Since its first applications published about 10 years ago, the combination of hydrogen exchange and mass spectrometry has been widely used in structural biology, providing views on protein structure and protein dynamics.1,2 For example, this method can probe the influence of mutations on the three-dimensional (3-D) structures of proteins,3 the conformational changes induced by ligand binding,4,5 the interactions in protein/protein6 or in peptide/membrane7 complexes, or the folding and unfolding of proteins.8,9
Under the usual conditions employed, only amide hydrogen exchange is observable. Exchangeable hydrogen atoms located on amino acid side chains, as well as those on the N- or C-terminus, have exchange rates which cannot be measured by most techniques. The exchange rates of amide hydrogens are highly dependent on their implication in secondary structures and on the solvent accessibility. Protons involved in hydrogen bonds in α-helices, in β-sheets, or present in the core of the protein, have lower rates of exchange than protons in unstructured regions or at the surface of the protein. The deuterium incorporation can be quantified by several methods (NMR and IR spectroscopy); however, mass spectrometry offers several advantages such as the limited amount of sample necessary, the speed of the analyses, and the possibility to study larger proteins.
The method used to determine the deuterium incorporation in a small region of a protein consists of incubating the protein in a deuterated solvent, digesting the protein with a protease, separating the resulting peptides by liquid chromatography, and analyzing on-line by mass spectrometry.10 A key aspect of this method is the quenching of isotope back-exchange because digestion, separation and mass spectrometric analyses are realized in hydrogenated solvents and in denaturating conditions. The rate constant for hydrogen exchange (kex) can be expressed in terms of kH and kOH, which are the rate constants for acid- and base-catalyzed exchange:
Studies on peptide models show that kOH is 108-fold higher than kH, giving, at pH > 3, a linear dependence of kex on pH, with a minimum exchange rate occurring at pH 2–3.11,12 After deuteration realized at neutral pH, decreasing the pH to 2.5 reduces the amide hydrogen exchange rate by four orders of magnitude. In addition to the pH sensitivity, the amide hydrogen exchange rate depends on the temperature. The exchange rate decreases 3-fold for each 10°C.12,13
For these reasons, proteases working at low pH and low temperature must be used. Conventionally, pepsin is used because the activity of this protease is maximum for a pH between 2 and 4 and it can still digest proteins at low temperature. However, with large proteins sequence coverage is not always complete, and the spatial resolution of the exchange rate is limited by the sizes of the resulting peptides (typically 5–30 amino acids). In this work two other proteases were found to work in conditions compatible with the deuteration studies, namely, protease type XIII from Aspergillus saitoi and protease type XVIII from Rhizhopus species.
As a model, a 77-kDa protein was selected, the penicillin-binding protein X, PBP-2X*; it is the primary target for β-lactam antibiotics and is involved in the resistance to these antibiotics. The 3-D structure of PBP-2X has been solved using X-ray crystallography,14,15 and therefore we could analyze our results and validate our method. In this study we determined the most accessible zones of the protein by the mass spectrometric measurement of the local isotope exchange. We show that the combination of three proteases increases the sequence coverage and the spatial resolution. To the best of our knowledge this is the first time that three different proteases have been used for H/D exchange experiments.
The water-soluble PBP-2X* (corresponding to amino acids 49–750) from Streptococcus pneumoniae was expressed as a fusion protein to glutathione transferase in a host strain of Escherichia coli and purified by affinity chromatography.16 The protein was in a 5 mM phosphate buffer (pH 6.8) at a final concentration of 2.1 mg/mL.
The protein was digested for 2 min at 0°C by three different proteases. Pepsin (Sigma Aldrich), protease type XIII from Aspergillus saitoi (Sigma Aldrich), and protease type XVIII from Rhizhopus species (Sigma Aldrich) were used at enzyme/PBP-2X* ratios (w/w) of 1, 10.5 and 17, respectively. To determine the optimum enzyme/PBP-2X* ratio to be used, we observed by electrophoresis the disappearance of the PBP-2X* at different concentrations of enzyme. In order to limit autolysis we chose for each protease the lowest concentration necessary for the complete digestion of the PBP-2X* in 2 min. Protease solutions were prepared in 0.11 M H3PO4 (pH 1.6).
PBP-2X* was separately digested as described above by each of the three enzymes. Peptides were separated by high-performance liquid chromatography (HPLC) on a C18 reversed-phase column (1 mm × 100 mm, Interchrom). Phase A was an 0.1% aqueous solution of TFA (Sigma Aldrich), and phase B was a 90:10:0.1 acetonitrile (SDS)/water/TFA mixture. The sample was first desalted on the column with 2% phase B at 50 μL/min. A 2–65% step gradient of phase B was used during the initial 45 min, then an increase to 100% phase B for 15 min, and finally a hold at 100% for 10 min. The HPLC system was run at 50 μL/min. Peptide assignment was performed by direct MS/MS analysis by splitting half of the flow to the mass spectrometer.
H/D exchange experiments on peptides
PBP-2X* (4 μL of a 2.1 mg/mL solution) was diluted 20-fold in 5 mM deuterated phosphate buffer (pD 6.8) made with Na2HPO4 and NaH2PO4 (Sigma Aldrich), dissolved in D2O (Sigma Aldrich). Isotope exchange was performed for 30 s at room temperature in order to exchange only the most accessible amide hydrogens of the protein. The addition of 9 μL of protease solution (pH 1.6) decreased the pD to 2.5, quenching the isotope back-exchange. Digestion was performed for 2 min at 0°C as described above. The sample was preconcentrated and desalted on a peptide MacroTrap column (Michrom Bioresources) with 2% phase B at a flow rate of 300 μL/min, prior to eluting on a C18 reversed-phase column (1 mm × 100 mm, Interchrom). The same separation conditions as described above were then used. To prevent back-exchange, buffers, valves and columns were cooled to 0°C. Direct measurement molecular masses of deuterated peptides were obtained without splitting. We performed in parallel the same experiments with a non-deuterated buffer, and could thus compare the retention times of the non-deuterated and deuterated peptides, which helped to identify the latter.
ESI-MS and ESI-MS/MS
Mass spectrometric analyses were performed using a quadrupole ion trap mass spectrometer (ESQUIRE 3000+, Bruker Daltonics) equipped with an ionspray source. The heated capillary was held at 4 kV, and end plate voltage set at 500 V. A 10 psi back-pressure of nitrogen sheath gas was used to stabilize the spray, and the drying gas flow of N2 was set at 8 L/min with a temperature of 250°C. For the MS/MS experiments the three most intense ions were fragmented and then excluded after one spectrum had been obtained. The isolation width was 4 Th and the fragmentation amplitude was set to 2 V. Mass spectra were acquired from m/z 50–2000. For analyses and treatment of the data we used the Bruker Daltonics softwares Data Analysis 3.0 and Biotools 2.1.
Peptide mapping experiments were performed for each enzyme by LC/MS/MS analyses. After the LC/MS/MS analysis, the Data Analysis software generated a compound list of automatically identified possible peptides. Within these possible peptides, some were selected with the Biotools software when the protein sequence was introduced; this gave a theoretical peptide mapping. To validate each peptide, we checked manually each corresponding MS/MS spectrum. Only peptides that were unambiguously identified were retained for the peptide mapping.
The sequences of the peptides obtained with each protease are shown in Fig. 1 as underlining bars under the PBP-2X* sequence. After separate digestions with pepsin, protease type XIII and protease type XVIII, we identified 149, 69 and 144 peptides, respectively, giving sequence coverages for PBP-2X* of 93, 40 and 84%, respectively. None of these proteases is really specific, but we found a very good reproducibility in the digestion patterns. Percentages of coverage were very different between the three enzymes, which might be due to their different activities and might also depend on the model protein which was digested. The use of only one enzyme did not provide complete coverage of the sequence. In contrast, if we combine the three enzymes, it is possible to cover 99.7% of the sequence. In addition, this combination gives several overlapping peptides which can be used to obtain higher resolution.
Isotope exchange on the peptides
For each enzyme, LC/MS experiments were realized in parallel with PBP-2X* in both deuterated and non-deuterated solvent. Elution profiles of peptides obtained from the digestion of non-deuterated protein were the same as those obtained from the digestion of deuterated protein. For each peptide previously identified we measured the extent of deuteration. The deuterated peptides obtained with each protease are shown in Fig. 2 as bars with different gray levels, according to their percentages of deuteration. After digestion with pepsin, protease type XIII and protease type XVIII, we could determine the deuterium incorporation for 84, 16 and 115 peptides, respectively, giving sequence coverages for deuterated PBP-2X* of 72, 12.5 and 81%, respectively. It was not possible to determine the deuterium incorporation for all the peptides identified by LC/MS/MS analyses. The signals for deuterated peptides are in general lower than those of the corresponding non-deuterated peptides because of the non-homogeneous deuteration, which creates several species for a given peptide. In addition, deuteration can lead to overlapping of mass peaks for co-eluting peptides. For these reasons, with pepsin alone we obtained information on only 72% of the amino acids, leaving the possibility of missing a major area of the protein in the remaining 28%. However, by combining the three analyses it was possible to obtain information about deuterium incorporation on almost 95% of the amide hydrogens. Furthermore, this combination provided overlapping peptides, enabling the calculation of the numbers of incorporated deuterium atoms for very small segments. By combining the results presented in Fig. 2 it is possible to determine the percentage of deuteration for each region of the protein; these data are presented in Fig. 3.
In particular, eight segments with a percentage of deuteration greater than 30% are identified, namely, fragments 49–63, 103–110, 153–168, 370–389, 492–501, 550–567, 616–624 and 634–645. These fragments correspond to the most easily exchanged regions of PBP-2X*.
The use of the three enzymes
To our knowledge this is the first time that three proteases have been used in such deuteration experiments in an attempt to further improve a method that was developed about 10 years ago. A recent work17 used protease type XIII and pepsin together, generating only one set of data, whereas we combined three sets of data after three independent proteolyses.
The first advantage of using the three proteases is the increase in the coverage of amino acid sequence. For example, if we use only pepsin the segment 84–92 is not covered (Fig. 1). In contrast, digestion with protease type XVIII generates several peptides covering this part of the protein. Likewise, the region 246–267 is not covered with type XVIII digestion, but type XIII and pepsin each generates several peptides for this zone. So, the combination of the three proteases enables acquisition of a sequence coverage close to 100%. Another advantage of using this set of proteases is the increase in the number of amide hydrogens for which we obtain data. It is not possible to determine the deuterium incorporation for all the peptides identified by LC/MS/MS analyses, but the combination of the data obtained with the three enzymes increases the coverage of the sequence of the deuterated protein. For example, no deuterated peptide is obtained with pepsin for the sequence 62–102, but we can obtain some with the protease type XVIII (Fig. 2). In contrast, type XVIII gave no deuterated peptides on the 568–573 fragment, but the deuterium incorporation can be calculated from the results obtained with pepsin and protease type XIII.
The use of the three proteases can also increase the spatial resolution of the results. Measured incorporations of deuterium are mean values for an area which can be longer than 10 amino acids. However, overlapping of the peptides generated by the three enzymes enables acquisition of data on smaller parts of the protein. For example, after the pepsin digestion, a fragment of 2272.1 Da is identified to be the fragment 634–654; deuteration experiments show that this peptide incorporates 7 deuteriums out of a possible 17. This region of the protein thus seems to be a region with a high number of deuteriums but the definition of the exchange area is not very precise. After digestion with the type XVIII protease we could identify five peptides from this region (fragments 645–655, 646–649, 648–657, 648–658 and 649–653) that incorporate on average 1, 0, 0.5, 1, and 0.5 deuterium atoms, respectively. These data permit reduction of the size of the highly deuterated zone to the 634–645 fragment (Fig. 4).
Furthermore, the combination of the three enzymes generates redundant data. For example, protease type XIII gave information that we could obtain with the two other proteases, but it permitted verification of this information, giving more reliable results.
Validation of the results with the PBP-2X 3-D structure
By using the combination of the three enzymes it is possible to obtain data about the location of deuterium atoms on the PBP-2X* after a short time of deuteration (Fig. 3). The use of a short period of exchange favors deuteration of the most accessible or unstructured regions of the protein. Eight highly deuterated (>30%) segments are identified, namely, fragments 49–63, 103–110, 153–168, 370–389, 492–501, 550–567, 616–624 and 634–645.
These results fit well with the 3-D structures obtained by crystallography.14,15 The most deuterated regions (deuteration incorporation more than 30%) are represented in black (Fig. 5). The first amino acid observed in electron density is Arg75, indicating a disordered N-terminus (49–74). This is in agreement with our observation for the first zone (49–63). The second fragment (103–110) corresponds to an accessible loop at the end of a tong of the N-terminal domain. The third fragment (153–168) is a weakly structured region. The fourth and the sixth fragments (370–389, 550–567) are disordered loops at each side of the base of the active site and are very accessible to the solvent. The fifth fragment (492–501) is a small loop from the transpeptidase domain. The seventh (616–624) and the eighth (634–645) fragments are parts of the linker between the transpeptidase domain and the C-terminal domain, for which the electron density map is absent in the structure at highest resolution (2.4 Å) indicating a high mobility and/or a large disorder.
Thus we can obtain information on the accessibility and the structures of the regions of the PBP-2X*, even for those which were not observed in the crystal state.
To improve a method that has been widely used for about 10 years, we used three different proteases (pepsin, protease type XIII, and protease type XVIII) in H/D exchange experiments combined with mass spectrometry. These proteases were used separately in conditions compatible with the deuteration (and the quenching of deuteration). We obtained a better coverage for the peptide mapping thus avoiding missing some potentially interesting regions of the protein. Furthermore, we obtained a better spatial resolution for deuterium incorporation data, specifying the deuterated regions more precisely. We found a good correlation between strong deuterium incorporation and accessible or unstructured regions of the PBP-2X* three-dimensional structure. This method is expected to help in the fine location of interacting ligands or proteins and thus in the design of drugs interfering in mechanisms involving these interactions.
We thank David Smith and Otto Dideberg for helpful discussions and Anne-Marie Di Guilmi for the gift of the protein PBP-2X*.