Conformational analysis of processivity clamps in solution demonstrates that tertiary structure does not correlate with protein dynamics

Summary

The relationship between protein sequence, structure, and dynamics has been elusive. We report one of the first comprehensive analyses using an in-solution experimental approach to study how the conservation of tertiary structure correlates with protein dynamics. Hydrogen exchange measurements of eight processivity clamp proteins from different species revealed that, despite highly similar three-dimensional structures, clamp proteins display a wide range of dynamic behavior. Differences were apparent both for structurally similar domains within proteins and for corresponding domains of different proteins. Several of the clamps contained regions that underwent local unfolding with different half-lives. We also observed a conserved pattern of alternating dynamics of the α-helices lining the inner pore of the clamps as well as a correlation between dynamics and the number of salt bridges in these α-helices. Our observations reveal that tertiary structure and dynamics are not directly correlated and that primary structure plays an important role in dynamics.

Introduction

Proteins are dynamic macromolecules with motions that are essential for biological function (Bahar et al., 2010; Frauenfelder et al., 1991) and may play an important role in evolution (Gerek et al., 2013; Maguid et al., 2006; Tokuriki and Tawfik, 2009). Substantial theoretical and experimental work has demonstrated that functional motions in proteins include both large-scale conformational changes, such as domain transitions (Hanson et al., 2007; Rothwell et al., 2013), and local motions, such as backbone fluctuations and localized unfolding and refolding (Chen et al., 2007; Eisenmesser et al., 2005; Shan et al., 2013). Such motions must be encoded in the structure of the proteins. Indeed, for certain enzymes, such as members of the RNase A superfamily, there seems to be a conservation of both structure and dynamics (Gagne et al., 2012; Gagne and Doucet, 2013). Although some computational investigations suggest that there is a correlation between protein structure, dynamics, and function (Gerek et al., 2013; Hensen et al., 2012; Pang et al., 2005), for most proteins this relationship remains unclear. It has also been shown that the same function can be performed by proteins with different structures (Galperin et al., 1998) and studies on various SH3 domains have provided some evidence suggesting that structurally conserved proteins may not share the same dynamic properties (Wales and Engen, 2006b).

Here, we have addressed the question of whether the sliding clamp proteins of highly similar structures also have similar dynamics. We have determined the dynamics of eight structurally conserved sliding clamp proteins using the in-solution experimental approach of hydrogen-deuterium exchange mass spectrometry (HX MS). Sliding clamps are conserved across all domains of life and some bacteriophage (Kelman and O'Donnell, 1995) and have highly similar structures at both the domain level and the whole protein level, despite having low sequence similarity (Figure 1 and S1). All sliding clamps have a highly conserved ring-shaped structure with pseudo-six-fold symmetry and a central pore large enough to accommodate double-stranded DNA.

Figure 1. Sliding Clamps Have Highly Conserved Structures and Low Sequence Similarity.

(A) Structural alignment of T4 gp45 trimer (green; PDB ID: 1czd), E. coli β clamp dimer (blue; PDB ID: 1mmi), T. kodakaraensis TK0582 trimer (yellow; PDB ID: 3lx2), and human PCNA trimer (red; PDB ID: 1vym). (B) Superposition of all 17 sliding clamp domains in this study. Domains were extracted from the PDB files and superimposed. PDB IDs: T4 gp45, 1czd; β clamp, 1mmi; yPCNA, 1plq; TK0582, 3lx2, TK0535, 3lx1; AtPCNA1, 2zvv; AtPCNA2, 2zvw; hPCNA, 1vym. (C) Plot of the percent sequence identity versus the rmsd from the pairwise structural alignments. Filled diamonds, full-length proteins; open squares, domains; blue squares, domains within the same protein. See also Figure S1.

Sliding clamps function as DNA polymerase processivity factors by encircling DNA and tethering the polymerase to the DNA template. Their dynamic properties are directly related to their functions since they need to be loaded onto DNA (Hedglin et al., 2013) and slide along the DNA helix (Barsky et al., 2011). Although there has been considerable effort to determine the dynamics of sliding clamps by computational methods (Adelman et al., 2010; Ivanov et al., 2006; Kazmirski et al., 2005), experimental data is still limited (De Biasio et al., 2011; Fang et al., 2011; Kochaniak et al., 2009; Tsutakawa et al., 2011). Besides their essential role in DNA replication, sliding clamps have been implicated in a large number of other cellular processes, in which they manage cellular responses to DNA damage and cell-cycle progression via their protein interactions (Vivona and Kelman, 2003). Most clamp-interacting proteins bind a hydrophobic pocket on the surface of the clamps via a conserved motif, although they may make unique contacts as well (Dalrymple et al., 2001; Warbrick, 1998).

Here we have used HX MS to probe the dynamics of sliding clamps from bacteriophage T4 (gp45), bacteria (Escherichia coli β clamp) (Fang et al., 2011), archaea (Thermococcus kodakaraensis TK0535 and TK0582), yeast (Saccharomyces cerevisiae yPCNA), plants (Arabidopsis thaliana AtPCNA1 and AtPCNA2), and humans (hPCNA). HX MS is a powerful tool for probing the dynamics of proteins in solution and is based on monitoring the mass increase of a protein when the backbone amide hydrogen atoms exchange with deuterium atoms in the solvent (Wales and Engen, 2006a; Zhang and Smith, 1993). The rate of HX is strongly dependent on hydrogen bonding and solvent accessibility and is therefore directly related to protein structure and dynamics (Englander et al., 1972; Englander and Kallenbach, 1983; Hvidt and Linderstrom-Lang, 1954). Molecular dynamics (MD) simulations were also performed to analyze the formation of hydrogen bonds and salt bridges in the different proteins. We present a detailed analysis showing that high conservation of tertiary structure of proteins and protein domains does not correlate with conservation of dynamics, suggesting that protein dynamics is more dependent on primary structure than tertiary structure.

Results

Different Sliding Clamps Have Different Dynamics in Solution

We measured the deuterium uptake of eight different sliding clamps by exposing each purified protein to deuterium for various periods of time, quenching the exchange, digesting the protein into peptides under quench conditions, and measuring the mass increase of each peptide by MS. We previously analyzed the dynamics of the E. coli β clamp using this strategy (Fang et al., 2011). The reported mass increase is due to the deuteration of backbone amide positions because deuterium incorporated into sidechains has a fast exchange rate and is washed away during the analysis steps (Zhang and Smith, 1993). Peptides were identified by a combination of accurate mass measurements and tandem mass spectrometry (MS/MS) using collision-induced dissociation. The resulting sequence coverage was over 91% for all proteins (Figure S2).

The fractional deuterium uptake for each peptide was mapped onto the crystal structures (Kontopidis et al., 2005; Krishna et al., 1994; Ladner et al., 2011; Moarefi et al., 2000; Oakley et al., 2003; Strzalka et al., 2009) of the respective proteins (Figure 2 and S2). These maps show that the clamps display strikingly different local dynamics; in particular, the dynamics are heterogeneous both within a particular protein and between proteins from different species. In general, the regions of the proteins that exhibited the most deuterium uptake during the first minute of exchange were solvent-exposed loops and some of the outer β-strands, whereas the α-helices in the inner pore exhibited relatively low deuterium uptake. The exception to this was seen in TK0535 where the α-helices in domain II exchanged with relatively high rates relative to the outer region of this protein (Figure 2 and S2). Strikingly, the dynamics of the domains in each monomer were also found to be different although they have nearly identical three-dimensional structures (Figure 1 and 2). For the trimeric clamps, including T4 gp45, domain II is more dynamic than domain I. The situation is reversed in the β clamp, where domain I is the most dynamic of the three domains.

Figure 2. Sliding Clamps Have Different Dynamics.

The relative percentage deuterium incorporation is shown using color gradients across different time points. Gray indicates regions that were not identified. Dashed lines indicate subunit interfaces. PDB IDs: T4 gp45, 1czd; β clamp, 1mmi; yPCNA, 1plq; TK0582, 3lx2; TK0535, 3lx1; AtPCNA1, 2zvv; hPCNA, 1vym. The melting temperatures were determined by a thermal shift assay as described in Experimental Procedures. The far right column highlights EX1 peptides and indicates proteins displaying only EX2 kinetics. HX MS data for AtPCNA2 (T_m = 51.6 °C) are shown in Figure S2. See also Figure S2

The eukaryotic PCNAs exhibited similar trends in deuterium incorporation although there were regional differences in different proteins (Figure 2). The human and plant PCNAs were more dynamic than yPCNA or the two archaeal PCNAs (Figure 2 and S2). After 10 seconds of exchange, the majority of peptides in hPCNA, AtPCNA1, and AtPCNA2 were more than 10% deuterated.

The Arabidopsis genome encodes two PCNAs (Strzalka et al., 2009); the sequence identity of AtPCNA1 and AtPCNA2 is 97% and the deuterium uptake of these proteins was highly similar (Figure 2 and S2). The main difference was observed in peptides 215-223 and 215-226, which are part of helix α4 and the following β-strand in domain II and contain one (AtPCNA1 Asp223 vs AtPCNA2 Glu223) of the eight amino acid differences in these two PCNAs (Figure S2); this region incorporated more deuterium in AtPCNA1 than in AtPCNA2. Even this one conservative amino acid substitution is sufficient to alter the deuterium uptake.

The T. kodakaraensis genome also contains two genes encoding PCNA homologs (TK0535 and TK0582) with 54% sequence identity (Figure S1B). The HX MS data showed that TK0535 and TK0582 have significantly different local dynamics (Figure 2 and S2). In particular the interdomain connector loop (IDCL) of TK0535 was less readily deuterated. This is consistent with the average number of salt bridges formed by IDCL residues during 1 ns interval of MD simulation, which was 2.1 in TK0535 and virtually none (0.02) in TK0582. The deuterium uptake of residues located at the subunit interface relative to the domain interface in TK0535 was higher, whereas the opposite was true for TK0582. This is consistent with a previous study indicating that TK0535 has a less stable subunit interface than TK0582 (Ladner et al., 2011).

Bacteriophage T4 gp45 was the most dynamic of the proteins investigated. During the first minute of exchange, much of the backbone of T4 gp45 was relatively protected from exchange while there was some exchange in the loops and some of the outer β-strands (Figure 2 and S2). However, after 10 min, some of the internal α-helices were over 50% deuterated, indicating a high degree of fluctuations in the central pore of the clamp, and at one hour large regions of T4 gp45 were over 60% deuterated. Exchange of this magnitude was not found in any of the other clamp proteins.

It could be argued that the variability in dynamics is derived solely from the primary structure, given the highly variable nature of the sequences for these proteins. The primary structure dictates the very local interactions and local stability of structural features so perhaps it should not be surprising that the hydrogen exchange level is not the same for variable sequences. As the intrinsic rate of exchange, which is that in the absence of secondary, tertiary and quaternary structure, can be calculated for any amino acid sequence (Bai et al., 1993), we compared the intrinsic rates of exchange under native and quench pH for the three domains of the β clamp to determine if there was a correlation of sequence variability with the dynamics we observed. The results (data not shown) revealed no correlation between intrinsic rates of exchange (or back-exchange) and the measured dynamics, arguing that the measured hydrogen exchange is a complex function that certainly involves higher-order structure.

To address whether the difference in dynamics that we found were related to stability, we determined the thermodynamic stability of the different proteins using a fluorescence thermal shift assay (Pantoliano et al., 2001). The melting temperature (T_m), defined as the midpoint of the melting transition, ranged from 47.8 °C to 89.0 °C for the different sliding clamps (Figure 2). There was not a strict correlation between T_m and deuterium uptake, but a consistent trend of less stable proteins being more dynamic was observed. Specifically, hPCNA and the two plant PCNAs were significantly more dynamic and less thermodynamically stable than yPCNA.

Dynamics at the Subunit Interfaces

All sliding clamps in this study form ring-shaped dimers (β clamp) or trimers (PCNAs and gp45) with the monomeric subunits arranged head-to-tail. Each subunit interface includes antiparallel β-strands that form an extended β-sheet that is part of the continuous layer of β-sheet structure on the outside of the clamps. In general, peptides located in the β-strands of the subunit interfaces exhibited at least as much deuterium uptake as peptides located in the domain interfaces (Figure 2 and S2), indicating a relatively dynamic behavior of the subunit interfaces. In the trimeric clamps, the β-strands in domain II that are part of the subunit interface were more readily deuterated than the β-strands in domain I of the adjacent monomer, except for T4 gp45 at the early time points (Figure 2 and S2). The β-strands in domain III of the β clamp were more deuterated than those in domain I of the adjacent monomer. Human PCNA has the most symmetric subunit interface of the clamps in terms of deuterium uptake of the β-strands. Although most of the sliding clamps seem to have relatively unprotected subunit interfaces, there were notable differences in the degree of deuterium uptake between different clamps, suggesting different dynamics. For example, yPCNA seems to have a less dynamic monomer interface than plant and human PCNAs.

The Interdomain Connector Loops

In each sliding clamp, domains are connected by a loop that spans the domain interface on the outside of the clamp. These loops were generally the most rapidly deuterated parts of the proteins (Figure 2 and S2). However, in several of the proteins these loops were relatively well protected during the first 10 seconds of exchange, suggesting that there is transient protection that could involve occlusion from solvent or the formation of hydrogen bonds between the backbone of the loop(s) and, perhaps, the globular domains. In the β clamp, the loop that connects domains I and II was more readily deuterated than the loop that connects domains II and III. It is noteworthy that in contrast to the other clamps, the IDCL of TK0535 was not the most readily deuterated part of the protein, suggesting that the backbone of this loop is significantly more protected from exchange relative to the IDCLs of the other clamps

The IDCLs of hPCNA and the plant PCNAs (residues 118-133) were rapidly labeled and were over 40% deuterated after only 10 seconds of exchange. However, in hPCNA, the C-terminal part of the IDCL (peptide 126-132) exhibited significantly faster exchange than the N-terminal part (peptide 116-124), suggesting that the backbone of the N-terminal part of the IDCL in hPCNA may be protected, perhaps by interactions with domain I. Similarly, these loops of TK0582 and the β clamp showed differential deuteration at some time points. In contrast, the corresponding peptides in yPCNA (peptides 119-125 and 126-130) showed similar levels of deuterium incorporation, suggesting that the protection of the IDCLs in yPCNA was more uniform. The IDCL of the plant PCNAs was covered by only one peptic peptide and we were therefore unable to distinguish the N- and C-terminal parts of the IDCL in these proteins. Because the IDCL is part of the binding site of the PCNA-interacting peptide motif (PIP box) present in most PCNA-interacting proteins (Warbrick, 1998), the heterogeneous deuteration that we observed in this loop may contribute to the diversity of protein-protein interactions of PCNA.

Local Unfolding

A benefit of continuous labeling HX MS experiments is that the mass spectra reveal distinct populations and report on exchange as a result of EX1 and EX2 kinetics (Hvidt and Nielsen, 1966; Weis et al., 2006). Under EX2 kinetics, which is the kinetic regime followed by most proteins in native conditions, many unfolding-refolding events or fluctuations may occur prior to isotope exchange and the mass spectra contain a single isotope distribution that increases in mass over time. In contrast, under EX1 kinetics, the rate of exchange is much larger than the rate of refolding so that an unfolded region becomes completely deuterated before it refolds. Thus, the mass spectra contain a bimodal isotope distribution representing one population that has not yet unfolded (lower mass) and one population that has unfolded (higher mass).

We observed EX1 kinetics in many peptides of T4 gp45, indicative of cooperative local unfolding (Figure 2 and 3A-C). Compared with the β clamp, where EX1 was observed in more peptides from domain I (Fang et al., 2011), the EX1 peptides in T4 gp45 were distributed in both domains I and II (Figure 2 and 3C). However, the distribution of EX1 kinetics was not symmetric in the two domains even though they possess similar tertiary structures. All EX1 peptides in T4 gp45 displayed similar patterns of dynamics and the half-life of the local unfolding was ~5 min (Figure 3A), significantly shorter than that observed in the β clamp (~3.5 h) (Fang et al., 2011). We repeated the analysis of the β clamp in parallel with these experiments; while the overall level of deuterium uptake was higher than in the previous work (Fang et al., 2011), the pattern of peptides exhibiting EX1 kinetics was remarkably similar (Figure 2, far right column).

Figure 3. EX1 Kinetics in T4 gp45 and Human PCNA.

(A) Mass spectra for a representative T4 gp45 peptide, residues 73-85 (m/z = 695.4, +2 charge state). The mass spectrum corresponding to the exchange time-point closest to the approximate half-life of the unfolding events is displayed in blue. (B) Time course of deuterium uptake of peptide 73-85. Each line represents an independent HX MS experiment. (C) All peptides showing EX1 kinetics in T4 gp45 (red) mapped onto the crystal structure (PDB ID: 1czd). (D) Mass spectra of the EX1 peptide in hPCNA, residues 170-182 (m/z = 630.3, +2 charge state). (E) Time course of deuterium uptake of peptide 170-182. (F) The EX1 peptide is part of the subunit interface and is shown in red on the crystal structure of hPCNA (PDB ID: 1vym). The two subunits are colored green and yellow, respectively.

The peptide covering residues 170-182 at the subunit interface of hPCNA showed a clear EX1 bimodal isotope distribution in the mass spectra (Figure 3D and E). We also observed similar bimodal isotope distributions for the overlapping peptides 158-182 and 163-182, although the mass spectra contained more EX2 character than for peptide 170-182 suggesting that the local unfolding is primarily localized to the C-terminal part of these peptides. The apparent half-life of this local unfolding is approximately 60 min and the relative deuterium level of residues 170-182 was 60% after 4 h of labeling. Residues 170-182 are located in the subunit interface and include the β strand in domain II that forms a continuous β sheet with β strands from domain I in the adjacent subunit (Figure 2 and 3F). The corresponding peptides in yeast and plant PCNAs followed solely EX2 kinetics, although in the plant PCNAs this peptide was highly deuterated, at 50% after 4 h. The large number and distribution throughout the proteins of peptides undergoing EX1 kinetics in T4 gp45 is consistent with this being the least stable and most dynamic of the clamps investigated. In general, the proteins that showed the most deuterium uptake overall also exhibited EX1 kinetics.

Variable dynamics of the α-helices in the inner pore

The function of sliding clamps is dependent on their ability to interact with DNA via interactions between the α-helices lining the inner pore of the clamps and the DNA backbone. During the first minute of HX, these α-helices were relatively protected from exchange in all the sliding clamps except for TK0535 (Figure 2 and Fig S2). However, in T4 gp45, β clamp, TK0582, AtPCNA1, AtPCNA2, and hPCNA, some of these α-helices reached deuterium levels that were among the highest in the protein, indicating a large degree of structural fluctuations in the inner pore. Although the extent of deuterium uptake differed significantly between the different proteins, the α-helices in the inner pore of all the eukaryotic PCNAs alternate between more dynamic and more stable (Figure 2 and S2). A similar pattern was observed in the β clamp. The inner α-helices in TK0582 exhibited a somewhat different pattern, in which the helices at the subunit interface were the most protected from exchange. Figure 4A shows the deuterium uptake for peptides located in the four α-helices of the inner pore of hPCNA. Helices α1 and α3 are relatively stable, whereas helices α2 and α4 are relatively dynamic. The same pattern of dynamics exists in the inner pore of yPCNA, AtPCNA1, AtPCNA2, and domain I and II of the β clamp (Figure 2 and S2), suggesting that it is a conserved pattern of dynamics. There is a strong evolutionary conservation of both positively and negatively charged residues in these α-helices (Barsky et al., 2011) and we observed that α-helices with both positively and negatively charged residues exhibited slower HX than α-helices that had exclusively positively or negatively charged residues. We performed MD simulations to determine the number salt bridges involving at least one residue of a given α-helix, which show that there is a negative correlation between the number of salt bridges and the deuterium uptake (Figure 4, S3, and Table S1). Many more salt bridges were formed in the relatively less dynamic α-helices of hPCNA, in particular, but the pattern also generally holds for other clamps.

Figure 4. Alternating Dynamics in the Inner Pore of Human PCNA.

(A) Average number of salt bridges formed by the inner pore α-helices of hPCNA (PDB ID: 1vym) during 1 ns of MD simulation. (B) HX MS data for peptides that are part of the four α-helices in the inner pore of human PCNA. The fractional deuterium level at each time point is plotted as a function of the midpoint position of each peptide. The first and last residues of each peptide are indicated above each data set. See also Figure S3 and Table S1.

Discussion

To address the question of how well tertiary structure correlates with dynamics, we analyzed sliding clamp proteins, which have highly conserved tertiary structure at both the domain level and the whole protein level. These proteins are nearly identical in terms of their tertiary structure even though they have very low sequence similarity. The hydrogen exchange measurements revealed that dynamics were not conserved in these proteins. Furthermore, there was a large degree of heterogeneity of the dynamics within each clamp, despite the internal structural symmetry of these proteins. As expected, the proteins/domains with the highest sequence similarity were generally the most similar in their dynamics. For example, hPCNA and AtPCNA1 are 64% identical and showed similar levels of deuterium uptake; however, the two TK clamps are 54% identical and showed quite different uptake at all but the earliest time points.

The biological function of sliding clamps is directly related to their structure and dynamics. These clamp proteins must be opened to be loaded onto DNA (Hedglin et al., 2013) and protein dynamics may play a role in the opening of the clamp. There is structural evidence that T4 gp45 and PCNA from archaea adopt spiral open conformations when bound to their respective clamp loaders (Kelch et al., 2011; Miyata et al., 2005) and it has been suggested that the β clamp and PCNA from eukaryotes adopt similar conformations based on the spiral geometry observed in the crystal structures of the respective clamp loaders (Bowman et al., 2004; Jeruzalmi et al., 2001). This would require opening of the interface and twisting of the β-sheet structure on the outside of the clamps. The cooperative local unfolding of the β-strands that form the extended β-sheet in T4 gp45, the β clamp, and hPCNA should facilitate clamp loading and unloading by providing the protein with localized conformational entropy, lowering the energy barrier for transition to the open, perhaps twisted, conformation of the clamp. In addition, this local unfolding is direct evidence that there can be intrinsic disruption of hydrogen bonds that stabilize the subunit interface in the absence of the clamp loading machinery. It has been shown that the residence times of the β clamp and hPCNA on circular DNA in the absence of other proteins in vitro are 72 and 24 minutes, respectively (Yao et al., 1996). In the same study, T4 gp45 dissociated from DNA too quickly to be measured unless DNA polymerase was present to stabilize the clamp on DNA. The residence times on DNA roughly correlate with the reported oligomer dissociation constants for T4 gp45 (~250 nM), hPCNA (21 nM), and the β clamp (<60 pM) (Yao et al., 1996), as well as the half-lives of the local unfolding that we observed in the subunit interfaces (T4 gp45, ~5 min; hPCNA, 1 h; β clamp, ~3.5 h (Fang et al., 2011)), suggesting that spontaneous dissociation from DNA may be facilitated by local unfolding of β-strands in the subunit interface. The short half-life of the local unfolding we observed in both domains I and II at the interface of T4 gp45 is consistent with previous work showing that T4 gp45 has one open interface in solution (Alley et al., 1999; Millar et al., 2004).

We show here that sliding clamps are generally highly dynamic in the region responsible for most protein-protein interactions, which is the IDCL in PCNA or the hydrophobic cleft between domains II and III in the β clamp. This high degree of flexibility may contribute to the diversity of protein-protein interactions of sliding clamps by exposing unique binding motifs at various times, thereby allowing for unique contacts with binding partners in addition to the canonical interactions. Moreover, our data indicate that the IDCL of PCNA is partially stabilized, perhaps by transient secondary structures or interactions with the globular domains. These structures could provide unique binding surfaces for PCNA-interacting proteins. Indeed, structural studies have shown that the IDCL can adopt different conformations upon interactions with peptides that represent partner proteins, such as peptides from p21 (Gulbis et al., 1996) or flap endonuclease 1 (Sakurai et al., 2005).

The function of sliding clamps is dependent on their ability to slide on DNA, which is facilitated through interactions between DNA and positively charged residues located in α-helices lining the inner pore of the clamps. An NMR study recently demonstrated faster backbone amide deuteration for hPCNA than for yPCNA and that the central α-helices were more dynamic than the outer β-sheets (De Biasio et al., 2011). However, only 15 out of 260 amide resonances for hPCNA and 90 out of 257 amide resonances of yPCNA were clearly observed after one hour in D₂O, which was the first exchange time point of the experiment. Because HX MS allows us to determine deuterium uptake after only a few seconds in D₂O, we were able to determine the deuterium levels after labeling periods of less than one hour. We clearly observed that the outer β-sheets of yeast and human PCNA were more dynamic than the central α-helices at the shorter time points of our experiment (Figure 2) whereas, consistent with the NMR study, the central α-helices had relatively high deuterium levels after one hour of labeling and hPCNA was clearly more dynamic than yPCNA.

Interestingly, we observed a pattern of alternating stable and dynamic α-helices lining the inner pore of all the eukaryotic PCNAs as well as the β clamp correlating with the number of potential salt bridges of the residues in each α-helix, suggesting that this pattern is conserved and related to functional interactions with DNA. There is a strong evolutionary conservation of both positively and negatively charged residues in these α-helices and it has been suggested that salt bridges in the inner pore of the clamps may play a role in sliding on DNA by allowing the positively charged residues to alternate interactions with negatively charged clamp residues and the phosphate groups on the DNA backbone (Barsky et al., 2011). Moreover, hPCNA slides on DNA with two apparently different rates, with the slower mode consistent with PCNA following the DNA groove (Kochaniak et al., 2009). Thus, the alternating pattern in dynamics that we observed may be a consequence of the number and positions of the salt bridges. It is known that salt bridges contribute to stability of proteins, particularly the stability of proteins from thermophiles (Ladenstein and Antranikian, 1998). In fact, we observed a correlation between thermal stability and the percentage of charged residues (Asp, Glu, Lys, Arg, His).

The question remains as to why in this family of proteins the tertiary structures are so similar, yet protein dynamics and motions are so variable. It is possible that, while maintaining the basic ring-shaped structure required for enhancement of DNA polymerase processivity, the dynamics of sliding clamps from different species have diverged as the requirements for processivity, protein-protein interactions, and clamp recycling have changed during evolution. Structural fluctuations, and in particular local unfolding, provide conformational entropy to the protein structure, reducing the energy barrier for dynamical transitions such as clamp opening (and perhaps twisting) and DNA translocation. In addition, structurally heterogeneous and dynamic binding sites increase binding entropy and facilitate transient protein-protein interactions, which are important for biological systems that require multiple levels of regulation, such as the eukaryotic cell cycle. Several examples of local unfolding related to the function of proteins have been demonstrated using both theoretical and experimental work (Shan et al., 2013; Whitford et al., 2007). Another explanation may be that dynamics are conserved only upon full assembly of the complexes in which these clamps function. Only more detailed studies of complexes will reveal if this is the case, but as discussed above the regions of high dynamics tend to coincide with the binding sites for other proteins.

Overall, we find that regions of the clamps involved in protein-protein interactions, specifically the IDCL and hydrophobic PIP-binding cleft in PCNA and the corresponding site in the β clamp, tend to be highly dynamic. We also observe that a specific pattern of alternating dynamics is present in the DNA-binding pore, which could explain mechanisms of clamp sliding on DNA. We show here that in a family of proteins of highly conserved tertiary structure but with low sequence conservation, the dynamics of the clamp proteins vary widely. It remains to be determined how general these findings will be for other families of proteins with similar properties.

Experimental Procedures

Cloning, Expression, and Protein Purification

The genes encoding T4 gp45, human PCNA, and A. thaliana PCNA1 and PCNA2 were amplified by PCR from the plasmids pUC9 (Rush et al., 1989) (a generous gift from Dr. William Konigsberg, Yale University), pAVR38 (Vidal et al., 2004) (a generous gift from Dr. Roger Woodgate, NICHD/NIH), and U21468 and U86753 (Arabidopsis Biological Resource Center at The Ohio State University), respectively. Each gene was cloned into plasmid pET11T using the NdeI and BamHI sites (Nguyen et al., 1993), followed by transformation of E. coli strain BL21(DE3). For the overexpression of yeast PCNA and the T. kodakaraensis PCNAs, E. coli Rosetta(DE3) pLysS cells were transformed using plasmids pMM119 (Ionescu et al., 2002) (a generous gift from Dr. Michael McAlear, Wesleyan University), and pET-TK0535 or pETTK0582 (Ladner et al., 2011) (a generous gift from Dr. Zvi Kelmam, Center for Advanced Research in Biotechnology, UMBC), respectively.

Cell culture conditions, protein overexpression, and lysis procedure were as described (Fang et al., 2011). The His-tagged TK PCNA proteins were purified as described (Ladner et al., 2011). Human PCNA, yeast PCNA, A. thaliana PCNAs, and T4 gp45 were purified mostly as described (Fang et al., 2011), with details given in Supplemental Information. The final purity and mass of all proteins was verified by electrospray mass spectrometry (LCT-Premier, Waters).

Deuterium Labeling

Protein stock solutions at 40-100 μM in 20 mM Hepes, 50 mM NaCl, 0.1 mM EDTA, and 1 mM TCEP (pH 7.5) were prepared and HX reactions were initiated by diluting the stock 18-fold (v/v) with D₂O buffer (20 mM Hepes, 50 mM NaCl, 0.1 mM EDTA, and 1 mM TCEP, pD 7.5) at 21 °C. The isotope exchange reaction was quenched after various periods of time (10 s to 4 h) by 1:1 dilution using ice cold sodium phosphate buffer (pH 2.1) containing 6 M guanidine HCl. The final pH of the quenched samples was 2.6. Samples were diluted 1:1 using ice cold 0.1% formic acid prior to injection into the LC system.

Mass Analysis

Deuterated protein samples were digested online in a self-packed immobilized pepsin column followed by separation using a Waters nanoACQUITY system with HDX technology for UPLC separation (Wales et al., 2008) and mass analysis using a Waters Q-Tof Premier mass spectrometer. We note that prior experiments with the β clamp (Fang et al., 2011) were performed with digestion in solution and therefore the reported deuterium levels and identified peptic peptides are slightly different. Protein samples (30-40 pmol) were injected onto a 2.1 mm × 50 mm stainless-steel column packed with immobilized pepsin (Wang et al., 2002) on POROS-20AL beads (PerSeptive Biosystems) with a flow rate of 100 μL/min in 0.1% formic acid at 15 °C. Peptides were trapped on a VanGuard Pre-Column (2.1 mm × 5 mm, ACQUITY UPLC BEH C18, 1.7 μm) for 4 min. The trap was then placed in-line with an ACQUITY UPLC BEH C18 1.7-μm 1.0 mm × 100 mm column (Waters) and an 8-40% gradient of acetonitrile over 7 min at a flow rate of 40 μL/min was used to separate the peptides at 0 °C. Formic acid (0.1%) was added to both mobile phases to maintain pH 2.5. Mass spectra of peptides were acquired in positive ion mode using electrospray ionization (ESI) on a Waters Q-Tof Premier mass spectrometer under optimized conditions. Spectra were acquired over an m/z range of 50-1700 and mass accuracy was maintained through continuous lock-mass correction using Glu-fibrinogen B peptide standard (Sigma).

Peptic peptides of triplicate undeuterated control samples were identified using a combination of accurate mass and MS^E, aided by Waters IdentityE software (Geromanos et al., 2009). Only peptides identified in at least two replicates were included in the analysis. Mass spectra of undeuterated and deuterated peptides at different time points were extracted and analyzed in Waters DynamX software. Regardless of whether peptides displayed EX1 or EX2 kinetics, the centroid mass of the entire isotope distribution was used to calculate the deuterium level. Relative deuterium levels were calculated by subtracting the average mass of the undeuterated control sample from that of the deuterated sample and plotted as a function of time. The data were not corrected for back-exchange and are therefore reported as relative (Wales and Engen, 2006a). The standard deviation of the relative deuterium level was generally less than 0.3 Da. The relative percent deuterium incorporation was mapped onto the protein structures using selected peptides to achieve maximum coverage of the primary structure.

Thermal Shift Assay

Samples containing 10 μM protein in 15 μL of assay buffer (20 mM Hepes, 50 mM NaCl, pH 7.5) and a final concentration of 25X Sypro Orange (Invitrogen) were assembled in 96-well PCR plates (Applied Biosystems) and analyzed using a CFX 96 Real-Time System (Bio-Rad) using the FRET channels. The temperature was raised from 20 °C to 100 °C with 0.2 °C increments and 10 s dwell times.

Structure and Sequence Analysis

Coordinate files for protein structures were downloaded from the Protein Data Bank. Structural alignments were performed using DaliLite (Hasegawa and Holm, 2009) and PROMALS3D (Pei et al., 2008). Structural analysis and images were prepared using PyMol (Schrödinger).

Molecular Dynamics Simulations

Protein structures from the Protein Data Bank were prepared for MD simulations using Maestro software (Schrödinger). MD simulations were performed using GROMACS software (version 4.5.5) (Hess et al., 2008). CHARMM22 force field (MacKerell et al., 1998) with CMAP correction (Mackerell et al., 2004) was employed to model protein atoms. The protein was immersed into a cubic box extending at least 10 Å from the protein. The rest of the box was filled with explicit TIP3P water molecules (Jorgensen et al., 1983). Sodium cations were added to neutralize charge of the protein and periodic boundary conditions were imposed for the simulations. Before running the simulation, the system was minimized for 1000 steepest descent iterations and then equilibrated for 20 ps with non-hydrogen protein atoms restrained. To investigate formation of salt bridges and hydrogen bonds, a 2 ns MD simulation was performed with a 2 fs time step. Pressure coupling was applied using the Parrinello-Rahman algorithm (Nose and Klein, 1983; Parrinello and Rahman, 1981), and temperature coupling was done with the V-rescale algorithm (Bussi et al., 2007). The particle-mesh Ewald (Darden et al., 1993) algorithm was used for long-range electrostatic interactions with a cutoff of 9 Å. A 9 Å cutoff was used for Van der Waals interactions. The MD simulation output between 1 ns and 2 ns (500 frames) was taken for further analysis of the hydrogen bonding/salt bridge network and its stability. The statistics were collected after 1 ns of MD simulation because it generally took 0.5-1 ns for the clamps to relax from the original structure and to effectively remove the artifacts of the crystallography and of the starting setup of the calculations. The hydrogen bonding in the resulting trajectories was investigated using g_hbond tool (in GROMACS), and salt bridges were explored using a plug-in in the VMD program (Humphrey et al., 1996).

Highlights.

Proteins with conserved tertiary structures display different dynamics in solution

Local unfolding was detected in T4 gp45, E. coli β clamp, and human PCNA

Clamps display a pattern of alternating dynamics in their inner pore helices