The crystal structures of variants of SNase with internal Lys-66, Lys-38 and Lys-92 were published previously (Harms et al., 2008
; Nguyen et al., 2004
; Stites et al., 1991
). Structures for 8 other variants with internal Lys have now been solved, in buffers at a pH in which the Lys side chains should have been neutral ( and ). The substitution of internal positions with Lys did not have an appreciable effect on any of the structures. The backbone of the 12 variants with internal Lys were almost perfectly superimposable and have an average Cα RMSD of less than 0.6 Å relative to the reference protein ( and ). The internal Lys side chains were ordered and visible in the electron density maps of all proteins. Most differences between structures are limited to the Ω loop (residues 44 to 49), which was not present in all the proteins studied. The observation that neutral Lys can be buried in the protein interior without inducing structural reorganization is consistent with previous studies of the resilience of SNase towards substitution of internal positions with bulky, flexible, and non-native amino acids (Wynn et al., 1996
; Daopin et al., 1991
Microenvironment of internal Lys residues in staphylococcal nuclease
Figure 1 (A) Stereo view of the superposition of crystal structures of twelve variants of SNase with internal Lys residue, overlaid on the structure of the Δ+PHS protein (PDB ID: 3BDC). (B) Composite stereo view of the side chains of twelve internal Lys (more ...)
Internal Lys residues with shifted pKa
values decrease the thermodynamic stability of proteins (Isom et al., 2008
; Isom, 2010b). This could have been mitigated through small structural changes to enhance penetration of water to hydrate the internal side chain or to allow extrusion of the internal side chain to bulk water, but this was not observed, perhaps because osmotic and packing forces in the crystal prevent this from happening.
The side chain of Lys-92 was poorly defined in the crystal structure (Nguyen et al., 2004
), but all other internal Lys side chains had strong and continuous electron density for all atoms. The internal Lys side chains could be modeled without clashes with nearby atoms and without requiring reorganization of the core of the protein. In a few cases (e.g. Lys-72) the conformation of the side chain is likely influenced by extensive contacts between the Nζ atom and polar backbone atoms. In all cases, the B factors of the internal Lys side chains were comparable to those of other nearby atoms. The B factors of the proteins with internal Lys residues were comparable to those of other SNase variants at comparable resolution. The rotameric states of these internal Lys side chains are not common (Dunbrack and Karplus, 1993
Solvent-accessible surface area (SASA) and depth of burial calculations showed that with the exception of Lys-72, all the Lys side chains and their Nζ atoms were completely inaccessible to water and buried, some quite far from bulk water (). The side chains of Lys-92 and Lys-104 exist in more than one conformation, suggesting that even when buried the Lys side chains can have considerable mobility. Most Lys side chains point either towards the naturally occurring microcavity at the primary hydrophobic core (Lys-25, Lys-36, Lys-62, Lys-66, Lys-92), or towards the secondary hydrophobic core of SNase (Lys-103, Lys-104, Lys-125) (). The microenvironments of the internal Lys side chains can be very different in different proteins ( and ). Seven of the internal Lys side chains were packed into predominantly hydrophobic regions with few or no polar atoms near the ionizable moiety (e.g. Lys-23, Lys-25, Lys-34, Lys-36, Lys-66, Lys-92, and Lys-125 have no polar atoms within 3.5 Å of the Nζ atom of lysine). Five lysines (Lys-38, Lys-62, Lys-72, Lys-103 and Lys-104) have Nζ atom within 3.5 Å of 1 to 3 polar groups or crystallographic water molecules ().
Figure 2 Microenvironment of the ionizable moieties of 11 internal Lys residues. (A) Δ+PHS SNase; (B) Lys-125 (PDB ID: 3c1e); (C) Lys-25 (PDB ID: 3erq); (D) Lys-36 (PDB ID: 3eji); (E) Lys-104 (PDB ID: 3c1f); (F) Lys-62 (PDB ID: 3dmu); (G) Lys-103 (PDB (more ...)
Lys-72 is unique among these internal Lys residues because it is buried shallowly. Its Nζ atom is only 7.9% accessible, 3.5 Å from a crystallographic water molecule, and within hydrogen bonding distance of three backbone carbonyl residues. Surprisingly, despite this highly polar microenvironment, Lys-72 titrates with an anomalous pKa
of 8.6. This contrasts with the case of Lys-38, which has a normal pKa
despite its Nζ atom being fully buried (Harms et al., 2008
). Crystal structures did not reveal a clear connection between measured shifts in pKa
values and the depth of burial or polarity of the microenvironments of Nζ atoms, suggesting that other structural properties must be the critical determinants of the pKa
values of these internal groups.
Structural consequences of substitution with neutral Lys
N HSQC spectra of the 25 Lys-containing variants were first measured under conditions of pH where the internal Lys residues were mostly neutral (Table S1 and Fig. S1–S3
). Twenty-four out of 25 variants exhibited chemical shift dispersion and well-resolved cross peaks characteristic of folded proteins, consistent with results from CD and Trp fluorescence (Isom et al., 2008
). The exception was the N100K variant, whose 1
N HSQC spectrum showed a significant loss of peaks consistent with some unfolding.
Assignment of HSQC spectra and other types of NMR spectroscopy experiments will be required to describe on a site-by-site basis the structural and dynamic consequences of substitution of internal positions with Lys residues. However, to identify the proteins worthy of further detailed study it was first necessary to characterize the structural consequences of substitution with and ionization of internal Lys residues in the 25 variants. To examine 1H-15N HSQC spectra qualitatively at pH values where the internal Lys was neutral we focused on the subset of resonances of each variant that are different from resonances in the spectrum of the reference protein at the same pH. The subset of resonances that was affected by the substitution differed from variant to variant. For qualitative purposes, peaks that overlapped those in the reference spectrum unambiguously were assigned by visual inspection. The total number of peaks assigned in this manner was normalized by the total number of assigned peaks in the 1H-15N HSQC spectrum of the reference protein. This provided a measure of how similar the HSQC spectrum of a variant was to that of the reference protein (). Although this analysis is strictly only qualitative, it contributed significant new insight into the type, magnitude and distribution of structural reorganization coupled to the presence and ionization of internal Lys residues.
Summary of global thermodynamic stability, pKa values of internal Lys residues and effects of pH on the HSQC spectra of variants of SNase with internal Lys residues
The 24 variants that are folded under conditions of pH where the Lys residues are supposed to be neutral had spectra with 42% to 87% similarity to the reference protein (). On average 65% of all peaks in the Lys-containing variants were superimposable with peaks in the reference protein. The relatively low level of overlap between the spectra of the Lys-containing variants and the reference protein reflects primarily the sensitivity of the amide chemical shift (Wishart and Case, 2001
). On the other hand, the absence of major differences between the 1
N HSQC spectra of the reference protein and of the Lys-containing variants is consistent with the similarity between crystal structures. Note that 1
N HSQC spectra of some variants of SNase with internal ionizable groups have been assigned and analyzed in terms of distribution of ϕ/ψ angles, and that in those cases the NMR data are fully consistent with the distribution of secondary structure observed in the crystals (Harms et al., 2009
; Harms et al., 2008
). In fact, the resonances that were affected by the substitutions (e.g.V66K or V66D) were not affected by the ionization of the internal Lys-66 or Asp-66 (Chimenti et al., 2011
The highest level of similarity between the reference protein and a Lys-containing variant was for the variants with the most solvent exposed Lys (I72K variant with 87% similarity with the spectrum of the reference protein), and for the variant with a Lys in a loop region (N118K variant, with 81% similarity). The variants that introduce Lys into the primary hydrophobic core of SNase at positions 25, 36, 62, and 66 also had HSQC spectra with 70–85% peak overlap with the reference spectrum, suggesting that the region of the protein between helix-1 and β-sheets 2 and 3 tolerate substitutions with Lys better than the region near helix-2 and helix-3, where substitutions with Lys yielded more highly perturbed HSQC spectra. No other general, qualitative trends were observed in the similarities in HSQC before and after substitution of internal positions with Lys.
Structural consequences of the ionization of internal Lys
N HSQC spectra were measured for each variant at pH values that bracketed the apparent pKa
of the internal Lys. The exception was the V23K, for which the highest pH studied corresponded to a form in which Lys-23 was only expected to be 50% neutral. The dispersion and resolution of the 1
N HSQC spectra indicate that 23 of the 25 variants were fully folded when the internal Lys residues were charged ( shows representative 1
N HSQC spectra before and after the ionization of an internal Lys). In general, the difference in amide proton chemical shifts (Δδ(1
)) in each variant between conditions where the internal Lys was charged and neutral were quite small (less than ~0.1 ppm). This is illustrated for three specific cases chosen to represent proteins with internal Lys residues with small (N118K), medium (I72K), and large (L125K) depressions in their pKa
(). With the exception of the variant with I92K discussed ahead, the amide shifts did not show evidence of any significant structural reorganization upon ionization of the internal Lys, not even in cases where the internal Lys residue has a highly depressed pKa
. This is consistent with results with optical spectroscopy showing that all but the I92K variant remained folded when the internal Lys residues were charged (Isom et al., 2008
). The variants appear to retain their native folded structure under conditions of pH where the internal Lys residues are charged.
Figure 3 1H-15N HSQC spectra acquired at pH values above (black) and below (color) the pKa value of internal Lys residues in four proteins. (A) Example of slow exchange with a fully unfolded form in the I92K variant (pKa ~ 5.3) when Lys-92 is ionized at pH 4 (red). (more ...)
Figure 4 ΔΔδ 1HN (ppm) calculated as ΔΔδ(ppm) = [δ(variant)neutral − δ(variant)charged] − [δ(Δ+PHS)neutral − δ(Δ+PHS)charged] (eqs. 2.1) and (more ...)
Local reorganization coupled to the ionization of internal Lys residues
The ionization of five Lys residues (Lys-25, Lys-34, Lys-58, Lys-66 and Lys-125) led to intermediate exchange behavior, consistent with increased dynamics or conformational fluctuations localized to a small part of the protein near the internal Lys side chain (). These cases are of special interest because they have information about the roles of conformational reorganization and dynamics as determinants of the pKa values of internal groups.
The ionization of internal Lys residues can increase the frequency and amplitude of their conformational fluctuations (Damjanovic et al., 2005
). The fluctuation of a charge around nearby amides may contribute to the line-broadening that was observed with some of the variants. However, the observation that the affected resonances do not remain in fast exchange and instead enter intermediate exchange is consistent with underlying slower local fluctuations of the backbone on the order of micro- to milliseconds, which promote conformational states where the internal charged Lys can become dynamic and solvated by water. This has been characterized in detail in the V66K variant (Chimenti et al., 2011
; Kitahara, 2010
); the Lys-66 side chain is internal when neutral, but still exchanges with water at a rapid rate of ~90 sec−1
at −1 °C, consistent with local unfolding or water penetration (Takayama et al., 2008
The L25K and L125K variants responded to the ionization of the internal Lys with the same intermediate exchange behavior observed in the V66K variant. Extreme line-broadening and large chemical shift changes were found in the HSQC spectra for a subset of resonances around a region immediately adjacent to the side chains of Lys-25 and Lys-125 (). Only ~20 resonances broadened upon ionization of the internal lysine of V66K and L125K. As many as 50 resonances broadened in response to the ionization of Lys-25, consistent with a larger dynamic response to ionization of this Lys residue. Many of the affected resonances in the L25K spectrum lost intensity without experiencing chemical shift changes, suggesting a purely dynamic process without structural rearrangement.
The F34K and V39K variants also showed significant intermediate exchange in the HSQC spectra. These variants had stability greater than 3 kcal/mol when the Lys was ionized, indicating that at certain positions in the protein, ionization of Lys can promote increased fluctuations even when the protein is relatively stable. Lys-39 is located in a loop region where fluctuations of the backbone may be normally less restricted. This observation of intermediate exchange dynamics in these more stable variants is counter to the idea that in the more stable variants the ionization of the internal Lys side chain does not increase local fluctuations. On the other hand, the majority of the more stable variants displayed no observable exchange behavior upon titration of the internal lysine (). Further studies will be necessary to examine further the relationship between ionization of internal Lys residues, promotion of local fluctuations, and global, thermodynamic stability.
Global unfolding coupled to the ionization of internal Lys
The 1H-15N HSQC spectra of the I92K and N100K variants at pH 6.7 and 9.4 respectively, where Lys-92 and Lys-100 are neutral, showed slow exchange behavior between the native state (in the case of I92K) or a partially folded form containing some native peaks (in the case of N100K), and an unfolded form that was characterized by a region of poorly-defined, broadened resonances from 8.0 to 8.4 ppm in 1HN (). The stability of I92K is ~0.2 kcal/mol at pH 4.9 (where Lys-92 is ionized) and the stability of N100K is ~1.3 kcal/mol at pH 7.0 (where Lys-100 is ionized). In both cases, when the Lys residue is >90% ionized, the unfolded form is the only form observable in the 1H-15N HSQC spectrum. In both cases, peaks from the native state in the 1H-15N HSQC did not show substantial chemical shift changes upon titration, indicating that the data are reporting a two-state exchange between folded (or partially folded, in the case of N100K) and unfolded forms without significant intermediates. The NMR data show unequivocally that ionization of Lys-92 and Lys-100 led to the global unfolding of the protein. The high apparent dielectric constants reported by Lys-92 and Lys-100 is simply a reflection of dielectric breakdown.
The HSQC spectra of the A58K, T62K, and V99K variants exhibited slow exchange behavior of a different type. In the case of T62K and V99K, the vast majority of the peaks remained at full intensity throughout the titration. However, at the low pH endpoint of the titration, new resonances appeared in the unfolded region of the 1H-15N HSQC spectrum. These resonances had lower intensity than the native peaks, consistent with the presence of a minor, partially unfolded form of the protein. In the case of the variant with Lys-58 several new resonances appeared at low pH, but they were generally found near native peaks, not in the unfolded region. This could indicate that the protein is switching between two very similar folded conformations on a slow timescale.
Relationship between global thermodynamic stability and the dielectric response
The type of structural response triggered by the ionization of Lys residues in different parts of the protein is summarized in . It is striking that the majority of the variants, including many (L36K, V23K, V74K, V104K, L103K, I72K, A90K, Y91K, A109K, and T41K) in which the Lys titrates with a depressed pKa value, showed no evidence of any major conformational changes coupled to the ionization of the internal Lys.
Figure 5 Summary of consequences of ionization of internal Lys residues in Δ+PHS SNase observed in 1H-15N HSQC data. (●) Ionization of Lys at these locations induced minor or no changes in chemical shifts and no slow or intermediate exchange behavior; (more ...)
All the variants with ΔG°H2O less than 3 kcal/mol at the pH where the internal Lys became fully ionized exhibited some form of exchange behavior in the HSQC spectra, consistent with increased fluctuations between alternative states (, ). These proteins can be divided into two classes depending on the nature of the response. Variants in one class (I92K, N100K, and to a lesser extent, T62K and V99K) exhibited exchange with a fully unfolded form. The A58K variant could be included in this set although the exchange behavior appears to affect only a small subset of peaks. The members of a second class (L25K, V39K, F34K, V66K, L125K) exhibit intermediate exchange only at a subset of residues consistent with exchange with a locally or partially unfolded form.
Whereas the structure of the native state is independent of global stability, the extent of the structural response of the protein to the ionization of an internal group appears to be somewhat dependent on the global thermodynamic stability of the protein. Partially unfolded states in which the internal charged moiety can be solvated through contact with protein or water should be populated more readily when the global stability of a protein is low. That is, in the case of variants with internal Lys residues with depressed pKa
the probability of populating excited states, or partially unfolded states, increases with decreasing pH because the energy gap between the native and the unfolded state decreases with decreasing pH. A similar inverse relationship between ΔG°H2O
and dynamic backbone fluctuations has been observed previously in SNase by NMR relaxation methods (Alexandrescu et al., 1996
) and native-state hydrogen exchange (Maity et al., 2003
Structural origins of the high apparent dielectric constant in the protein interior
The Lys side chains introduced in SNase are internal and buried deeply when they are neutral (). The majority of them are embedded in highly hydrophobic microenvironments that are inconsistent with the high apparent dielectric reported by the measured pKa values. For this reason it was natural to invoke increased dynamics or structural reorganization to explain the high dielectric constants consistent with the experimental data. The crystal structures showed no evidence of structural reorganization. Relying on the atomic resolution afforded by NMR spectroscopy we have now shown that 15 out of 25 internal Lys residues in SNase ionize without triggering detectable changes in chemical shifts or in exchange behavior. Neither the substitution of a core position with a Lys nor the ionization of the Lys had any detectable structural or dynamic perturbations in these 15 cases. Six of these 15 Lys residues titrate with a normal pKa (L38K, A132K, A58K, L37K, G20K, N118K) and with the exception of Lys-38, which is known to be buried, these could well end up being trivial cases where the Lys side chains are not actually buried. Crystal structures are needed to establish this. However, the other nine Lys side chains have depressed pKa values. Regardless of what the origins of the apparent dielectric constants reported by these Lys residues might be, they are sufficiently subtle so as to be invisible in HSQC spectra. Three of the four proteins where the amino moiety of the internal Lys residue makes contacts with polar atoms fall under this category; in these cases this high local polarity might be sufficient to solvate an internal charged group thereby preventing conformational changes in response to the ionization event. The lack of evidence for processes in the intermediate exchange regime in the 1H-15N HSQC spectra of a majority of Lys-containing variants of SNase can only be used to exclude very slow and intermediate timescale fluctuations as important elements of the dielectric response probed by the internal Lys residues. Contributions from very fast dynamics or from minor structural forms in solution (<10% populated) can still play a role; this is being examined.
Another possibility is that in the 15 cases where there is no evidence of structural or dynamic perturbation, the Lys side chain is embedded in microenvironments where buried water molecules hydrate the internal ionizable moiety, essentially creating a favorable dielectric environment for the polar group. Internal water molecules near the internal ionizable moieties were absent near most internal Lys residues, but have been observed near other ionizable groups (Schlessman et al., 2008
). Alternatively, the minor chemical shift changes upon ionization of the Lys residues in these 15 variants may reflect local electrostatic effects or increased local dynamics on the fast timescale that would be invisible in the 1
N HSQC data. Molecular dynamics (MD) simulation on the crystal structure of the L38K variant provided strong evidence of local unfolding and water penetration to solvate the Lys-38 side chain on ns timescales (Harms et al., 2008
). A more rigorous experimental characterization of these 15 proteins with relaxation measurements and native-state hydrogen exchange will be needed to obtain more detailed insight into mechanisms of dielectric relaxation.
The substitution of internal positions with Lys decreases the thermodynamic stability of SNase. The 1
N HSQC spectra of the most destabilized variants were the ones that exhibited exchange behavior consistent with either increased dynamics at backbone amides or global unfolding in response to the ionization of the group. This suggests that the character of the apparent dielectric response of a protein can be influenced by its global stability (Karp et al., 2010
). In the proteins with the lowest thermodynamic stability the partially unfolded states in which the ionized side chains of internal Lys can contact bulk water can be sampled more readily through increased local fluctuations. Because the local fluctuations occur on the intermediate or slow timescales, this increase in local fluctuations was not observed directly but was manifested indirectly in the 1
N HSQC spectra. Many peaks in these spectra also exhibited large chemical shift changes upon ionization that were consistent with small structural perturbations or electrostatic effects emanating from the fluctuating electrostatic field from the fluctuating Lys side chain. Efforts are underway to demonstrate that the pKa
values of internal ionizable groups are sensitive to the global thermodynamic stability and that for this reason stability can affect the apparent dielectric effect. If this idea is borne out by further experimental data, it has the potential to redirect efforts in structure-based calculation of electrostatic energies.
The increased dynamics and intermediate or slow exchange observed in 10 of 25 variant proteins (variants I92K, V66K, L125K, L25K, V99K, F34K with low stability and T62K, N100K, V39K, A58K with high stability) is consistent with the idea that the high apparent polarizibility reported by the internal Lys residues in some of these proteins is partly a reflection of increased dynamics or structural reorganization concomitant with the ionization event. This type of structural response to the ionization of an internal group is essential for biological function in protein such as bacteriorhodopsin (Lanyi and Luecke, 2001
), in ATPase (Rastogi and Girvin, 1999
), and in the photoactive yellow protein (Xie et al., 2001
). Our smaller and more tractable protein will allow benchmarking of algorithms need to examine mechanism of H+
-driven conformational changes in many of the proteins that use internal ionizable groups for purposes of biological energy transduction.
This NMR spectroscopy survey of structural consequences of the ionization of internal Lys residues in SNase contributes novel insight into the range of possible structural reorganization that can be triggered by the ionization of internal groups, and how this is affected by details of the microenvironment of the ionizable moiety or by thermodynamic properties of the protein. It also contributes a set of constraints useful to benchmark the performance and accuracy of computational methods for structure-based calculations of pKa
values of ionizable groups in dehydrated environments. The ideas about structural origins of dielectric effects and molecular determinants of pKa
values of internal groups that emerge from this study have important, general implications for computational efforts. Computational methods are going to have to be able to identify cases in which the ionization of an internal group can trigger structural rearrangement or increased dynamics, and they should be capable of identifying the specific conformational states stabilized through the ionization of buried groups. The possibility that the structural response to the ionization of an internal group is actually governed by the global stability of the protein raises several challenges for algorithms for structure-based pKa
calculations. The most challenging problem is the need for improved sampling capabilities so alternative conformational states that might be populated when an internal ionizable group becomes charged are actually sampled during a calculation. On-going efforts by some groups suggest this is still a daunting task (Kato and Warshel, 2006
). Another challenge is the need to be able to discriminate among alternative states and to identify the ones that are relevant. Structure-based calculation of pKa
values has in general been focused exclusively on accurate calculation of electrostatic forces. If the suggestion that the global thermodynamic stability of a protein can affect pKa
values is confirmed by experimentation, computational methods will have to involve accurate calculation of thermodynamic stability (i.e. ΔG°H2O
). Free energy determines the probability of population of a conformational state relative to another; therefore, this thermodynamic parameter will be necessary to elucidate molecular determinants of pKa
values and mechanisms of H+
-driven conformational transitions in proteins.