Biochemical properties of the human guanylate binding protein 5 and a tumor-specific truncated splice variant


C. Herrmann, Ruhr-Universität Bochum, Physikalische Chemie I – AG Proteininteraktionen, Universitätsstraße 150, 44800 Bochum, Germany
Fax: +49 2343214785
Tel: +49 2343224173


The human guanylate binding protein 5 (hGBP5) belongs to the family of interferon-γ-inducible large GTPases, which are well known for their high induction by pro-inflammatory cytokines. The cellular role of this protein family is unclear at this point, but there are indications for antiviral and antibacterial activity of hGBP1. hGBP5 exists in three splice variants, forming two different proteins, of which the tumor-specific one is C-terminally truncated by 97 amino acids, and therefore lacks the CaaX motif for geranylgeranylation. Here we present biochemical data on the splice variants of hGBP5. We show that, unlike hGBP1, hGBP5a/b and hGBP5ta do not bind GMP or produce any GMP during hydrolysis despite the fact the residues involved in GMP production from hGBP1 are conserved in hGBP5. Hydrolysis of GTP is concentration-dependent and shows weak self-activation. Thermodynamic studies showed strongly negative entropic changes during nucleotide binding, which reflect structural ordering in the protein during nucleotide binding. These structural changes were also observed during changes in the oligomerization state. We observed only a minor influence of the C-terminal truncation on hydrolysis, nucleotide binding and oligomerization of hGBP5. Based on these similarities we speculate that the missing C-terminal part, which also carries the geranylgeranylation motif, is the reason for the dysregulation of hGBP5′s function in lymphoma cells.

Structured digital abstract


guanosine 5′-(β,γ-imino)-triphosphate

GTPγS guanosine



human guanylate binding protein


isothermal titration calorimetry




A large variety of cellular processes, including signal transduction, regulation of transcription and membrane deformation, are regulated by GTP-binding proteins [1,2]. Usually these proteins act as ‘molecular switches’ that strongly interact with their effector proteins in the GTP-bound state (‘on’ state) but only very weakly in their GDP-bound state (‘off’ state) [3]. By hydrolysis of the GTP to GDP and Pi, these proteins are able to ‘switch off’ intrinsically, or GTP cleavage may be promoted by interaction with a GTPase-activating protein. Small regulatory GTP-binding proteins from the Ras superfamily usually only exchange the bound nucleotide very slowly, but the exchange rate is dramatically increased in the presence of guanine nucleotide exchange factors, and can be regulated. GTPases from the dynamin superfamily (reviewed in [4]), on the other hand, interact dynamically with the bound nucleotide, and usually bind nucleotides in the micromolar range. In addition to these striking differences in nucleotide binding, these large GTPases are catalytically very active, show self-activating GTP hydrolysis, and are functional as oligomers rather than monomers.

Human guanylate binding protein 5 (hGBP5) belongs to the family of interferon-γ-induced p65 GTPases, which has seven members in the human genome [5]. This family of guanylate binding proteins was originally identified by its ability to bind to immobilized guanine nucleotides with similar affinities for GTP, GDP and GMP [6,7]. Within this family, the best characterized member, human guanylate binding protein 1 (hGBP1), has been shown to exhibit a unique hydrolytic activity – cleavage of not only the γ-phosphate but also the β-phosphate in a successive step [8,9]. In contrast to the small regulatory GTP-binding proteins of the Ras superfamily, hGBP1 is characterized by highly dynamic nucleotide binding, low affinities for all three guanine nucleotides, and high, self-activated GTPase activity, which is promoted by oligomerization. These features are generally conserved in high-molecular-weight GTPases such as dynamin and the antiviral protein, Mx. Because of similarities in the molecular architecture of GBPs and members of the dynamin family, GBPs are classified as a part of the dynamin superfamily. Although their biochemical characteristics are well understood, there is only an incomplete picture of the cellular functions of hGBP1. Previous studies showed an antiviral effect against specific viruses [10,11], and an anti-Chlamydia effect [12], inhibition of endothelial cell proliferation [13] and their subcellular localization [14] have also been investigated recently. Similar to the recent report of an anti-Chlamydia effect of hGBP1, there are indications of murine GBP5 co-localization with membrane ruffles formed by invading Salmonella enterica, and a positive regulation of pyroptosis to defend against infection by S. enterica and possibly other bacterial pathogens [15]. Although no other proteins from the guanylate-binding protein family are known to exist as more than one splice variant, hGBP5 does exist as three splice variants, which form two different proteins [16]. These two proteins differ with respect to the presence or absence of the C-terminal 97 amino acids including the C-terminal geranylgeranylation motif. If a fold similar to that of hGBP1 is assumed, this deletion corresponds to deletion of extended helices 12 and 13. In healthy cells, only the a/b splice variant is expressed, but the truncated splice variant has been detected in all melanoma and most of lymphoma cell lines tested.

In this study, we focus on the biochemical properties of both splice variants of hGBP5, hGBP5a/b (amino acids 1-586) and the C-terminally truncated hGBP5ta (amino acids 1-489), using isothermal titration calorimetry (ITC), concentration-dependent GTPase assays, fluorescence titrations and analytical gel filtration.

Results and Discussion

Hydrolytic activity of hGBP5a/b and hGBP5ta

In light of previous observations for hGBP1 and other large GTPases, we analysed the enzymatic activity of hGBP5a/b and hGBP5ta in a concentration-dependent manner. Various concentrations of purified hGBP5a/b and hGBP5ta were incubated with 350 μm of GTP, and aliquots were taken after various durations and analysed by C18 reverse-phase HPLC. Initial rates of GTP hydrolysis were normalized to the protein concentration (specific activity) and plotted against the protein concentration (Fig. 1).

Figure 1.

 Initial rates of GTP hydrolysis were normalized to the protein concentration (specific activity), and plotted against the protein concentration. Both hGBP5a/b (circles) and hGBP5ta (squares) show an approximately two-fold self-activation of GTP hydrolysis.

We observed weak concentration-dependent self-activation (approximately two-fold) from a basal specific activity at low protein concentrations to maximum turnovers at 25 °C of 0.054 and 0.077 s−1 for hGBP5a/b and hGBP5ta, respectively. Increasing the temperature from 25 °C to 37 °C resulted in an approximately three-fold increase in specific activity (see Fig. S1). Both splice variants exhibited positive cooperativity in specific activity, with Hill coefficients of 2.4 and 1.9 and dissociation constants (KdHill) of 4 and 2 μm for hGBP5a/b and hGBP5ta, respectively. The proteins are very similar in terms of their enzymatic activity, so the activity is not affected by deletion of the C-terminal part. In contrast to results obtained for hGBP1 and hGBP2 [17], we observed absolutely no formation of GMP at any protein concentration or time, indicating a different mechanism of hydrolysis compared to hGBP1. To confirm that hGBP5 does not require higher temperatures to form GMP, we performed additional hydrolysis measurements at 37 °C, but did not detect GMP formation (see Fig. S1). As described previously [18], certain positions in the hGBP1 large GTPase domain are crucial for the formation of GMP by hGBP1. Surprisingly, these residues (namely R48, H74, K76, E99, K106 and D112 in the primary sequence of hGBP1) are conserved in the primary structure of the GTPase domain of hGBP5, so the formation of GMP must be impaired in an as yet unknown manner. Sequence alignments of the large GTP-binding domains of hGBP1, hGBP2 and hGBP5 showed many substitutions specific to hGBP5, so we cannot conclude at present which residues are responsible for the lack of GMP formation (see Fig. S2). The similar hydrolytic activities of the a/b and C-terminally truncated forms of hGBP5 are in contrast to that of the analogous deletion mutant of hGBP1 (S. Kunzelmann, Ruhr-University Bochum and C. Herrmann, unpublished results), which resulted in a 2.5-fold increase in specific GTPase activity and increased formation of GMP during hydrolysis.

Thermodynamics and stoichiometry of nucleotide binding

We used ITC to investigate the thermodynamics of nucleotide binding. Using this method, it is possible to determine the stoichiometry, enthalpy change, dissociation constant and thereby the change in entropy in a single experiment. No label is required in these experiments as the heat change of the reaction is measured.

Using ITC, we found a 1 : 1 stoichiometry of nucleotide to protein with a deviation of less than 10%. We found comparable dissociation constants for GDP and GppNHp [guanosine 5′-(β,γ-imino)-triphosphate], with Kd values of 11 and 5 μm for hGBP5ta and 7.2 and 2.6 μm for hGBP5a/b, respectively. For the non-hydrolysable analog GTPγS [guanosine 5′-O-(γ-thio)triphosphate], the dissociation constant is slightly higher (see Table 1). In all experiments, nucleotide binding yielded large exothermic peaks, with corresponding ΔH values ranging from −14.7 kcal·mol−1 for GDP binding to hGBP5a/b and −38.1 kcal·mol−1 for binding of GTPγS to hGBP5ta. In light of the thermodynamic data and the oligomerization behavior, we attribute the rather pronounced negative changes in enthalpy not simply to nucleotide binding, but to a combination of several exothermic effects occurring at the same time, most likely nucleotide binding and coupled oligomerization.

Table 1.   Thermodynamic parameters of nucleotide binding by hGBP5a/b and hGBP5ta at 25 °C as measured by ITC.
NucleotideKdm)ΔH (kcal·mol−1)T·ΔS (kcal·mol−1)
GDP7.2 ± 0.3−14.7 ± 0.1−7.7 ± 0.2
GppNHp2.6 ± 0.3−31.3 ± 0.4−23.5 ± 0.6
GTPγS15 ± 3−21.4 ± 0.8−14.9 ± 0.9
GDP11 ± 1−16.5 ± 0.1−9.7 ± 0.2
GppNHp5 ± 0.2−25.3 ± 0.3−18.2 ± 0.3
GTPγS26 ± 4−38.1 ± 0.8−31.9 ± 0.9

All nucleotide binding experiments showed a negative change in ΔS, which is compensated for by the large and negative ΔH value. This negative change of entropy is most pronounced for hGBP5a/b using the GTP analog GppNHp (T·Δ= −23.5 kcal·mol−1) and for hGBP5ta using GTPγS (T·Δ= −31.9 kcal·mol−1). The T·ΔS values for GDP are less negative (> −10 kcal·mol−1) for both splice variants. We attribute these pronounced entropic compensations to a loss of conformational freedom upon nucleotide binding, and, especially when using the GTP analogs, to the formation of higher-order oligomers. In the case of hGBP1, the entropy changes for nucleotide binding are generally positive but less than 5 kcal·mol−1. Despite the fact that both hGBP5 and hGBP1 form oligomers in a nucleotide-dependent manner, the entropic contributions of nucleotide binding and the coupled oligomerization show opposing signs. In the case of hGBP1, release of water from the nucleotide binding pocket and the protein surface may counterbalance the changes in conformational flexibility imposed by nucleotide binding, while in the case of hGBP5, conformational restrictions after nucleotide binding and oligomer formation appear to result in strong entropic penalties.

In contrast to the findings for hGBP1 [8], we did not observe any signals in ITC experiments when using GMP. Neither the a/b form nor the truncated splice variant exhibit any GMP binding in the micromolar range (see Fig. 2 for a representative experiment). The missing signals in our ITC experiments are not due to a very low ΔH at 25 °C, but are due to the absence of GMP binding, as confirmed by competitive fluorescence measurements.

Figure 2.

 Representative isothermal titration calorimetry runs using hGBP5ta at 25 °C. hGBP5ta (150 μm) was titrated with 2.25 mm of nucleotide (GMP, open circles; GDP, filled squares; GppNHp, open squares). The raw experimental data and the processed data (including the fitted curves using a one-site binding model) are shown in the upper and lower panels, respectively.

Binding and dynamics of fluorescent nucleotide analogs

To investigate the binding of fluorescent nucleotide analogs, we used fluorescence titrations as described previously for hGBP1 [8,19]. When titrating increasing amounts of hGBP5a/b or hGBP5ta with a solution of N-methylanthraniloyl-GDP (mant-GDP) or mant-GTPγS, an increase in fluorescence intensity was observed, as described previously for hGBP1 [8]. The observed approximately 3.5-fold increase in fluorescence is indicative of a non-solvent-accessible binding pocket, as found in the crystal structure of hGBP1 [20,21]. Analysis of the fluorescence intensities using a quadratic binding equation yielded the dissociation constants summarized in Table 2. The splice variants hGBP5a/b and hGBP5ta did not show any significant differences in mant-nucleotide binding. The observed dissociation constant for mant-GTPγS (11 μm) is higher than that for mant-GDP (3 μm), which is the converse of the tendency observed with unlabeled nucleotides in ITC experiments (see Fig. 3). To further confirm the lack of GMP binding by hGBP5 splice variants as observed in the ITC experiments, we performed fluorescence titrations using mant-GMP. Similar to the ITC experiments with GMP, we only observed a small effect, i.e. a marginal increase in mant-GMP fluorescence in the presence of 50 μm hGBP5 (data not shown). Furthermore, we attempted to displace bound mant-GDP using a high excess (3000-fold) of GMP, but only observed a decrease in fluorescence of less than 10%, indicating a much higher dissociation constant for GMP than for mant-GDP. In contrast, displacement of mant-GDP from hGBP5 was efficient at low concentrations of competing GDP, GppNHp and GTPγS (data not shown). When investigating mant-GppNHp binding to hGBP5a/b or hGBP5ta, we found very low reaction rate constants that were at least 100-fold smaller than for all other nucleotides tested (data not shown). Because of the slow reaction of mant-GppNHp and hGBP5, fluorescence measurements were not feasible due to long-term protein instability. Similar to these fluorescence experiments, a rather slow return to baseline in the ITC experiments was observed, but, without the mant label, GppNHp binding is fast enough to allow equilibration after each injection.

Table 2.   Dissociation constants of mant-labeled nucleotides as measured by fluorescence titrations.
mant-GDP3.1 ± 0.33.2 ± 0.2
mant-GTPγS11.3 ± 0.410.8 ± 0.4
Figure 3.

 Fluorescence titration data showing similar dissociation constants for binding of mant-GDP (hGBP5a/b, filled circles; hGBP5ta, open circles) and for binding of mant-GTPγS (hGBP5a/b, filled squares; hGBP5ta, open squares) to hGBP5a/b or hGBP5ta, respectively. Relative fluorescence values are plotted against the concentration of added protein.

To investigate the dynamics of nucleotide binding, constant concentrations of the nucleotides were mixed with increasing amounts of hGBP5a/b or hGBP5ta (see Fig. 4 for representative data). In the case of mant-GTP, increasing amounts of mant-GTP were mixed with a small concentration of hGBP5a/b or hGBP5ta to avoid fluorescence changes due to nucleotide hydrolysis.

Figure 4.

 Nucleotide dynamics determined using stopped flow measurements. (A) Representative fluorescence traces of mant-GDP binding to hGBP5ta. The hGBP5 concentrations increase from 2.5 μm to 25 μm from the bottom up. (B) The observed rates of mant-GDP (filled circles) and mant-GTPγS (open circles) association are plotted against the protein concentration. (C) Observed association rates of mant-GTP with hGBP5a/b (open circles) or hGBP5ta (filled circles).

The association experiments for all nucleotides measured only show a single rate constant. In contrast, the displacement experiments using mant-GTPγS revealed two distinct rate constants for dissociation, with the relative amplitudes of 60% for the faster process and 40% for the slower process, which is indicative of a difference in binding dynamics of the 2′/3′-OH mant-labeled GTPγS as the two isomers occur in approximately this ratio. We did not find any evidence for two dissociation rates for mant-GDP (see Fig. S3). Using hGBP5, we found rate constants that were approximately 100-fold lower for mant-GTP and approximately 30-fold lower for mant-GDP than those reported for hGBP1 [9]. The association rate constants (kon) for mant-GDP (0.046 μm−1·s−1), mant-GTPγS (0.099 μm−1·s−1) and mant-GTP (0.014 μm−1·s−1) were quite similar. In contrast, we found a five-fold faster dissociation of mant-GTPγS (1.11 s−1) compared to mant-GDP (0.23 s−1) and mant-GTP (0.62 s−1). The three-fold lower dissociation rate (koff) for mant-GTPγS was comparable to that for mant-GDP, but is not due to partial hydrolysis of the nucleotide, as verified by HPLC analysis (Table 3). We found association rate constants for hGBP5ta with mant-GDP, mant-GTPγS and mant-GTP of 0.066, 0.072 and 0.025 μm−1·s−1, respectively. The dissociation rate constant obtained from the intercept of the plot in Fig. 4, parts B and C (inline image) and the directly measured dissociation rate constants from displacement experiments (inline image) in Fig. S3 are similar, and show a five- to 10-fold lower rate for dissociation of mant-GDP (0.129/0.23 s−1) from hGBP5ta than for mant-GTPγS (1.15 s−1), and the value for mant-GTP is between those two values (0.7 s−1). Similarly, the slower dissociation rate constant for mant-GTPγS was comparable to that for mant-GDP (see Table 3).

Table 3.   Nucleotide binding dynamics of hGBP5a/b and hGBP5ta. The dissociation constants were calculated using the relationship Kd = koffdiss/kon, except for mant-GTP, for which koffintercept was used.
Nucleotidekonm−1·s−1)koffintercept (s−1)koffdiss (s−1)Kdm)
  1. a Corresponding relative amplitude = 0.6. b Corresponding relative amplitude = 0.4.

mant-GDP0.046 ± 0.0030.28 ± 0.040.23 ± 0.025.0 ± 0.8
mant-GTPγS0.099 ± 0.0032.08 ± 0.041.11 ± 0.01a11.2 ± 0.5
   0.19 ± 0.02b1.9 ± 0.3
mant-GTP0.014 ± 0.0010.62 ± 0.0144 ± 4*
mant-GDP0.066 ± 0.0040.13 ± 0.010.23 ± 0.013.5 ± 0.3
mant-GTPγS0.072 ± 0.0061.4 ± 0.11.15 ± 0.05a16 ± 2
   0.20 ± 0.03b2.8 ± 0.7
mant-GTP0.025 ± 0.0050.72 ± 0.0730 ± 10*

Oligomerization of hGBP5a/b and hGBP5ta

Large GTPases, especially of the dynamin superfamily, are well known for their ability to oligomerize and the resulting stimulation of their hydrolytic activity. In particular, hGBP1 is known to form homodimers after GppNHp binding and larger oligomers with the transition state analog GDP and aluminum fluoride (AlFx) [18]. We used analytical size-exclusion chromatography to characterize the nucleotide-dependent oligomerization state of the protein. Protein (20 μm) was preincubated with a 10-fold excess of the nucleotide for 30 min on ice prior to injection onto a Superdex 200 column equilibrated with buffer containing 200 μm of nucleotide.

The C-terminally truncated splice variant hGBP5ta (theoretical molecular mass 55 kDa) shows an apparent molecular mass (Mapp) in the GDP-bound form of approximately 70 kDa, most likely due to an elongated shape of the protein. The same behavior is observed in the presence of GDP·AlFx. Because of the unexpected observation of a monomeric protein species with both GDP and GDP·AlFx, we tested AlFx binding to the protein by adding 300 μm AlCl3 and 10 mm NaF to a solution containing hGBP5a/b or hGBP5ta and GTP, and analysed GTPase activity. We observed a reduction of GTPase activity of approximately 40%. This small inhibitory effect suggests weak binding of AlFx to the protein. In the GTP-bound (as well as the nucleotide-free) state, a Mapp of approximately 120 kDa is observed, which probably corresponds to a dimer. When bound to GppNHp, the Mapp of hGBP5ta is increased to approximately 200 kDa, which corresponds to a tetramer, while the presence of GTPγS leads to an oligomer size intermediate between those with GTP and GppNHp (see Fig. 5A). The putative transition state hGBP5·GDP·AlFx resembles the product state more closely than a GTP-bound form does. Similar to the monomeric species for the GDP-bound state of hGBP5ta, a recent study found that dynamin oligomers are repeatedly dissociated while GTP hydrolysis occurs [22].

Figure 5.

 Size-exclusion chromatography of the splice variants hGBP5ta (A) and hGBP5a/b (B), respectively, with various nucleotides [red, GDP; green, GDP·AlFx; magenta, GTP; blue, GppNHp; cyan, GTPγS] and in the nucleotide-free state (black). The absorbance at 280 nm was plotted against the elution volume (Ve) normalized to the exclusion volume (V0).

In contrast to hGBP5ta, the hGBP5a/b splice variant does not show a monomer species in the presence of any nucleotide. When bound to GDP or GDP·AlFx, a dimeric form is observed (taking elongated shape into account). The size of the protein complex is only slightly increased when nucleotide-free or in the presence of GTP. As with hGBP5ta, the complex is increased to the size of a tetramer after binding of GppNHp, and binding of GTPγS again shows an intermediate behavior (Fig. 5B). The elution profiles of both hGBP5ta and hGBP5a/b in the presence of various nucleotides show a small fraction of protein complexes eluting at higher apparent masses. This might be an indication that equilibria between various oligomer complexes are established that are controlled by the bound nucleotide. In contrast to hGBP1, our observations suggest that, in the course of hGBP5 catalyzed GTP hydrolysis, a nucleotide-free hGBP5 dimer binds GTP, leading to a larger complex (represented by GppNHp/GTPγS). GTP cleavage leads to dissociation of the protein complex shown by the apparent dimer, resulting in monomeric product complex in the case of hGBP5ta·GDP. In contrast, the hGBP5a/b splice variant remains dimeric. This difference may be caused by the additional amino acids at the C-terminus of hGBP5a/b.

Concluding remarks

We observed biochemical properties for hGBP5 that are strongly different from those of the well-characterized isoform hGBP1. The GTPase activity is much lower, and, in particular, the concentration-dependent activation is not as pronounced. In contrast to hGBP1, we did not observe any GMP binding by hGBP5, and the enzymatic hydrolysis of GTP leads to GDP rather than GMP. The nucleotide affinities of hGBP5 are grossly similar to those of hGBP1, but there are differences in thermodynamic and kinetic details. The entropy changes are more negative in the case of hGBP5, possibly reflecting larger structural changes accompanying nucleotide binding. The rate constants for nucleotide association and dissociation are both much smaller for hGBP5 compared to hGBP1, reflecting a less dynamic behavior but leading to similar binding strength. Intriguingly, all biochemical parameters addressed in this work and summarized above are similar for hGBP5a/b and the splice variant truncated at the C-terminus. At this point, we can only speculate about the origin of the difference in cells, which is not based on different enzymatic activities or different characteristics with respect to nucleotide binding and subsequent formation of oligomers. Rather we assume that lack of the prenylation site in hGBP5ta is responsible for some of the alteration of the biological function.

Experimental procedures

Expression and purification

hGBP5a/b (GenBank accession number AAN39036.1) and hGBP5ta (GenBank accession number AAO40731.1) were subcloned into pQE80L expression vectors (Qiagen, Hilden, Germany), transformed into Escherichia coli Rosetta 2 (DE3) pLysS (Merck, Darmstadt, Germany), grown in TB medium to an attenuance at 600 nm of 0.6, and expression induced by adding 100 μm isopropyl thio-β-d-galactoside (AppliChem, Darmstadt, Germany). The final step of purification, which has been described previously for hGBP1 [8], was size-exclusion chromatography using buffer C (50 mm Tris pH 7.9, 5 mm MgCl2, 2 mm dithiothreitol). Pure protein fractions were concentrated to approximately 1 mm, frozen in liquid nitrogen, and stored at −80 °C. Concentrations of the purified proteins were measured using UV absorbance at 280 nm (ε = 45 380 and 38 880 (m·cm)−1 for hGBP5a/b and hGBP5ta, respectively).

Hydrolysis assays

Hydrolysis measurements were performed as described previously [19] using 350 μm GTP (Sigma-Aldrich, Munich, Germany) and increasing concentrations of hGBP5a/b or hGBP5ta in buffer C containing 50 μm BSA (Sigma-Aldrich) for protein stabilization at 25 °C. Aliquots were taken after defined incubation periods, injected onto a Chromolith RP18e HPLC column (Merck), and elution was followed by determination of absorption at 254 nm using an MD5100plus diode array detector (Jasco, Gross-Umstadt, Germany). The running buffer was composed of 100 mm potassium phosphate, 10 mm tetrabutylammonium bromide, 0.2 mm sodium azide and 1.25% acetonitrile at pH 6.5. Elution times were measured using GMP, GDP and GTP (all purchased from Sigma-Aldrich) as calibration standards. Data were fitted using the Hill-like model shown in Eqn (1) where Smin and Smax represent minimum and maximum specific activity, respectively, and n the Hill coefficient:


Isothermal titration calorimetry (ITC)

All ITC experiments were performed at 25 °C in buffer C using a Microcal AutoITC200 (GE Healthcare, Munich, Germany). The cell was loaded with 150 μm of protein and was titrated against a 3-fold excess of nucleotide by 20 injection steps of 1.8 μL each using 2.25 mm of each nucleotide. All nucleotides used were purchased from Jena Biosciences (Jena, Germany), the concentration was determined by UV absorption at 254 nm (ε = 13 700 (M·cm)−1), and their purity was checked by HPLC analysis (> 98% for all used nucleotides). Data analysis was performed using an ITC-Origin calorimeter (Microcal/GE Healthcare).

Size-exclusion chromatography

Analytical gel filtration experiments were performed using a Superdex 200 10/300 column (GE Healthcare). The elution buffer (50 mm Tris pH 7.9, 5 mm MgCl2, 150 mm NaCl) contained 200 μm of the nucleotide, and 300 μm AlCl3 and 10 mm NaF were added in the case of GDP·AlFx. Protein (20 μm) was preincubated in the elution buffer for 30 min on ice before being injected onto the gel filtration column. Size calibration was carried out using standard proteins with masses in the range of 29–669 kDa (the corresponding elution volumes are indicated on the graphs by arrows). Elution was followed by monitoring the absorbance at 280 nm using an Äkta Purifier system (GE Healthcare).

Fluorescence titration

Fluorescence titrations were performed at 25 °C using an SFM25 fluorospectrometer (Kontron, Zurich, Switzerland) and mant-labeled nucleotides (Jena Biosciences). The excitation and emission wavelengths were 366 and 435 nm, respectively. mant-labeled nucleotide (0.5 μm) was titrated using protein solutions (typically approximately 100 μm) containing 0.5 μm of the mant-labeled nucleotide to avoid dilution of the fluorophore. The data were analysed using a quadratic binding equation as described previously [19].

Stopped flow measurements

All measurements were performed using an SFM400 stopped flow apparatus (Bio-Logic, Claix, France). For association kinetic experiments, 38 μL of a nucleotide solution (0.5 μm) and 38 μL of hGBP5a/b or hGBP5ta (increasing concentrations starting from a 10-fold molar excess) were mixed at 14 mL·s−1. In the case of mant-GTP association, it was necessary to use increasing concentrations of the nucleotide because of nucleotide hydrolysis (0.5 μm of protein was mixed with increasing concentrations of mant-GTP starting from a 15-fold molar excess). Fluorescence was excited at 295 nm (mant-GTP) or 366 nm (all others), and recorded using a 420 nm cut-off filter. For mant-GTP, we used a fluorescence resonance energy transfer approach (primary excitation of the tryptophans in the G-domain and transfer to the mant-group of the nucleotide) to minimize excitation of unbound nucleotide and thereby loss of signal quality. The traces were fitted using a single rate constant, and the resulting rates (kobs) were plotted against the protein or mant-GTP concentration. Using a linear fit, the association rate constants were extracted from the slope (kon), and the intercept represents the dissociation rate (koff). In the case of displacement experiments, 10 μm protein was preincubated with 0.5 μm of mant-nucleotide. The mant-nucleotide was displaced by mixing with a 1000-fold excess of unlabeled GDP, and the resulting rate constant corresponds to kdiss. Fluorescence traces were fitted using a single rate constant (mant-GDP) or two rate constants (mant-GTPγS). Corresponding dissociation constants are calculated from the relationship Kd = koff/kon.


We thank Professor Dr Michael Stürzl for providing the cDNA for hGBP5, the Deutsche Forschungsgemeinschaft for financial support, and the Ruhr-University Research School for a full scholarship to M.W.