Arg240 and Arg244 for hGBP1 self-activation by dimerization. Whereas the changes in Kdimer caused by these mutations were about two orders of magnitude, the changes in smax were small, and did not exceed threefold. Hence, a high GTP turnover rate can still be achieved by these mutants if high concentrations of hGBP1 are present to saturate the monomer–dimer equilibrium. This indicates that the GTP hydrolysis step itself is not impaired.

Mutations of two other guanine cap residues, Arg245 or Asp259, to alanine indicated that each residue is responsible for about 10% of the binding energy. As described above, mutations outside the guanine cap (D192A, E105A, and S186A) have only a small effect on the binding energy of the dimer (up to 7% for D192A). In addition, two other residues in the switch regions, Glu72 (switch I) and Asn109 (switch II), that form intermolecular contacts in the crystal structure were mutated previously, and they did not show a large difference Kdimer [22]. This identifies the guanine cap as the major element responsible for dimer formation in hGBP1.

Mutations of the guanine cap do not affect nucleotide binding

Given the fact that the guanine cap forms a hydrophobic pocket for the guanine base moiety, less self-stimulation might be also caused by lower nucleotide binding affinities and not exclusively by decreased dimerization. In order to exclude any effects of different nucleotide binding affinities, we performed fluorescence titrations with N-methylanthraniloyl (mant)-labelled nucleotides. Representative fluorescence titrations are shown in Figs 2 and S2, and the obtained dissociation constants are summarized in Table S1. The observed nucleotide dissociation constants showed only marginal changes as compared with those for the wild-type reported earlier [19,22,23]. As for the wild-type, the relative binding affinities were GMP > guanosine 5′-(βγ-imino)-triphosphate (GppNHp) > GDP. Our original assumption of an exclusive effect on dimerization is strongly supported by these observations of similar nucleotide-binding properties.

Figure 2.

 Representative mant-nucleotide binding experiment with wild-type hGBP1 (upper panel) or the hGBP1 mutant D255A (lower panel). A solution containing 0.5 μm mant-GMP (circles), mant-GDP (triangles) or mant-GppNHp (squares) was titrated with wild-type hGBP1 and the D255A mutant, respectively, and the observed fluorescence was normalized to the fluorescence of the nucleotide alone. The obtained nucleotide dissociation constants of all mutants used in this study are summarized in Table S1. WT, wild-type.

Dimerization at the LG domains is inhibited, but tetramers are still formed

After analysing the concentration-dependent self-stimulation of the mutants, as shown above, we were interested in the ability of hGBP1 to form oligomers in complex with various nucleotides and nucleotide analogues. The wild-type has been shown to form dimers in complex with the GTP analogue GppNHp via an interaction of the LG domains [17,21]. In recent studies, we have identified a second interaction interface at the helical domain, which is accessible only as a result of GTP hydrolysis [25,26]. This leads to the formation of tetramers, which can be trapped by the complex of GDP and aluminium fluoride [17,25,26]. Thus, we used size-exclusion chromatography to investigate whether contact formation of two LG domains and of two helical domains occur independently.

To directly analyse the oligomerization behaviour of the mutants, we performed analytical gel filtration experiments. In agreement with the hydrolysis data, we observed that mutants with strongly increased Kdimer values did not form or only partly formed dimers in presence of GppNHp (Figs 3 and S3; Table 2). With a protein concentration (20 μm) that saturates dimer formation of the wild-type, no sign of dimer formation was observed with the R240A mutant. The R244A, R245A and D259A mutants showed increasing fractions of a dimeric species, but > 90% of the protein was in a monomeric state. The D192A and E105A mutants exhibited very similar behaviour to the wild-type. These mutants formed mainly dimers, and only a very small fraction of monomeric protein was observed, in good agreement with the marginal inhibition of self-stimulation in concentration-dependent GTPase activity. Thus, the GppNHp-dependent dimerization reflects the results of self-activation obtained by GTP hydrolysis experiments, and directly proves the involvement of Arg240 and Arg244 in dimer formation of hGBP1.

Figure 3.

 Size-exclusion chromatography experiment with wild-type hGBP1 (upper panel) or the guanine cap mutant R240A (lower panel) in the nucleotide-free (dotted lines), GppNHp (solid black line) and GDP aluminium fluoride (grey line) states. The molecular masses of standard proteins are indicated by arrows. Elution of all proteins was followed by absorbance at 280 nm, and the elution volume (Ve) was normalized to the exclusion volume (V0). The results for other mutants used in this study are summarized in Table 2 and in more detail in Table S2. WT, wild-type.

Table 2.   Oligomerization of hGBP1 mutants. WT, wild-type.
 Nucleotide-freeGppNHpGDP AlFx
WTMonomerDimerTetramer
K62AOligomer ≥ 200 kDaOligomer ≥ 200 kDaOligomer ≥ 200 kDa
E105AMonomerDimerTetramer
S186AMonomerDimerTetramer
D192AMonomerDimerTetramer
R240AMonomerMonomerMostly tetramer
R244AMonomerMonomerMostly tetramer
R245AMonomerMostly dimerMostly tetramer
D255AOligomer ≥ 200 kDaOligomer ≥ 200 kDaOligomer ≥ 200 kDa
D259AMonomerMostly dimerMostly tetramer

As described above, hGBP1 forms tetramers in the ‘trapped state’ of GTP hydrolysis with GDP aluminium fluoride [22,23]. All mutants that showed only minor effects on self-activated hydrolysis and dimerization were also able to form tetramers in complex with GDP aluminium fluoride, similar to the wild-type (Tables 2 and S2). Surprisingly, proteins with mutations that significantly impaired self-stimulation and dimerization were also able to form tetramers in the presence of GDP aluminium fluoride. In contrast to the wild-type, these mutants showed an elution profile corresponding to the tetramer and, partially, the monomer (Fig. 3). The ability of the R240A and R244A mutants to form higher oligomers, even though the dimerization was strongly impaired, can be explained by the presence of two independent interaction sites, as described above. In the case of the wild-type, the first dimerization at the LG domains occurs as a result of GTP binding. In the course of hydrolysis, the secondary interface becomes accessible, and additional interaction at the C-terminal part occurs, resulting in the formation of a tetramer, i.e. a dimer of a dimer. In contrast to the behaviour of the wild-type, mutations of the LG dimer interface weaken the first dimerization step, but the mutants are still able to form the contact at the secondary interface. This interaction of the C-terminal parts of hGBP1 results in two weak dimerization sites in proximity, facilitating tetramer formation.

The guanine cap indirectly senses the bound nucleotide

By analysing the guanine cap conformation in different nucleotide states, we found that, in the GMP-bound state of the LG domain, the guanine cap has an ‘open’ conformation, which does not allow dimerization (Fig. 4). X-ray structures of GTP-bound or GDP-bound hGBP1 are not available, as these nucleotides are hydrolysed. The hGBP1 structures of the trapped states in complex with one of the analogues, GppNHp, GDP aluminium fluoride, or GMP aluminium fluoride, show the guanine cap in a ‘closed’ conformation that facilitates dimer formation of two LG domains. Thus, there is a relationship between the guanine cap conformation and the presence of a β-phosphate. Detailed analysis of the LG domain X-ray structures reveals an intramolecular electrostatic contact between the guanine cap Asp255 and Lys62 located in the region between the P loop (residues 45–52) and switch I (residues 65–77). This contact is found in the GTP analogue structures but is lost in the GMP-bound LG domain, leading to the hypothesis that it could be responsible for the ‘closed’ guanine cap conformation resulting in dimer formation and, subsequently, GTPase stimulation. It was shown earlier that GTP binding and not hydrolysis, i.e. binding of nonhydrolysable GppNHp, results in dimer formation of hGBP1. Thus, the loss of the Asp255–Lys62 contact should result in loss of self-stimulation capability, and the GTPase activity should be at a low level and concentration-independent; that is, only the unstimulated monomer activity should be observed. Indeed, after introducing an alanine mutation at either side of the contact (K62A or D255A), we found a nearly constant specific activity at the unstimulated level of GTP turnover over three orders of magnitude of hGBP1 concentration (Fig. 5). By fluorescence titration, we showed that the nucleotide binding is similar to that of the wild-type, indicating that the effect on hydrolysis is not caused by changed nucleotide binding. Using size-exclusion chromatography, we found the oligomer state of both the K62A and D255A mutants to be independent of the bound nucleotide. Control of defined hGBP1 oligomer formation seems to be lost, because, with every nucleotide or nucleotide analogue used, we observed similar elution profiles, showing oligomers with molecular masses > 200 kDa corresponding to a size of more than a tetramer. This demonstrates that the contact between Lys62 and Asp255 or Ala255 is essential for the control of oligomer formation by the nucleotide state. We conclude that the contact between Lys62, in proximity to the essential switch I Thr75, and the guanine cap Asp255 is responsible for the switch-like character of the guanine cap, and is of crucial importance for the establishment of the guanine cap conformation. As a result of this active conformation, the LG domains are able to dimerize, being mainly stabilized by Arg240 and Arg244. This leads to the rearrangement of the catalytically important residues, such as Arg48 and Ser72 [21], and thereby self-stimulation of GTPase activity.

Figure 4.

 Conformations of hGBP1’s guanine cap in different nucleotide states. (A) Structural overview of the GppNHp-bound hGBP1 LG domain dimer [Protein Data Bank (PDB) 2bc9 [20]]. The two monomers (blue and green) are facing each other, and the guanine caps are highlighted in purple or yellow. (B) The electrostatic contact Lys62–Asp255 is lost when GMP is bound, and the guanine cap adopts a relaxed conformation that does not support dimer formation. The colours of the protein chain used are blue for GppNHp (PDB 2bc9) and white for GMP (PDB 2d4h).

Figure 5.

 Concentration-dependent GTP hydrolysis of wild-type hGBP1 (solid squares) and the K62A (filled circles) and D255A (empty triangles) mutants. Data were treated similarly to the data shown in Fig. 1.

Concluding remarks

Previous studies on the oligomerization of hGBP1 focused on the role of dimerization in GMP production [27] and dimer formation in living cells [28]. However, the crucial residue positions for dimerization and the resulting self-activation remain elusive. In this study, we show the importance of hGBP1’s guanine cap for the self-stimulation of GTPase activity in solution, which was shown to be similar to GTPase activity of fully modified hGBP1 bound on lipids [29]. By mutagenesis, we were able to identify Arg240 and Arg244 as the major determinants of GTP-induced dimerization, which, in turn, leads to efficient self-stimulation of the GTPase activity. Using the apparent dimerization constant, we were able to estimate the relative contributions of single residues to dimerization, and this showed that about 50% of the energy is attributable to the guanine cap residues 240 and 244. By further analysis of the guanine cap conformation in the published crystal structures, an electrostatic contact between the guanine cap (Asp255) and the switch I region (Lys62) was identified. Mutation of the residues in this contact completely abolished self-activation of hGBP1. We conclude that, after GTP binding to the LG domain, the contact between switch I and the guanine cap is formed, which ‘transmits’ the information about the nucleotide state to the guanine cap. As a result, the guanine adopts its ‘active’ conformation and the LG domains form a dimer. This dimerization leads to the well-described activation of GTP hydrolysis by rearrangement of the catalytically important residues Arg48 and Ser72. The results presented in this study give insights into the early processes of hGBP1’s self-stimulation, and the mutations characterized might be valuable tools for further understanding hGBP1’s cellular function.

Experimental procedures

Site-directed mutagenesis and protein purification

Wild-type hGBP1 and the mutants used in this study were cloned into pQE80L expression vectors (Qiagen, Hildesheim, Germany), expressed in Escherichia coli BL21(DE3), and purified as described previously [19]. Mutations were introduced by QuikChange site-directed mutagenesis (Stratagene, Amsterdam, The Netherlands), according to the manufacturer’s instructions. All introduced mutations were verified by DNA sequencing with a 3130xl sequencer (Applied Biosystems, Foster City, CA, USA). Concentrations of the purified proteins were measured by UV absorption at 276 nm [ε276 nm = 45 400 (m·cm)−1] [19].

Hydrolysis assay

Hydrolysis assays were performed as described previously [22,23], with 350 μm GTP (Sigma-Aldrich, Munich, Germany) and different concentrations of hGBP1 in buffer C (50 mm Tris, pH 7.9, 5 mm MgCl2, and 2 mm dithiothreitol) at 25 °C. The high concentration of GTP ensures near complete saturation of the protein molecules with nucleotide. Aliquots were injected onto a Chromolith RP18e HPLC column (Merck, Darmstadt, Germany) and eluted with HPLC buffer [100 mm potassium phosphate, 10 mm tetrabutylammonium bromide, and 1.25% (v/v) acetonitrile], and absorption was analysed at 254 nm with a MD-2015 diode array detector (Jasco, Gross-Umstadt, Germany). The initial (linear) phase of steady-state hydrolysis (remaining GTP > 60%) was analysed by linear regression, and the resulting rates were normalized to the protein concentration, yielding the specific activity. Data were analysed as described previously, with a quadratic binding equation that gives two parameters: Kdimer, the apparent dissociation constant of hGBP1 dimers, and smax, the specific activity at saturating protein concentrations [22–24].

Size-exclusion chromatography

Analytical gel filtration experiments were performed with a Superdex 200 10/300 (GE Healthcare, München, Germany) gel filtration column. The elution buffer (50 mm Tris, pH 7.9, 5 mm MgCl2, and 2 mm dithiothreitol) contained 200 μm of the nucleotide and, in the case of GDP/GMP aluminium fluoride, additionally 300 μm AlCl3 and 10 mm NaF. Protein at a concentration of 20 μm was preincubated in the elution buffer for 5 min on ice prior to injection. Size calibration was carried out using standard proteins with molecular masses between 29 kDa and 200 kDa (the corresponding elution volumes are marked by arrows in the plots). The void volume (V0) was measured with the use of Blue Dextran. Elution was followed by monitoring the absorbance at 280 nm with an MD2015 diode array detector (Jasco).

Fluorescence titrations

Fluorescence titrations were performed at 25 °C with a Kontron SFM25 fluorospectrometer (Kontron, Zürich, Switzerland) and 2′/3′-mant-labelled nucleotides (Jena Bioscience, Jena, Germany). The excitation and emission wavelengths were 366 nm and 435 nm, respectively. Mant-labelled nucleotide (0.5 μm) was titrated with protein solutions (typically 100 μm) containing 0.5 μm mant-labelled nucleotide. The data were analysed with a quadratic binding equation as described previously [19,22,23].

Acknowledgement

This work was financially supported by Deutsche Forschungsgemeinschaft (DFG).

Ancillary