Article pubs.acs.org/JPCB
Using Hydrogen−Deuterium Exchange to Monitor Protein Structure in the Presence of Gold Nanoparticles Ailin Wang, Tam Vo,† Vu Le,‡ and Nicholas C. Fitzkee* Department of Chemistry, Mississippi State University, Mississippi State, Mississippi 39762, United States S Supporting Information *
ABSTRACT: The potential applications of protein-functionalized gold nanoparticles (AuNPs) have motivated many studies characterizing protein−AuNP interactions. However, the lack of detailed structural information has hindered our ability to understand the mechanism of protein adsorption on AuNPs. In order to determine the structural perturbations that occur during adsorption, hydrogen/deuterium exchange (HDX) of amide protons was measured for two proteins by NMR. Specifically, we measured both slow (5−300 min) and fast (10−500 ms) H/D exchange rates for GB3 and ubiquitin, two well-characterized proteins. Overall, amide exchange rates are very similar in the presence and absence of AuNPs, supporting a model where the adsorbed protein remains largely folded on the AuNP surface. Small differences in exchange rates are observed for several loop residues, suggesting that the secondary structure remains relatively rigid while loops and surface residues can experience perturbations upon binding. Strikingly, several of these residues are close to lysines, which supports a model where positive surface residues may interact favorably with AuNP-bound citrate. Because these proteins appear to remain folded on AuNP surfaces, these studies suggest that it may be possible to engineer functional AuNP-based nanoconjugates without the use of chemical linkers.
■
INTRODUCTION Substantial research has been devoted to the study of gold nanoparticles (AuNPs) because of their use as a medical diagnostic tool or a drug delivery vector.1−3 A key step is to functionalize AuNPs by coating them with proteins. It is wellknown that protein can spontaneously associate with gold nanoparticles,4 and this may result in unexpected aggregation of either protein5 or nanoparticles. One solution to this problem is to employ chemical linkers that attach proteins to AuNPs.6 An alternative strategy is to rely on the protein’s own affinity for AuNPs, but to do this, one must understand the chemical nature of the protein−AuNP interaction. Additionally, when protein-functionalized AuNPs are placed in biological fluids, other proteins will also compete for binding;7,8 therefore, even in chemically tethered nanoconjugates, protein−AuNP interactions remain an important consideration. This was highlighted in a recent study by Salvati et al., which demonstrated that nanoparticles may lose their functionality in vivo when exposed to blood serum.9 It is thus imperative to understand the mechanism of protein−nanoparticle adsorption. Arguably, understanding of the nature of the protein−AuNP interaction is not possible without knowing the structure of proteins on AuNP surfaces.10 In practice, studying structural changes of AuNP-bound proteins is challenging. While circular dichroism (CD) can be used to study gross structural changes upon adsorption,11 it does not provide atomistic structural detail. Nevertheless, it can be used to monitor large scale structural changes like protein unfolding.12 The prerequisite for development of functionalized protein−nanoparticles is, to a © 2014 American Chemical Society
certain extent, to identify precisely how the interaction with AuNPs may perturb the adsorbed protein structure. The ultimate goal would be to predict whether a particular loop or allosteric site on the protein would be affected by nanoparticle binding. Because of its ability to monitor specific protein regions precisely, NMR spectroscopy could potentially reveal the early steps of protein−nanoparticle adsorption. In particular, NMRbased hydrogen/deuterium exchange (HDX) is a powerful tool for studying protein structure, stability, and dynamics in complex systems undergoing chemical exchange.13,14 Under the appropriate conditions, one can monitor amide protons as they exchange with solvent protons, and this process is very sensitive to structural changes and other environmental factors.15 Experimentally, one can measure HDX by rapid solvent exchange from H2O to D2O or by employing one of several NMR experiments optimized for measuring fast HDX.16−20 An earlier study by Engel et al. measured the hydrogen/deuterium exchange (HDX) rates of bovine αlactalbumin following adsorption onto polystyrene nanospheres.21 This work revealed how adsorption can induce protein conformational changes. At present, however, this powerful tool has not yet been adequately applied to investigations of other nanoparticles. In particular, the surface Special Issue: Spectroscopy of Nano- and Biomaterials Symposium Received: June 30, 2014 Revised: September 29, 2014 Published: September 29, 2014 14148
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
protein signal will dominate, hindering the detection of exchanging proteins. NMR samples were transferred to a 600 MHz Bruker Avance III cryoprobe-equipped NMR spectrometer, pre-equilibrated to 25 °C. The dead time between the addition of D2O buffer and the first NMR measurement was 7 min. A series of 2D 15N−1H HSQC spectra were collected to monitor the exchange reactions at various time points up to 960 min. The 15N acquisition time for each experiment was 25 ms, collected over 50 complex points. The total experiment time for each spectrum was 15 min. HSQC spectra were processed using NMRPipe31 and a series of in-house scripts designed to process a batch data set of 64 spectra. Chemical shift assignments for GB325 and ubiquitin32,33 (Ubq) were obtained from previous studies. For K19C GB3, the assignments were nearly identical to wild-type (WT) GB3, and the backbone chemical shifts for residues 18−20 were determined using standard NOESY-HSQC and TOCSYHSQC experiments.34,35 For real-time HDX measurements, the intensities of each cross peak were normalized by dividing by the peak intensity at the first time point. Then the timedependent behavior of each peak I(t) was modeled by an exponential decay with a baseline offset:
chemistry of AuNPs is very different from that of polymer nanoparticles, and there is no reason to expect that proteins will behave similarly on AuNPs. In this study, NMR was used to measure HDX rates of two proteins in the presence and absence of 15 nm citrate-stabilized AuNPs. Previously, we used NMR to quantify protein−gold nanoparticle adsorption,22 and we found that no significant chemical shift or obvious line broadening occurred in the presence of AuNPs. This suggested that the binding is slow, on the time scale of milliseconds or longer. Additionally, we found evidence that proteins remain globular on the AuNP surface. Here, we use HDX to further explore nanoparticle-induced structural changes. Real-time NMR-based HDX measurements are employed to monitor large structural perturbations to protein secondary structure over the time scale of minutes to hours. Additionally, faster HDX rates are monitored using 15 H/D N SOLEXSY (solvent exchange spectroscopy), which can detect exchange in loops ranging from 0.2 to 2 s−1.19 Since thiol groups are thought to play a significant role in protein adsorption,22−24 we also investigate the behavior a protein containing a surface cysteine residue. In general, we find that HDX rates are largely unperturbed in the presence of AuNPs, strongly suggesting that the proteins used in this study retain their native topology on the nanoparticle surface. Several exceptions to this rule are observed, and while the differences are limited in their statistical significance, the combined results suggest that positively charged residues may favorably interact with the citrate-stabilized AuNPs used in this study. The data presented here reveal that some proteins can retain their structure on AuNPs with little to no deformation. If this finding is general, it may be possible to design protein-functionalized AuNPs without the use of chemical tethers, an attractive yet currently elusive goal.
I(t ) = I0e−kt + b
In this equation, k is the observed hydrogen−deuterium exchange rate, and b is the baseline offset, included to account for the ∼4% H2O concentration introduced by the initial dilution. Each measurement was repeated on three independent samples, and the values for k are reported as the average of these three measurements. Uncertainties in k are given as the sample standard deviation. Fast HDX Measurements (SOLEXSY). For solvent exchange spectroscopy (SOLEXSY), 15N,13C-labeled proteins were prepared in a 50% mixture of H2O and D2O containing 20 mM sodium phosphate at pH* 6.5 (uncorrected). In the absence of AuNPs, 500 μM WT GB3, K19C GB3, and Ubq were used to maximize the observed signal. Measurements were repeated in the presence of 150 nM AuNPs in which the protein concentration was 80 μM (the visible, unbound concentration was ∼40 μM). For samples of K19C, tris(2carboxyethyl)phosphine (TCEP) was added in equimolar amounts to prevent disulfide formation, and the buffer was changed to 20 mM PIPES, pH* 6.5, instead of sodium phosphate. Adding 1 mM TCEP to 150 nM citrate-stabilized AuNPs did not induce aggregation, nor did it shift the plasmonic peak at 520 nm (data not shown). After preparation, all samples were incubated for at least 2 h at room temperature before performing SOLEXSY experiments. Interleaved SOLEXSY experiments were recorded at 600 MHz as described by Chevelkov et al.19 A series of spectra were collected employing the mixing times of 1, 41, 71, 111, 161, 221, 501, and 1001 ms, and the 15N acquisition time was 117 ms over a total of 250 complex points for each mixing time. All other parameters and settings were identical to those described previously.19 The final interleaved experiments were transformed using NMRPipe, and spectra were processed using Sparky.36 Assignments were made based on the most intense peak in the series of time points, and Sparky’s “relaxation fitting” module was used to determine the intensity of a given peak as a function of mixing time. A graphical interface was implemented which takes the output from Sparky and feeds it directly into the original MATLAB fitting scripts developed by Chevelkov et al.,19 substantially simplifying the extraction of
■
EXPERIMENTAL SECTION Protein Samples. 15N and 15N,13C-labeled wild-type and K19C GB3 were expressed and purified as described previously.22,25 Similarly, isotopically labeled ubiquitin (Ubq) was expressed and purified using perchloric acid extraction.26 All protein concentrations was determined by UV absorbance using the calculated extinction coefficient at 280 nm.27 After purification, protein samples were buffer exchanged into water using a HiPrep 26/10 desalting column and subsequently lyophilized. Lyophilized proteins were stored at −80 °C until use. Citrate-Stabilized AuNPs. Gold(III) chloride trihydrate and sodium citrate dihydrate were purchased from SigmaAldrich. Gold nanoparticles (AuNPs) were synthesized by the citric acid reduction method.28 Characterization and electron microscopy of AuNPs has been described previously.22,29 The size of 15 nm was confirmed based on TEM measurements (Supporting Information) as well as UV−visible spectroscopy, where a maximum absorbance was observed at 520 nm.30 Real-Time Measurement of Hydrogen−Deuterium Exchange (HDX). 15N lyophilized protein was resuspended in water to a concentration of 1 mM. The sample was rapidly diluted in D2O to a final concentration of 40 μM in 50 mM sodium phosphate, pH* (uncorrected) 6.5, with and without 100 nM AuNPs. At these concentrations, we expect 50% of protein will be bound to AuNPs.22 This fractional saturation was chosen because it will maximize sensitivity to bound protein: at lower concentrations, it will be difficult to detect any protein signal, and at higher concentrations, the unbound 14149
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
fitting parameters. This interface is available freely on the Internet at http://fitzkee.chemistry.msstate.edu/. As implemented previously,19 the uncertainties of all HDX rates from SOLEXSY measurements is reported as the standard deviation of 20 Monte Carlo simulations, where the uncertainty in each data point is estimated from the NMR spectral noise.
AuNP systems (for an early example, see De Roe et al.4). For the 15 nm AuNPs used in this study, NMR relaxation precludes the direct observation of bound proteins. However, since binding is in equilibrium, structural perturbations can be observed indirectly by monitoring the unbound (free) protein (Figure 1). If protein deformation occurs on the surface of AuNPs, HDX rates will increase, and this will be reflected in a reduction in the NMR signal when the bound protein detaches from the AuNP. Thus, by comparison of HDX rates in the presence and absence of AuNPs, it is possible to monitor structural perturbations in the AuNP-bound, invisible state. One expects different effects on the HDX rates depending on the time scale of the binding: If binding and deformation is slow, on the order of seconds to minutes, HDX rates in this regime will demonstrate a visible effect. On the other hand, if AuNP binding occurs on a faster time scale (∼100 ms), faster HDX rates will be affected as well. Our experiments are designed to probe both time scales of interaction. Placing a fully protonated protein into buffered D2O initiates HDX, and exchange rates can be measured from an exponential fit of each HSQC peak as it decays over time (Figure 2). Two
■
RESULTS AND DISCUSSION In our previous studies, we observed a good agreement between the observed adsorption capacity and the predicted adsorption capacity using the native state radius of gyration.22 This observation suggests that proteins adsorbed onto AuNPs remain globular and compact, although they potentially adopt a non-native topology on the AuNP surface. To test whether native topology was present, we employed a series of hydrogen−deuterium exchange (HDX) measurements. If a protein experiences a structural perturbation on the AuNP surface, otherwise buried residues would become exposed to solvent, accelerating the HDX rates for exposed residues (Figure 1). Regions where the protein is unperturbed should
Figure 1. Hydrogen/deuterium exchange (HDX) rate perturbations in the presence of AuNPs. (a) If the protein structure is unperturbed on the AuNP surface, HDX rates are expected to be similar to those in the absence of AuNPs. (b) If the protein unfolds on the AuNP surface, regions experiencing structural perturbations will be exposed to solvent and have elevated HDX rates compared to the folded protein. Folded regions (green helices) should retain native-like HDX rates.
not experience elevated HDX rates. Thus, by comparing the difference between HDX rates in the presence and absence of AuNPs, we hypothesized that structural changes on the AuNP surface could be monitored. NMR spectroscopy allows amide proton HDX rates to be measured for many residues, allowing us to detect which regions in the protein are most dramatically affected. To increase the generality of our results, two different model proteins were studied: the third IgG-binding domain of streptococcal protein G (GB3)37,38 and human ubiquitin (Ubq).39−41 Additionally, both slow (∼0.1 min−1 or slower) and fast (1−50 s−1) HDX rates were measured, enabling us to probe structural changes over a wide range of time scales. Monitoring Protein Topology with Slow HDX. Amide proton hydrogen−deuterium exchange occurs when local unfolding or solvent penetration results in the chemical exchange from NH to ND.15 GB3 and Ubq are both stable proteins (ΔG̅ 0 ≥ 5 kcal mol−1),31,42 and they each possess many slowly exchanging amide protons.43,44 When proteins are folded in their native topology, backbone protons, particularly those protons in regular secondary structures, will be protected from HDX. If secondary structure is disrupted or if the protein topology is perturbed, the HDX rates will change. Our previous experiments,22 combined with calorimetric studies of protein− AuNP binding,45,46 suggest that AuNP adsorption is reversible for proteins lacking surface cysteine residues. Additionally, equilibrium behavior has been reported for many protein−
Figure 2. Decay of peak intensity for GB3 residues F30 and N37 after initiation of HDX. The behavior of F30 is plotted in free solution (a) and in the presence of AuNPs (b). Similar data are plotted for residue N37, which exchanges at a much faster rate (c, d). Solid lines represent exponential fits to the data. Exchange rates are determined as an average of three independent samples.
typical examples are residues F30 and N37 in WT GB3. F30 decays much more slowly than N37, corresponding to its burial in the middle of the α-helix.43 In the presence of AuNPs, the decay rates for both of these residues are statistically identical (Figure 2b,c), suggesting that these residues are unperturbed on the AuNP surface. The experimental design allows HDX rates to be measured ranging from 0.1 × 10−2 to 0.5 × 10−2 min−1. Residues exchanging faster than this have zero intensity in the first HSQC, whereas slower residues do not decay substantially during each 16 h experiment. In GB3, 27 out of 56 residues fall into this range, and they all fit well to a monoexponential decay. Similar behavior is observed for Ubq, where 30 out of 76 residues are measurable in this range. Comparing all of the slow exchanging residues, after performing each experiment in triplicate, we observe no statistically significant difference between GB3 and Ubq HDX rates in the presence and absence of AuNPs (Figure 3). Experimental conditions were designed so that half of all protein in the sample would be bound to AuNPs. These 14150
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
Figure 3. HDX rates of WT GB3 (a) and ubiquitin (b) with and without AuNPs. The white and gray columns represent the exchange rates of protein in the absence and presence of AuNPs, respectively. Residues with no detectable signal after the dead time (7 min) are not displayed. Error bars represent the standard deviation from three independent measurements. No statistically significant differences are observed for either protein.
structure and stability, it is highly likely that this behavior applies to a large set of globular proteins. Exceptions may exist, and these data are not sufficient to rule out small changes in the fast-exchanging loop regions. For example, DSC studies have shown that AuNPs increase the melting temperature of the protein unfolding transition,50 implying that AuNPs would slow HDX rates. On the other hand, circular dichroism measurements indicate a slight loss of secondary structure, which would increase rates.11 Both of these effects are expected to be most evident for residues near loop regions (5 min−1 > kex > 0.1 min−1), outside the detectable range of our experiments. Thus, while slight perturbations are possible, we expect many proteins will retain their native topology when bound to the AuNP surface. It is known that several proteins show reduced enzymatic activity when bound to AuNPs. This behavior is observed for lysozyme, chymotrypsin, and glucose oxidase, among others.46,51 Neither GB3 nor Ubq contains a native enzymatic function, making it impossible to test whether these proteins are “active” in the traditional sense. However, in light of our data above, we hypothesize that proteins that lose their activity on AuNP surfaces are preferentially oriented with their active site occluded on the AuNP surface. Alternatively, the small perturbations described above may be sufficient to deform the active site, even when it is accessible to solvent. Work is ongoing to differentiate these two possibilities. Using Fast HDX To Monitor Transient Interactions. Our experiments suggest that the adsorbed layer of protein on the AuNP surface exchanges slowly with free protein in solvent (with time scales from seconds to minutes).22 However, it has been proposed that the biocorona consists of multiple protein layers,9,52,53 and these layers may experience faster exchange kinetics. A recent study, which investigated protein binding behavior on polystyrene NPs, found evidence for at least two additional layers of bound protein beyond the initial adsorbed monolayer.54 It is possible that AuNPs exhibit a similar behavior: after initial adsorption, additional layers of protein could interact with those proteins already present on the AuNP surface. Additional layers of weakly bound protein could
conditions were chosen to maximize the effect of binding while still retaining adequate signal for NMR experiments. With the exception of residues 8−24 in GB3, we were able to measure HDX rates for all secondary structure elements in both proteins. Slow exchange is observed for the α-helix and strands β1, β3, and β4, and elevated rates are observed at the interface between secondary structure and loops (e.g., residues 50, 35− 37). The interface between strands β1 and β2 is known to have a slightly smaller protection factor, which may explain why β2 exhibits faster hydrogen exchange in our experiments. In Ubq, a greater number of slow rates are observed, commensurate with its larger unfolding free energy (10 vs 4.5 kcal mol−1).44 However, for both proteins, our data are sufficient to monitor perturbations throughout the entire peptide chain (Supporting Information). In addition to observing no statistically significant difference, we observe no systematic trend in the HDX rates. Some rates are slightly greater in the presence of AuNPs, and some rates are slightly less. This is consistent with random statistical fluctuations about a common mean value for both samples. Importantly, no new residues appeared in our experiments when AuNPs were added. All residues displaying fast behavior in the absence of AuNPs (kex > 0.5 min−1) exhibited similarly fast behavior when AuNPs were present, exchanging in the dead time of the experiment. Thus, while it is impossible to compare loop regions directly using these experiments, AuNPs do not appear to substantially protect loop regions from HDX. Measuring faster HDX rates (see below) may reveal whether loops preferentially bind to the AuNP, thus enabling the determination of protein orientation on the nanoparticle surface. Protein Structure on AuNP Surfaces. These data strongly suggest that for GB3 and Ubq, protein topology (the conformation of helices and strands) is not significantly perturbed upon binding to AuNP surfaces. This finding supports a model where proteins remain globular when bound to AuNPs,22 and it explains several studies where enzymatic activity is preserved upon binding.47−49 Given the usefulness of GB3 and Ubq as model systems for protein 14151
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
experience structural perturbations on a very fast time scale (ms), much faster than could be observed using the experiments described above. To test for this effect, we employed solvent exchange spectroscopy (SOLEXSY)19 to measure differential HDX rates in the presence and absence of AuNPs for ubiquitin and two sequence variants of GB3. SOLEXSY is suitable for the measurement of exchange rates on the order of approximately 0.2−20 s−1. Experiments are performed in a mixture of 50% D2O and 50% H2O, and exchange is measured at equilibrium as proteins de- and reprotonate. Other experiments, such as CLEANEX-PM,18 WEX-2,17 and WEX-3,20 measure HDX indirectly through saturation transfer and therefore can be complicated by fast relaxation and NOEs on the AuNP surface. SOLEXSY, because it measures chemical exchange directly, is largely immune to these effects. If an amide proton exchanges in D2O, this will modulate the observed peak intensity in the SOLEXSY experiment, regardless of conditions at or near the AuNP surface. Therefore, SOLEXSY is ideal for monitoring HDX in solutions of AuNPs. Fast HDX Effects in GB3 and Ubiquitin. SOLEXSY experiments were performed on samples of 50% bound GB3 and Ubq in solution at 25 °C (Figure 4 and Supporting Information). SOLEXSY measures two peaks for each residue, one that originates from a 15N nucleus which is attached to 2H during the exchange mixing time, and one from 15N nuclei attached to 1H during this mixing time.19 These peaks exhibit different chemical shifts based on the deuterium isotope effect. During a series of interleaved experiments, the mixing time is increased, and HDX causes the 1H peak to decay and the 2H peak to increase in intensity. Additionally, T1 relaxation contributes to the decay of both peaks (Figure 4). Fitting the Bloch−McConnell equations to the data yields both the HDX rate as well as an estimate for T1 (Supporting Information). In both GB3 and Ubq, HDX in secondary structures is generally too slow to observe using SOLEXSY. Instead, exchange rates are measured for flexible loops in both proteins. The exception to this rule is β2 in GB3, which is susceptible to HDX at SOLEXSY time scales and therefore serves as a useful probe for measuring the effect of AuNPs on secondary structure. A total of 22 rates (out of 56) were measured in GB3, and 30 rates (out of 76) were measured in Ubq in the absence of AuNPs. Fewer rates are observed in samples containing AuNPs because of the substantially lower concentration of these samples. A lower protein concentration (∼40 μM vs 500 μM without AuNPs) is required because of the relatively low solubility of AuNPs (200 nM).22 A direct comparison of HDX rates in the presence and absence of AuNPs (pH* 6.5) reveals that most residues of WT GB3 behave similarly in the presence and absence of AuNPs (Figure 5a). A correlation plot reveals no significant outliers within 2 standard deviations. A similar conclusion is reached for Ubq (Figure 5b). Thus, both fast and slow HDX rates appear to be fairly constant regardless of the presence of AuNPs. This supports a model where structural perturbations are small as proteins interact with AuNPs. It has been suggested that multiple layers of proteins can bind the surface of certain nanoparticles.54 While we do not see evidence for this on AuNPs,22 if multiple additional layers of protein are forming at a fast time scale, these layers also do not appear to perturb the structure, at least for WT GB3 and Ubq. While most of the residues in GB3 and Ubq are not significantly different, several HDX rates appear to be slower in
Figure 4. Typical SOLEXSY data for WT GB3 (a), ubiquitin (b), and K19C GB3 (c). Two residues are plotted for each protein, represented as triangles and circles. Closed, red symbols are data points plotted from protein in the absence of AuNPs. Open, black symbols are plotted from samples where 50% of protein was bound to AuNPs. Lines represent the best-fit model for the data. Solid lines are fit to the peak originating from proton-attached 15N nuclei, and dashed lines originate from deuterium-attached 15N nuclei. As described in the text, both peaks are recorded simultaneously in a single, interleaved experiment.
the presence of AuNPs. For example, in GB3, T11 has an observed rate of 0.76 ± 0.04 s−1 in the absence of AuNPs, compared to 0.31 ± 0.17 s−1 (p < 0.01) when nanoparticles are present (Figure 4). A similar situation is observed for residues K19 (1.17 ± 0.08 s−1 vs 0.65 ± 0.32 s−1) and V21 (1.69 ± 0.09 s−1 vs 1.19 ± 0.27 s−1), although they are not statistically significant. In protein−AuNP mixtures, there is more uncertainty in HDX rates because of the lower protein concentration; however, all of these residues are clustered around lysine residues. Given the negative charge of the citratestabilized AuNPs, this suggests that these positive lysine residues may preferentially interact with the nanoparticle surface and protect nearby residues from HDX. A similar trend is observed in Ubq, where residue 73 exhibits slower 14152
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
found that MBI can displace the bound WT GB3 but not K19C GB3.22 Presumably, this increased stability is a result of the formation of an Au−S thiolate complex.23 To test the effects of this additional stabilizing interaction on the properties of GB3, we used SOLEXSY to measure fast HDX on K19C GB3 in the presence and absence of AuNPs. All experiments were performed in the presence of tris(2-carboxyethyl)phosphine (TCEP) to reduce disulfide bonds. Unlike 2-mercaptoethanol (BME) and dithiothreitol (DTT), TCEP did not induce aggregation in AuNPs, nor did it shift the plasmonic peak at 520 nm (data not shown). In the absence of AuNPs, the HDX rates of WT GB3 and K19C GB3 compare favorably (Supporting Information). This is in agreement with previous work that indicates the structure of K19C is unperturbed.24 However, several residues in K19C GB3 exhibit slower exchange, including V21. This is likely because the positively charged K19 in WT GB3 increases the effective pH for nearby residues, catalyzing faster exchange.56 When this lysine is substituted with cysteine, the exchange rate would be expected to decrease, as is observed. An initial comparison of exchange rates with and without AuNPs in K19C GB3 revealed that many rates were statistically indistinguishable. This finding is expected, because the cysteine residue should prevent adsorbed proteins from exchanging with free proteins. Given the lack of exchange with solvent, it is difficult to imagine a mechanism by which HDX rates could differ significantly. Importantly, the same protection that was observed for T11 and V21 in WT GB3 was not observed here, suggesting that these residues are not stabilized in the presence of AuNPs. However, several rates appeared to be elevated in AuNP solutions (Figure 5c). These include residues T16 (p < 0.0061), V21 (p < 0.0444), F30 (p < 0.1064), F52 (p < 0.0001), and V54 (p < 0.0209). A close inspection of these residues reveals that they are broadly distributed throughout the structure, and no pattern can be observed with respect to nearby charge distribution or secondary structure. The two most significant residues, T16 and F52, most likely represent a real difference between the two conditions. Of the other residues, F30 is not significantly different, and V21 and V54 are just within the 5% level of confidence. Presently, the differences observed for T16 and F52 remain unexplained. One possibility is the chemical change that occurs upon the formation of Au−S thiolate complexes. It is known that organothiol adsorption to AuNPs can dramatically reduce the pH,23 and even in buffered solutions this may be sufficient to affect protein stability or catalyze disulfide formation, which could affect the exchange rates of these residues on the time scale of our NMR experiments. Even with TCEP in the sample, some disulfide formation did occur during the ∼48 h SOLEXSY experiment, although this fraction was estimated at
Using Hydrogen−Deuterium Exchange to Monitor Protein Structure in the Presence of Gold Nanoparticles Ailin Wang, Tam Vo,† Vu Le,‡ and Nicholas C. Fitzkee* Department of Chemistry, Mississippi State University, Mississippi State, Mississippi 39762, United States S Supporting Information *
ABSTRACT: The potential applications of protein-functionalized gold nanoparticles (AuNPs) have motivated many studies characterizing protein−AuNP interactions. However, the lack of detailed structural information has hindered our ability to understand the mechanism of protein adsorption on AuNPs. In order to determine the structural perturbations that occur during adsorption, hydrogen/deuterium exchange (HDX) of amide protons was measured for two proteins by NMR. Specifically, we measured both slow (5−300 min) and fast (10−500 ms) H/D exchange rates for GB3 and ubiquitin, two well-characterized proteins. Overall, amide exchange rates are very similar in the presence and absence of AuNPs, supporting a model where the adsorbed protein remains largely folded on the AuNP surface. Small differences in exchange rates are observed for several loop residues, suggesting that the secondary structure remains relatively rigid while loops and surface residues can experience perturbations upon binding. Strikingly, several of these residues are close to lysines, which supports a model where positive surface residues may interact favorably with AuNP-bound citrate. Because these proteins appear to remain folded on AuNP surfaces, these studies suggest that it may be possible to engineer functional AuNP-based nanoconjugates without the use of chemical linkers.
■
INTRODUCTION Substantial research has been devoted to the study of gold nanoparticles (AuNPs) because of their use as a medical diagnostic tool or a drug delivery vector.1−3 A key step is to functionalize AuNPs by coating them with proteins. It is wellknown that protein can spontaneously associate with gold nanoparticles,4 and this may result in unexpected aggregation of either protein5 or nanoparticles. One solution to this problem is to employ chemical linkers that attach proteins to AuNPs.6 An alternative strategy is to rely on the protein’s own affinity for AuNPs, but to do this, one must understand the chemical nature of the protein−AuNP interaction. Additionally, when protein-functionalized AuNPs are placed in biological fluids, other proteins will also compete for binding;7,8 therefore, even in chemically tethered nanoconjugates, protein−AuNP interactions remain an important consideration. This was highlighted in a recent study by Salvati et al., which demonstrated that nanoparticles may lose their functionality in vivo when exposed to blood serum.9 It is thus imperative to understand the mechanism of protein−nanoparticle adsorption. Arguably, understanding of the nature of the protein−AuNP interaction is not possible without knowing the structure of proteins on AuNP surfaces.10 In practice, studying structural changes of AuNP-bound proteins is challenging. While circular dichroism (CD) can be used to study gross structural changes upon adsorption,11 it does not provide atomistic structural detail. Nevertheless, it can be used to monitor large scale structural changes like protein unfolding.12 The prerequisite for development of functionalized protein−nanoparticles is, to a © 2014 American Chemical Society
certain extent, to identify precisely how the interaction with AuNPs may perturb the adsorbed protein structure. The ultimate goal would be to predict whether a particular loop or allosteric site on the protein would be affected by nanoparticle binding. Because of its ability to monitor specific protein regions precisely, NMR spectroscopy could potentially reveal the early steps of protein−nanoparticle adsorption. In particular, NMRbased hydrogen/deuterium exchange (HDX) is a powerful tool for studying protein structure, stability, and dynamics in complex systems undergoing chemical exchange.13,14 Under the appropriate conditions, one can monitor amide protons as they exchange with solvent protons, and this process is very sensitive to structural changes and other environmental factors.15 Experimentally, one can measure HDX by rapid solvent exchange from H2O to D2O or by employing one of several NMR experiments optimized for measuring fast HDX.16−20 An earlier study by Engel et al. measured the hydrogen/deuterium exchange (HDX) rates of bovine αlactalbumin following adsorption onto polystyrene nanospheres.21 This work revealed how adsorption can induce protein conformational changes. At present, however, this powerful tool has not yet been adequately applied to investigations of other nanoparticles. In particular, the surface Special Issue: Spectroscopy of Nano- and Biomaterials Symposium Received: June 30, 2014 Revised: September 29, 2014 Published: September 29, 2014 14148
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
protein signal will dominate, hindering the detection of exchanging proteins. NMR samples were transferred to a 600 MHz Bruker Avance III cryoprobe-equipped NMR spectrometer, pre-equilibrated to 25 °C. The dead time between the addition of D2O buffer and the first NMR measurement was 7 min. A series of 2D 15N−1H HSQC spectra were collected to monitor the exchange reactions at various time points up to 960 min. The 15N acquisition time for each experiment was 25 ms, collected over 50 complex points. The total experiment time for each spectrum was 15 min. HSQC spectra were processed using NMRPipe31 and a series of in-house scripts designed to process a batch data set of 64 spectra. Chemical shift assignments for GB325 and ubiquitin32,33 (Ubq) were obtained from previous studies. For K19C GB3, the assignments were nearly identical to wild-type (WT) GB3, and the backbone chemical shifts for residues 18−20 were determined using standard NOESY-HSQC and TOCSYHSQC experiments.34,35 For real-time HDX measurements, the intensities of each cross peak were normalized by dividing by the peak intensity at the first time point. Then the timedependent behavior of each peak I(t) was modeled by an exponential decay with a baseline offset:
chemistry of AuNPs is very different from that of polymer nanoparticles, and there is no reason to expect that proteins will behave similarly on AuNPs. In this study, NMR was used to measure HDX rates of two proteins in the presence and absence of 15 nm citrate-stabilized AuNPs. Previously, we used NMR to quantify protein−gold nanoparticle adsorption,22 and we found that no significant chemical shift or obvious line broadening occurred in the presence of AuNPs. This suggested that the binding is slow, on the time scale of milliseconds or longer. Additionally, we found evidence that proteins remain globular on the AuNP surface. Here, we use HDX to further explore nanoparticle-induced structural changes. Real-time NMR-based HDX measurements are employed to monitor large structural perturbations to protein secondary structure over the time scale of minutes to hours. Additionally, faster HDX rates are monitored using 15 H/D N SOLEXSY (solvent exchange spectroscopy), which can detect exchange in loops ranging from 0.2 to 2 s−1.19 Since thiol groups are thought to play a significant role in protein adsorption,22−24 we also investigate the behavior a protein containing a surface cysteine residue. In general, we find that HDX rates are largely unperturbed in the presence of AuNPs, strongly suggesting that the proteins used in this study retain their native topology on the nanoparticle surface. Several exceptions to this rule are observed, and while the differences are limited in their statistical significance, the combined results suggest that positively charged residues may favorably interact with the citrate-stabilized AuNPs used in this study. The data presented here reveal that some proteins can retain their structure on AuNPs with little to no deformation. If this finding is general, it may be possible to design protein-functionalized AuNPs without the use of chemical tethers, an attractive yet currently elusive goal.
I(t ) = I0e−kt + b
In this equation, k is the observed hydrogen−deuterium exchange rate, and b is the baseline offset, included to account for the ∼4% H2O concentration introduced by the initial dilution. Each measurement was repeated on three independent samples, and the values for k are reported as the average of these three measurements. Uncertainties in k are given as the sample standard deviation. Fast HDX Measurements (SOLEXSY). For solvent exchange spectroscopy (SOLEXSY), 15N,13C-labeled proteins were prepared in a 50% mixture of H2O and D2O containing 20 mM sodium phosphate at pH* 6.5 (uncorrected). In the absence of AuNPs, 500 μM WT GB3, K19C GB3, and Ubq were used to maximize the observed signal. Measurements were repeated in the presence of 150 nM AuNPs in which the protein concentration was 80 μM (the visible, unbound concentration was ∼40 μM). For samples of K19C, tris(2carboxyethyl)phosphine (TCEP) was added in equimolar amounts to prevent disulfide formation, and the buffer was changed to 20 mM PIPES, pH* 6.5, instead of sodium phosphate. Adding 1 mM TCEP to 150 nM citrate-stabilized AuNPs did not induce aggregation, nor did it shift the plasmonic peak at 520 nm (data not shown). After preparation, all samples were incubated for at least 2 h at room temperature before performing SOLEXSY experiments. Interleaved SOLEXSY experiments were recorded at 600 MHz as described by Chevelkov et al.19 A series of spectra were collected employing the mixing times of 1, 41, 71, 111, 161, 221, 501, and 1001 ms, and the 15N acquisition time was 117 ms over a total of 250 complex points for each mixing time. All other parameters and settings were identical to those described previously.19 The final interleaved experiments were transformed using NMRPipe, and spectra were processed using Sparky.36 Assignments were made based on the most intense peak in the series of time points, and Sparky’s “relaxation fitting” module was used to determine the intensity of a given peak as a function of mixing time. A graphical interface was implemented which takes the output from Sparky and feeds it directly into the original MATLAB fitting scripts developed by Chevelkov et al.,19 substantially simplifying the extraction of
■
EXPERIMENTAL SECTION Protein Samples. 15N and 15N,13C-labeled wild-type and K19C GB3 were expressed and purified as described previously.22,25 Similarly, isotopically labeled ubiquitin (Ubq) was expressed and purified using perchloric acid extraction.26 All protein concentrations was determined by UV absorbance using the calculated extinction coefficient at 280 nm.27 After purification, protein samples were buffer exchanged into water using a HiPrep 26/10 desalting column and subsequently lyophilized. Lyophilized proteins were stored at −80 °C until use. Citrate-Stabilized AuNPs. Gold(III) chloride trihydrate and sodium citrate dihydrate were purchased from SigmaAldrich. Gold nanoparticles (AuNPs) were synthesized by the citric acid reduction method.28 Characterization and electron microscopy of AuNPs has been described previously.22,29 The size of 15 nm was confirmed based on TEM measurements (Supporting Information) as well as UV−visible spectroscopy, where a maximum absorbance was observed at 520 nm.30 Real-Time Measurement of Hydrogen−Deuterium Exchange (HDX). 15N lyophilized protein was resuspended in water to a concentration of 1 mM. The sample was rapidly diluted in D2O to a final concentration of 40 μM in 50 mM sodium phosphate, pH* (uncorrected) 6.5, with and without 100 nM AuNPs. At these concentrations, we expect 50% of protein will be bound to AuNPs.22 This fractional saturation was chosen because it will maximize sensitivity to bound protein: at lower concentrations, it will be difficult to detect any protein signal, and at higher concentrations, the unbound 14149
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
fitting parameters. This interface is available freely on the Internet at http://fitzkee.chemistry.msstate.edu/. As implemented previously,19 the uncertainties of all HDX rates from SOLEXSY measurements is reported as the standard deviation of 20 Monte Carlo simulations, where the uncertainty in each data point is estimated from the NMR spectral noise.
AuNP systems (for an early example, see De Roe et al.4). For the 15 nm AuNPs used in this study, NMR relaxation precludes the direct observation of bound proteins. However, since binding is in equilibrium, structural perturbations can be observed indirectly by monitoring the unbound (free) protein (Figure 1). If protein deformation occurs on the surface of AuNPs, HDX rates will increase, and this will be reflected in a reduction in the NMR signal when the bound protein detaches from the AuNP. Thus, by comparison of HDX rates in the presence and absence of AuNPs, it is possible to monitor structural perturbations in the AuNP-bound, invisible state. One expects different effects on the HDX rates depending on the time scale of the binding: If binding and deformation is slow, on the order of seconds to minutes, HDX rates in this regime will demonstrate a visible effect. On the other hand, if AuNP binding occurs on a faster time scale (∼100 ms), faster HDX rates will be affected as well. Our experiments are designed to probe both time scales of interaction. Placing a fully protonated protein into buffered D2O initiates HDX, and exchange rates can be measured from an exponential fit of each HSQC peak as it decays over time (Figure 2). Two
■
RESULTS AND DISCUSSION In our previous studies, we observed a good agreement between the observed adsorption capacity and the predicted adsorption capacity using the native state radius of gyration.22 This observation suggests that proteins adsorbed onto AuNPs remain globular and compact, although they potentially adopt a non-native topology on the AuNP surface. To test whether native topology was present, we employed a series of hydrogen−deuterium exchange (HDX) measurements. If a protein experiences a structural perturbation on the AuNP surface, otherwise buried residues would become exposed to solvent, accelerating the HDX rates for exposed residues (Figure 1). Regions where the protein is unperturbed should
Figure 1. Hydrogen/deuterium exchange (HDX) rate perturbations in the presence of AuNPs. (a) If the protein structure is unperturbed on the AuNP surface, HDX rates are expected to be similar to those in the absence of AuNPs. (b) If the protein unfolds on the AuNP surface, regions experiencing structural perturbations will be exposed to solvent and have elevated HDX rates compared to the folded protein. Folded regions (green helices) should retain native-like HDX rates.
not experience elevated HDX rates. Thus, by comparing the difference between HDX rates in the presence and absence of AuNPs, we hypothesized that structural changes on the AuNP surface could be monitored. NMR spectroscopy allows amide proton HDX rates to be measured for many residues, allowing us to detect which regions in the protein are most dramatically affected. To increase the generality of our results, two different model proteins were studied: the third IgG-binding domain of streptococcal protein G (GB3)37,38 and human ubiquitin (Ubq).39−41 Additionally, both slow (∼0.1 min−1 or slower) and fast (1−50 s−1) HDX rates were measured, enabling us to probe structural changes over a wide range of time scales. Monitoring Protein Topology with Slow HDX. Amide proton hydrogen−deuterium exchange occurs when local unfolding or solvent penetration results in the chemical exchange from NH to ND.15 GB3 and Ubq are both stable proteins (ΔG̅ 0 ≥ 5 kcal mol−1),31,42 and they each possess many slowly exchanging amide protons.43,44 When proteins are folded in their native topology, backbone protons, particularly those protons in regular secondary structures, will be protected from HDX. If secondary structure is disrupted or if the protein topology is perturbed, the HDX rates will change. Our previous experiments,22 combined with calorimetric studies of protein− AuNP binding,45,46 suggest that AuNP adsorption is reversible for proteins lacking surface cysteine residues. Additionally, equilibrium behavior has been reported for many protein−
Figure 2. Decay of peak intensity for GB3 residues F30 and N37 after initiation of HDX. The behavior of F30 is plotted in free solution (a) and in the presence of AuNPs (b). Similar data are plotted for residue N37, which exchanges at a much faster rate (c, d). Solid lines represent exponential fits to the data. Exchange rates are determined as an average of three independent samples.
typical examples are residues F30 and N37 in WT GB3. F30 decays much more slowly than N37, corresponding to its burial in the middle of the α-helix.43 In the presence of AuNPs, the decay rates for both of these residues are statistically identical (Figure 2b,c), suggesting that these residues are unperturbed on the AuNP surface. The experimental design allows HDX rates to be measured ranging from 0.1 × 10−2 to 0.5 × 10−2 min−1. Residues exchanging faster than this have zero intensity in the first HSQC, whereas slower residues do not decay substantially during each 16 h experiment. In GB3, 27 out of 56 residues fall into this range, and they all fit well to a monoexponential decay. Similar behavior is observed for Ubq, where 30 out of 76 residues are measurable in this range. Comparing all of the slow exchanging residues, after performing each experiment in triplicate, we observe no statistically significant difference between GB3 and Ubq HDX rates in the presence and absence of AuNPs (Figure 3). Experimental conditions were designed so that half of all protein in the sample would be bound to AuNPs. These 14150
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
Figure 3. HDX rates of WT GB3 (a) and ubiquitin (b) with and without AuNPs. The white and gray columns represent the exchange rates of protein in the absence and presence of AuNPs, respectively. Residues with no detectable signal after the dead time (7 min) are not displayed. Error bars represent the standard deviation from three independent measurements. No statistically significant differences are observed for either protein.
structure and stability, it is highly likely that this behavior applies to a large set of globular proteins. Exceptions may exist, and these data are not sufficient to rule out small changes in the fast-exchanging loop regions. For example, DSC studies have shown that AuNPs increase the melting temperature of the protein unfolding transition,50 implying that AuNPs would slow HDX rates. On the other hand, circular dichroism measurements indicate a slight loss of secondary structure, which would increase rates.11 Both of these effects are expected to be most evident for residues near loop regions (5 min−1 > kex > 0.1 min−1), outside the detectable range of our experiments. Thus, while slight perturbations are possible, we expect many proteins will retain their native topology when bound to the AuNP surface. It is known that several proteins show reduced enzymatic activity when bound to AuNPs. This behavior is observed for lysozyme, chymotrypsin, and glucose oxidase, among others.46,51 Neither GB3 nor Ubq contains a native enzymatic function, making it impossible to test whether these proteins are “active” in the traditional sense. However, in light of our data above, we hypothesize that proteins that lose their activity on AuNP surfaces are preferentially oriented with their active site occluded on the AuNP surface. Alternatively, the small perturbations described above may be sufficient to deform the active site, even when it is accessible to solvent. Work is ongoing to differentiate these two possibilities. Using Fast HDX To Monitor Transient Interactions. Our experiments suggest that the adsorbed layer of protein on the AuNP surface exchanges slowly with free protein in solvent (with time scales from seconds to minutes).22 However, it has been proposed that the biocorona consists of multiple protein layers,9,52,53 and these layers may experience faster exchange kinetics. A recent study, which investigated protein binding behavior on polystyrene NPs, found evidence for at least two additional layers of bound protein beyond the initial adsorbed monolayer.54 It is possible that AuNPs exhibit a similar behavior: after initial adsorption, additional layers of protein could interact with those proteins already present on the AuNP surface. Additional layers of weakly bound protein could
conditions were chosen to maximize the effect of binding while still retaining adequate signal for NMR experiments. With the exception of residues 8−24 in GB3, we were able to measure HDX rates for all secondary structure elements in both proteins. Slow exchange is observed for the α-helix and strands β1, β3, and β4, and elevated rates are observed at the interface between secondary structure and loops (e.g., residues 50, 35− 37). The interface between strands β1 and β2 is known to have a slightly smaller protection factor, which may explain why β2 exhibits faster hydrogen exchange in our experiments. In Ubq, a greater number of slow rates are observed, commensurate with its larger unfolding free energy (10 vs 4.5 kcal mol−1).44 However, for both proteins, our data are sufficient to monitor perturbations throughout the entire peptide chain (Supporting Information). In addition to observing no statistically significant difference, we observe no systematic trend in the HDX rates. Some rates are slightly greater in the presence of AuNPs, and some rates are slightly less. This is consistent with random statistical fluctuations about a common mean value for both samples. Importantly, no new residues appeared in our experiments when AuNPs were added. All residues displaying fast behavior in the absence of AuNPs (kex > 0.5 min−1) exhibited similarly fast behavior when AuNPs were present, exchanging in the dead time of the experiment. Thus, while it is impossible to compare loop regions directly using these experiments, AuNPs do not appear to substantially protect loop regions from HDX. Measuring faster HDX rates (see below) may reveal whether loops preferentially bind to the AuNP, thus enabling the determination of protein orientation on the nanoparticle surface. Protein Structure on AuNP Surfaces. These data strongly suggest that for GB3 and Ubq, protein topology (the conformation of helices and strands) is not significantly perturbed upon binding to AuNP surfaces. This finding supports a model where proteins remain globular when bound to AuNPs,22 and it explains several studies where enzymatic activity is preserved upon binding.47−49 Given the usefulness of GB3 and Ubq as model systems for protein 14151
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
experience structural perturbations on a very fast time scale (ms), much faster than could be observed using the experiments described above. To test for this effect, we employed solvent exchange spectroscopy (SOLEXSY)19 to measure differential HDX rates in the presence and absence of AuNPs for ubiquitin and two sequence variants of GB3. SOLEXSY is suitable for the measurement of exchange rates on the order of approximately 0.2−20 s−1. Experiments are performed in a mixture of 50% D2O and 50% H2O, and exchange is measured at equilibrium as proteins de- and reprotonate. Other experiments, such as CLEANEX-PM,18 WEX-2,17 and WEX-3,20 measure HDX indirectly through saturation transfer and therefore can be complicated by fast relaxation and NOEs on the AuNP surface. SOLEXSY, because it measures chemical exchange directly, is largely immune to these effects. If an amide proton exchanges in D2O, this will modulate the observed peak intensity in the SOLEXSY experiment, regardless of conditions at or near the AuNP surface. Therefore, SOLEXSY is ideal for monitoring HDX in solutions of AuNPs. Fast HDX Effects in GB3 and Ubiquitin. SOLEXSY experiments were performed on samples of 50% bound GB3 and Ubq in solution at 25 °C (Figure 4 and Supporting Information). SOLEXSY measures two peaks for each residue, one that originates from a 15N nucleus which is attached to 2H during the exchange mixing time, and one from 15N nuclei attached to 1H during this mixing time.19 These peaks exhibit different chemical shifts based on the deuterium isotope effect. During a series of interleaved experiments, the mixing time is increased, and HDX causes the 1H peak to decay and the 2H peak to increase in intensity. Additionally, T1 relaxation contributes to the decay of both peaks (Figure 4). Fitting the Bloch−McConnell equations to the data yields both the HDX rate as well as an estimate for T1 (Supporting Information). In both GB3 and Ubq, HDX in secondary structures is generally too slow to observe using SOLEXSY. Instead, exchange rates are measured for flexible loops in both proteins. The exception to this rule is β2 in GB3, which is susceptible to HDX at SOLEXSY time scales and therefore serves as a useful probe for measuring the effect of AuNPs on secondary structure. A total of 22 rates (out of 56) were measured in GB3, and 30 rates (out of 76) were measured in Ubq in the absence of AuNPs. Fewer rates are observed in samples containing AuNPs because of the substantially lower concentration of these samples. A lower protein concentration (∼40 μM vs 500 μM without AuNPs) is required because of the relatively low solubility of AuNPs (200 nM).22 A direct comparison of HDX rates in the presence and absence of AuNPs (pH* 6.5) reveals that most residues of WT GB3 behave similarly in the presence and absence of AuNPs (Figure 5a). A correlation plot reveals no significant outliers within 2 standard deviations. A similar conclusion is reached for Ubq (Figure 5b). Thus, both fast and slow HDX rates appear to be fairly constant regardless of the presence of AuNPs. This supports a model where structural perturbations are small as proteins interact with AuNPs. It has been suggested that multiple layers of proteins can bind the surface of certain nanoparticles.54 While we do not see evidence for this on AuNPs,22 if multiple additional layers of protein are forming at a fast time scale, these layers also do not appear to perturb the structure, at least for WT GB3 and Ubq. While most of the residues in GB3 and Ubq are not significantly different, several HDX rates appear to be slower in
Figure 4. Typical SOLEXSY data for WT GB3 (a), ubiquitin (b), and K19C GB3 (c). Two residues are plotted for each protein, represented as triangles and circles. Closed, red symbols are data points plotted from protein in the absence of AuNPs. Open, black symbols are plotted from samples where 50% of protein was bound to AuNPs. Lines represent the best-fit model for the data. Solid lines are fit to the peak originating from proton-attached 15N nuclei, and dashed lines originate from deuterium-attached 15N nuclei. As described in the text, both peaks are recorded simultaneously in a single, interleaved experiment.
the presence of AuNPs. For example, in GB3, T11 has an observed rate of 0.76 ± 0.04 s−1 in the absence of AuNPs, compared to 0.31 ± 0.17 s−1 (p < 0.01) when nanoparticles are present (Figure 4). A similar situation is observed for residues K19 (1.17 ± 0.08 s−1 vs 0.65 ± 0.32 s−1) and V21 (1.69 ± 0.09 s−1 vs 1.19 ± 0.27 s−1), although they are not statistically significant. In protein−AuNP mixtures, there is more uncertainty in HDX rates because of the lower protein concentration; however, all of these residues are clustered around lysine residues. Given the negative charge of the citratestabilized AuNPs, this suggests that these positive lysine residues may preferentially interact with the nanoparticle surface and protect nearby residues from HDX. A similar trend is observed in Ubq, where residue 73 exhibits slower 14152
dx.doi.org/10.1021/jp506506p | J. Phys. Chem. B 2014, 118, 14148−14156
The Journal of Physical Chemistry B
Article
found that MBI can displace the bound WT GB3 but not K19C GB3.22 Presumably, this increased stability is a result of the formation of an Au−S thiolate complex.23 To test the effects of this additional stabilizing interaction on the properties of GB3, we used SOLEXSY to measure fast HDX on K19C GB3 in the presence and absence of AuNPs. All experiments were performed in the presence of tris(2-carboxyethyl)phosphine (TCEP) to reduce disulfide bonds. Unlike 2-mercaptoethanol (BME) and dithiothreitol (DTT), TCEP did not induce aggregation in AuNPs, nor did it shift the plasmonic peak at 520 nm (data not shown). In the absence of AuNPs, the HDX rates of WT GB3 and K19C GB3 compare favorably (Supporting Information). This is in agreement with previous work that indicates the structure of K19C is unperturbed.24 However, several residues in K19C GB3 exhibit slower exchange, including V21. This is likely because the positively charged K19 in WT GB3 increases the effective pH for nearby residues, catalyzing faster exchange.56 When this lysine is substituted with cysteine, the exchange rate would be expected to decrease, as is observed. An initial comparison of exchange rates with and without AuNPs in K19C GB3 revealed that many rates were statistically indistinguishable. This finding is expected, because the cysteine residue should prevent adsorbed proteins from exchanging with free proteins. Given the lack of exchange with solvent, it is difficult to imagine a mechanism by which HDX rates could differ significantly. Importantly, the same protection that was observed for T11 and V21 in WT GB3 was not observed here, suggesting that these residues are not stabilized in the presence of AuNPs. However, several rates appeared to be elevated in AuNP solutions (Figure 5c). These include residues T16 (p < 0.0061), V21 (p < 0.0444), F30 (p < 0.1064), F52 (p < 0.0001), and V54 (p < 0.0209). A close inspection of these residues reveals that they are broadly distributed throughout the structure, and no pattern can be observed with respect to nearby charge distribution or secondary structure. The two most significant residues, T16 and F52, most likely represent a real difference between the two conditions. Of the other residues, F30 is not significantly different, and V21 and V54 are just within the 5% level of confidence. Presently, the differences observed for T16 and F52 remain unexplained. One possibility is the chemical change that occurs upon the formation of Au−S thiolate complexes. It is known that organothiol adsorption to AuNPs can dramatically reduce the pH,23 and even in buffered solutions this may be sufficient to affect protein stability or catalyze disulfide formation, which could affect the exchange rates of these residues on the time scale of our NMR experiments. Even with TCEP in the sample, some disulfide formation did occur during the ∼48 h SOLEXSY experiment, although this fraction was estimated at
