2.1. General MD simulation setup

MD simulations were conducted using the GROMACS 5.0.5 suite [36] under a constant temperature and pressure (NPT) ensemble. The CHARMM36 force field [37] was employed to characterize both the protein and DNA, alongside the TIP3P water model [38]. The system was solvated in a cubic box, ensuring a minimum distance of 10 Å from the box boundaries, with periodic boundary conditions applied in all directions. The equation of motion was integrated with a time step of 2 fs. The temperature was maintained at 300 K using the Berendsen thermostat, and the pressure was controlled isotropically at 1 bar using the Parrinello-Rahman barostat [39]. Neighbor searching was performed with a cutoff radius of 10 Å, and the non-bonded pair list was updated every 40 steps. Short-ranged interactions were truncated beyond 10 Å, with dispersion corrections applied. Long-range electrostatic interactions were calculated using the Particle Mesh Ewald (PME) method [40] with a grid spacing of 1.6 Å and interpolation of order 4. Using the SETTLE [41] and LINCS [42] algorithms, covalent bonds were constrained to their equilibrium geometries. The steepest decent method was used to minimize the energy of the system. The energy was minimized for 10,000 steps to remove unfavorable contacts from water and ion placement. After energy minimization, the systems were equilibrated in the NVT ensemble at 300 K with positional restraints for 10 ns to ensure stability, followed by another 10 ns of equilibration under the NPT ensemble at 1 bar and 300 K. The production run was performed for 1 μ s (1000 ns) for each system under unrestrained NPT conditions at 300 K and 1 bar. The coordinates are recorded every 2 ps for subsequent analysis.

For the structural analysis of the systems, we employed various built-in GROMACS functions [36], including gmx rms, gmx gyrate, gmx rmsf, and gmx hbond, on the trajectories generated for each DNA pol β complex system after MD simulations. These functions were used to calculate the root mean square deviation (RMSD), radius of gyration (R_g), root mean square fluctuations (RMSF), and the number of hydrogen bonds (H-bonds). The criteria for H-bonds were set as a donor-acceptor distance d ≤3.5 Å, and the angle between the donor and acceptor is >30⁰. Matplotlib v3.3.2 [43] was used to plotting the results whereas, the structural images were produced using visual molecular dynamics (VMD) [44] and PyMOL [45]. Fig S1 (in the supplementary information) shows the flow chart of the MD simulation.

2.2. Principal component analysis

Principal component analysis (PCA) is a mathematical technique used to reduce the dimensionality of large datasets by transforming a large number of variables into a smaller set while preserving the original information. In the context of molecular dynamics (MD) simulations, PCA, also known as essential dynamics [46], is employed to analyze information about the direction of motions of the protein residues from a set of conformations produced by MD simulation. PCA works by diagonalizing the covariance matrix, where the elements are represented by . This covariance matrix is constructed from the Cartesian coordinates of the structures obtained from the MD simulations:

(1)

where i and j represent all possible pairs of 3N Cartesian coordinates, where N is the number of atoms. In this work, the PCA was performed using atoms of the protein and phosphate atoms of DNA. The covariance matrix, constructed from the atomic fluctuations of both the protein and DNA, was diagonalized. The resulting eigenvalues and eigenvectors from PCA can be used to describe the motion of the system. The first 20 eigenvectors were examined to access their cosine content. Of these, the two eigenvectors designated as PC1 and PC2, with cosine content less than or equal to 0.1, were used. These principal components were utilized to generate the free energy landscape (FEL) using the gmx sham function in GROMACS [36], which computes the minimum free energy configuration based on PC1 and PC2. The FEL map was subsequently plotted using MATLAB [47].

2.3. Perturbation response scanning

Perturbation response scanning (PRS) analysis [48,49] was conducted to identify the most critical and sensitive residues of methylated DNA polymerase β, using the anisotropic network model (ANM). In the ANM approach [50], the protein is represented as a network of nodes corresponding to the positions of atoms for each residue, with inter-residue interactions modeled by harmonic springs within a threshold distance of 12 Å. PRS, based on linear response theory, evaluates the impact of perturbing individual residues by measuring changes in the protein’s dynamics. In PRS, each residue in the network is sequentially subjected to a force of fixed magnitude, and the overall network response is calculated using Hooke’s law (ΔR = H⁻¹F), where F is the applied force and H is the Hessian matrix obtained from ANM. The displacement of all residues, ΔR⁽ⁱ⁾, is computed for each residue under the applied force, where i ranges from 1 to N. The PRS analysis focuses on the first 20 slowest non-zero eigenvalues of the ANM modes, which characterize the most energetically favorable collective motions in the protein structure. This method allows for the identification of key residues in DNA polymerase β.

3.1. Structural stability and flexibility of the methylated systems

Methylation is known to enhance DNA polymerase activity by improving DNA binding. Nevertheless, the understanding of its impact on the stability of the DNA polymerase β complex is not fully elucidated. To evaluate the structural stability and simulation convergence under methylation, we computed mean square deviations (RMSDs) of atoms of protein and backbone phosphates for DNA from the initial structure. Fig 1B depicts the time evolution of RMSD of atoms of protein of DNA polymerase for meR83, meR152, and meR83,152. Initial fluctuations are observed in all systems but eventually stabilize after 100 ns, indicating simulation convergence. For meR83,152, there is a sudden increase in RMSD value (4.4 Å) at 510 ns, before reaching a plateau at 550 ns. A small dip occurs at 790 ns followed by an increase, maintaining the same state up to 1000 ns. In meR152, stability is achieved at 50 ns, while meR83 displays a jump in RMSD at 400 ns (3.8 Å), gradually returns to its initial state within a few ns, and remains stable. The figure illustrates that methylation at both R83 and R152 sites enhances system flexibility, while methylation at R152 drives the protein into a more compact state compared to the WildType (WT). Mean RMSD values for meR83, meR152, and meR83,152 are 1.78 (0.29) Å, 1.75 (0.19) Å, and 2.4 (0.56) Å, respectively, while the average WT RMSD is 1.93 (0.53) Å. Fig 1C shows the time evolution of RMSD of backbone phosphate atoms of DNA fragments bound to DNA polymerase β. Similar to the protein RMSDs, the DNA phosphate backbone atoms exhibit a similar behavior. The average RMSD value of the DNA fragment for the WT system is 2.86 (0.95) Å, whereas, for the meR83, meR152, and meR83,152, the average RMSD value is 3.51 (0.76) Å, 2.98 (0.67) Å, and 5.09 (0.98) Å, respectively. A higher value was observed in the DNA fragment for all methylated systems compared to the WT. This indicates that the methylated sites interact with the backbone DNA phosphate and make it fluctuate more.

As noted above, when DNA is present, the DNA polymerase β complex demonstrates enhanced stability. A pertinent inquiry arises regarding how the methylated residues interact with various domains and subdomains of pol β in the absence of DNA. To address this question, we conducted a microsecond-long MD simulation for these systems without DNA fragments. Fig S2 (in the SI) illustrates the time evolution of RMSD of atoms of DNA polymerase protein for meR83, meR152, and meR83,152. Unlike the pol β complex in the presence of DNA, the absence of DNA results in a notable jump in RMSDs. In the absence of DNA, the mean values for the methylated systems are 4.3 (0.90) Å, 10.98 (3.8) Å, and 15.01 (4.7) Å, respectively. This unequivocally indicates that in the absence of DNA, methylated residues interact differently with the protein, rendering the system less stable.

To explore the impact of methylated sites on the structural compactness of DNA polymerase β, we have calculated the radius of gyration (). Fig 1D shows the probability distribution of of meR83, meR152, and meR83,152. The probability distribution of both meR83, and meR152 is sharply defined, yet the peak position differs. In contrast, the probability distribution for meR83,152 is broader. These findings suggest that DNA polymerase β undergoes a conformational transition and transits into a more closed state for meR152 as a result of methylation.

Methylation at distinct sites on the DNA polymerase β induces structural changes. To identify the subdomains most affected by the methylation, we have calculated the root mean square fluctuations (RMSF) for atoms of protein residues and backbone phosphate atom of DNA. The RMSF value shows the average fluctuation of each considered atom over the total time of the simulation. Fig 2A illustrates the difference in the RMSF value of the protein for meR83 (blue dotted line), meR152 (red dashed line), and meR83,152 (green solid line) with respect to the WT, while the black circle indicates methylation sites. The figure reveals that methylation enhances RMSF values in certain subdomains while reducing them in others compared to the WT. For meR152, RMSF values of subdomains decreased, whereas methylating R83 significantly enhances residues that correspond to the catalytic subdomain. In both meR83 and meR152, the RMSF values for the Lyase domain, DNA binding subdomain, and N-terminal base pairing subdomain decrease, indicating a transition of DNA pol β to a more compact state. Simultaneous methylation at both R83 and R152 increases the RMSF value of the residues linking the Lyase domain and DNA binding subdomain, as well as the linker region between the DNA binding subdomain and catalytic subdomain. Additionally, the RMSF value of the residues corresponds to the catalytic subdomain, and the N-terminal base pairing domain also increased. On the contrary, in the absence of DNA, notable changes occur due to methylation. Fig S3 (in SI) displays the RMSF value of the protein for meR83, meR152, meR83,152, and WT in the absence of the DNA. In the absence of DNA, methylating both sites increases the RMSF value almost 3-fold compared to in the presence of DNA. We also looked at the impact of methylation on the DNA fragment. Fig 2B demonstrates the difference in RMSF value of the DNA fragment for meR83, meR152, meR83,152 from WT. While methylation at the individual sites has a modest impact on DNA fragment RMSF, simultaneous methylation at both R83 and R152 results in a significant increase in fluctuations, possibly attributed to methyl group interactions with the DNA backbone atoms.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Difference in the Root Mean Square Fluctuations (RMSF) value of

atoms of DNA polymerase β concerning WT. (A) meR83 (blue dotted line), meR152 (red dashed line), and meR83,152 (green solid line). The rectangle shown in the horizontal panel reflects the DNA polymerase β domains and subdomains and are colored as follows: Lyase domain (blue) and three subdomains: D (orange), C (red), and N (green). (B) Difference in P atom of DNA backbone of DNA polymerase β concerning WT: meR83 (blue dotted line), meR152 (red dashed line), and meR83,152 (green solid line).

https://doi.org/10.1371/journal.pone.0318614.g002

The crystal structure of DNA polymerase β revealed that the enzyme adopts a closed structure facilitated by a hydrogen bond (H-bond) between S44 (located in the Lyase domain) and E335 (lies in the N-terminal base paring subdomain). This H-bond is absent in experimentally resolved structures where DNA was absent. This observation implies that both DNA and this H-bond interaction play a vital role in stabilizing the enzyme in the closed state. To analyze how methylation influences this H-bond, we looked at the H-bond occupancy using our MD trajectory. The H-bond occupancy is defined as the fraction of time during which the residue S44 forms a H-bond with the S335 residue. The GROMACS hbond tool was employed to compute the H-bond occupancy with a threshold distance of 3.5 Å and an angle cutoff of 30⁰. Results indicate that the H-bond was present in the WT system for 80% of the simulation duration, while methylation does impact the H-bond occupancy. Specifically, the H-bond occupancy for the methylated systems is 75% (meR83), 55% (meR152), and 40% (meR83, 152), respectively. The average donor-acceptor distance between S44 and E335 for all methylated systems compared to WT is presented in Table S1 (in SI). The reduction in H-bond occupancy for the methylated system compared to WT is likely associated with the presence of the methyl group.

Furthermore, an important question arises: does methylation impact the polar interactions between different domains and subdomains? To address this, we calculated the number of H-bonds formed between the methylated systems and the various domains of DNA pol β complex. We found the number of H-bonds varied throughout all three simulated systems. Specifically, for meR83, the Lyase domain formed an average of 2.3 (0.7) H-bonds with the methylated residues, while the D-subdomains formed 0.2 (0.006) H-bonds, and no H-bonds were observed with other subdomains. For meR152, an average of 1.3 (0.4) H-bonds were formed with the C-subdomain, and no H-bonds were detected between different subdomains of the enzyme. In the case of meR83,152, the Lyase domain formed an average of 1.9 (0.9) H-bonds, the D-subdomain formed 0.01 (0.002) H-bonds, and the C-subdomain formed 1.6 (0.6) H-bonds. Interestingly, when both R83 and R152 were methylated, the average number of H-bonds meR83,152 formed with the Lyase domain decreased, while the H-bonds formed between meR83,152 and the C-subdomain increased. Additionally, we explored how methylation affects the DNA fragment by calculating the center of mass (COM) distances between methylated systems (meR83, meR152, and meR83,152) and DNA. The findings revealed that DNA comes closer to the methylated residues compared to the WT. The average COM distance between meR83, meR152, and meR83,152 are 32.6 Å (1.1), 34.7 Å (1.1), and 26.3 Å (1.4), respectively, while, for WT is 34.9 Å, 35.8 Å, and 36.3 Å.

Experimental evidence suggests that methylation affects the DNA binding property of the DNA polymerase β complex. This raises an additional question: does methylation impact the interaction between DNA and different subdomains of the enzyme? To address this, we computed the number of H-bonds formed between DNA and various subdomains. For meR83, meR152, and meR83,152, the average number of H-bonds between DNA and D-subdomain is 2.9, 3.0, and 3.1; between DNA and C-subdomain is 4.6, 4.9, and 5.9; and between DNA and N-subdomain is 4.2, 4.8, and 5.2. In contrast, for WT, the average number of H-bonds between DNA and D-subdomain is 2.7; DNA and C-subdomain is 4.6; DNA and N-subdomain is 2.7. These findings suggest that methylation causes DNA to move closer to the polymerase domain due to a significant increase in the number of H-bonds compared to WT. The time evolution of H-bonds between DNA and different subdomains for all three methylated systems is depicted in Fig S4 (in SI). These results underscore the crucial role of H-bonds in the structural transitions observed in all three methylated systems, which will be discussed in the following section.

4.1. New salt-bridge formed due to the methylation

PCA and H-bonds analysis reveals that methylated residues induced structural transitions in DNA pol β complex. Consequently, methylated residues are expected to form new salt bridges with neighboring residues and different subdomains. We conducted a salt bridge analysis in the presence and absence of DNA fragments. Table 1 shows the list of salt bridges formed between the methylated residues and the different domains/subdomains of DNA polymerase β.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. List of the salt bridges formed between the methylated residues and the different domains/subdomains of DNA polymerase β. The percentage of time each salt bridge was observed during the simulation is shown in brackets.

https://doi.org/10.1371/journal.pone.0318614.t001

In the presence of DNA, meR83 forms a new salt bridge with E71, which was not present in the absence of DNA, indicating that the presence of DNA stabilizes the Lyase domain. Compared to meR83, when both residues are methylated, the salt bridge formed with E71 is broken. This occurs because meR152 induces a twist in the helical region of the Lyase domain, causing the salt bridge between meR83 and E71 to break. This could indicate that methylation may occur sequentially with meR83 occurring first to stabilize the lyase domain, followed by meR152 to induce salt-bridge breakage to relieve this stabilization. In meR83,152, with DNA, the methylated residues form four salt bridges with residues located in the Lyase domain or C-subdomain, whereas, without DNA, it forms seven salt bridges. Notably, salt bridges formed with residues located in the D- and N-subdomains are absent in the presence of DNA, likely due to the DNA fragments interacting with these subdomains and increasing the distance from the methylated residues. Fig S7 (in SI) illustrates the schematic representation of salt bridges formed in both the presence and absence of DNA for the meR83,152 system. To investigate the effect of methylation on salt bridge formation, we calculated the average O-N distance between the methylated arginine and E75, E86, E153, and E186 residues in the presence of the DNA and compared with the WT. We observed that methylating residues decrease the distance between these residues. In the meR83,152 system, the average distance between meR83 and E75 is 9.4 Å, between meR83 and E86 is 5.9 Å, between meR152 and E153 is 8.1 Å, and between meR152 and E186 is 7.3 Å, while for the WT, the average distance between meR83 and E75 is 7.8 Å, between meR83 and E86 is 7.5 Å, between meR152 and E153 is 9.3 Å, and between meR152 and E186 is 8.0 Å, respectively. These results suggest that in the methylated system, the Lyase domain expands while the D-subdomain and C-subdomain transition to a more compact state due to the interaction with DNA.

To explore how the methylated residues influence the polymerase domain, we examined the interaction of the methylated system with active sites (D190, D192, D256) with surrounding residues. Our analysis revealed that methylation induces the formation of a new salt bridge between D190 and R254 which is absent in the WT system. This interaction leads to the C-subdomain approaching the D-subdomain, resulting in a more rigid polymerase domain compared to the WT.

4.2. Cross-correlation map shows the correlation enhanced between different subdomains upon methylating both R83 and R152

To characterize how the methylation affected the coupling between the distant subdomain of the DNA pol β complex, we calculated the cross-correlation between the positions of different residues in 3-dimensional space. This method is often used to study the allosteric effect on ligand binding [51–54]. Conventional implementations (Pearson Correlation Coefficient) compute the dynamic cross-correlation map (DCCM) of the position vector atoms of amino acids [46]. However, DCCM ignores correlated motion in orthogonal directions [55]. This problem can be avoided by using linear mutual information (LMI) based cross-correlation metric, which we use in this study. The cross-correlation matrix elements are, is defined as

(2)

where d is equal to 3. The coefficient of is zero for fully uncorrelated motions and assumes values up to 1 for fully correlated motions. The is the LMI calculated as

(3)

where H is closely related to the Shannon entropy,

(4)

where, i and j corresponds to residue i and residue j, and are the 3D Cartesian vector of atomic coordinates of the corresponding for proteins, P atoms for DNA fragments, whereas and shows, respectively, the marginal probability density for and the joint probability density of and [51,56]. The PCA analysis reveals that conformational transitions in DNA pol β are influenced by methylation. Comparing the LMI cross-correlation among subdomains aids in identifying residues exhibiting distinct behavior across various DNA pol β conformations. Notably, the difference in correlation between the WT state and methylated states can pinpoint the residues gaining or losing contact due to the methylations, consequently initiating the conformational transitions. In our subsequent analysis, we incorporated an unbiased trajectory, comparing the correlation heat maps of all residues in the WT form with the three methylated systems (meR83, meR152, and meR83,152), to understand the coupling between the subdomains and protein residue fluctuations as a consequence of methylation.

Fig 4 illustrates the cross-correlation map for WT (A), meR83 (B), meR152 (C), and meR83,152 (D). The LMI correlation is notably higher for meR83,152 compared to WT, as indicated by the increased prevalence of blue and green colors (Fig 4A and 4D). Both meR83 and meR152 exhibit weak intra-subdomain correlation, with meR152 displaying almost no correlations between the D-subdomain and N-subdomain. In contrast, methylating both R83, and R152 enhances both inter- and intra-subdomain correlations. Notably, a substantial increase in LMI values between the protein domains and DNA fragments upon methylation is observed, consistent with the experimental findings that methylation enhances DNA binding.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. (A-D) Cross correlation matrices of the fluctuations of the atoms and phosphates atoms of DNA polymerase

β for four systems (A) WT; (B) meR83; (C) meR152; (D) meR83,152) computed using LMI. (E-G) The difference in cross-correlation matrix element for the (E) meR83, (F) meR152, and (G) meR83,152 with respect to the WT. Highly correlated blocks (G) indicate that methylating both residues shows a signature of long-distance correlation motions.

https://doi.org/10.1371/journal.pone.0318614.g004

Fig 4(E-G) displays the change in correlation coefficient (Δ correlation) relative to the WT for the three methylated systems. Residues gaining correlations are depicted in blue color while those losing correlations are shown in red color. It is evident from the figure methylating R152 site results in residues corresponding to different domains and sub-domains losing correlations, whereas meR83 shows inter-subdomain loss of correlation between various subdomains, especially Lyase domain and N-subdomains and C- and N-subdomain. Methylating both R83 and R152 demonstrates a gain in correlation between different domains and subdomains. Intriguingly, methylation increases the correlation between the protein and DNA fragment, except for the meR152 system, possibly due to the location of R152. Additionally, in the case of meR83,152, a significant gain in correlation is observed in the linking region of the C- and N-subdomain, which is quite far from the methylated sites. This clearly shows the methylation-induced long-range correlation in the DNA pol β complex.

4.3. Structural Network Analysis

As seen above, methylation induced structural transition in the DNA polymerase β complex. To gain more profound insight related to the functional movement and information exchange mechanism, we built a protein connectivity graph network model (GNM). This methodology involves a weighted network whose nodes represent the atoms of the amino acids in the protein and phosphate atom in DNA, and the correlation between them is the edges connecting the nodes. We then calculated the betweenness centrality (BC), a graph theoretical measure that provides a way to quantify the amount of information that flows via the nodes and edges of a network. If a node i acts as a bridge between two other nodes along the shortest path joining them, then the BC of a given node i is given by the following equation:

(5)

where represents the total number of shortest paths joining nodes a to b, out of which path pass through node i [52]. By analyzing the BC values, we can identify the residues that are involved in the information-propagating pathway. Centrality analysis has been extensively employed to determine the functionally significant residues in communication pathways, metabolic networks, disease networks, and allosteric communications [57–60]. The change in BC has been used to explore the dynamics of the spike protein [61]. In this work, we used the difference in BC as a metric to identify key residues that gain or lose relative importance along the information pathway due to methylation. The difference in the normalized BC is measured by comparing the methylated states with the WT (i.e., BC^meR - BC^WT, for every residue and nucleotide of DNA polymerase β complex) from our all-atom trajectory with explicit solvation. Fig 5(A-C), displays the difference in BC value per residues for the methylated system in comparison to the WT. The result indicates that some residue loss and others gain importance along the information pathways.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. (A-C) The values of the difference in betweenness centrality (BC) with respect to the WT state plotted as a function of residue index for three systems(A), meR83; (B), meR152; (C), meR83,152.

The horizontal lines correspond to the value ±0.1. The data are colored according to domain/subdomain color as shown in Fig 1A. Data shows a large change in normalized BC value due to the methylation. Residues with the largest positive change in normalized BC value for (D) meR83, (F) meR152, and (H) meR83,152. Residues with the largest negative change in normalized BC value for (E) meR83, (G) meR152, and (I) meR83,152. The DNA polymerase β subdomains corresponding to those residues are also mentioned next to the residue number.

https://doi.org/10.1371/journal.pone.0318614.g005

For meR83, and meR152 the residues with significant (e.g., >0.1) change in the BC value are mostly from the D-subdomain or C-subdomain, whereas, for meR83,152 the significant change in BC value is either from the D-subdomain or the N-subdomain (Fig 5D-I). This suggests that allosteric information flows through Lyase and the C-subdomain for meR83, whereas for meR152, and meR83,152, the information flows through the Lyase and the D-subdomain. Any mutation in these regions can break the allosteric network and affect the functionality of DNA pol β complex. This further supports that the methylation may occur sequentially with meR83 occurring first to stabilize the lyase domain and hasten DNA repair, followed by meR152 to induce salt-bridge breakage and relieve this stabilization.

In culmination, the most interesting aspect is the striking large change in BC of the residues that are distant from the methylated sites in the 3D structure. Significant gain or loss of BC is observed in residues R253, R149, and L19, whereas, for meR83.152, it is D17, L19, T20, E21, K249. There is no significant loss or gain is observed in the meR152 system. For meR83, 152, the first 5 residues correspond to the Lyase domains whereas K249 corresponds to the loop region of the catalytic subdomain. For the meR83 system, the residue R253 corresponds to the β -sheet region of the catalytic subdomain, and R149 corresponds to the linker region of the DNA binding subdomain and Lyase domain. These results suggest that methylation has a long-range effect and not just on the methylation target site in the Lyase domain, the effect of this perturbation felt by the residues corresponds to the catalytic subdomain.

4.4. Methylation R83, R152, and R83,152 together changes the effector/sensitivity profile of DNA polymerase β complex

Centrality network reveals that methylation modulates the information pathways of DNA polymerase β complex. To investigate the impact of methylation on the residues responsible for the long-range transmission signal, we used the perturbation response scanning (PRS) method on the meR83, and meR83,152 systems. Anisotropic network model (ANM) from the representative structures (every 10 ns interval) of meR83, and meR83,152 was constructed, representing the protein as elastic mass and springs networks. The PRS map was generated using the covariance matrices from the first 20 nonzero ANM frequency modes. These lower frequency modes reveal the collective changes in structure close to the native state. The PRS method (details are given in the method section) probes the response of each residue to perturbation in every other residue. Fig 6 displays the PRS maps of meR83 (A) and meR83,152 (B), drawn using the MD trajectory. The representative structures were taken over every 10 ns interval resulting in a total of 100 representative structures. The present map is the averaged PRS map generated using 100 representative structures. Each element in the PRS matrix represents the response of residue j to a perturbation on residue i. The rows of the PRS map provide information about the influence of a particular residue in transmitting signals when subjected to the unit perturbation, while the column provides information about the sensitivity of a given residue to those signals. The effector residues (peaks in the right ordinate bar plot) and sensor residues (peaks in the upper abscissa bar plot) are colored by domain/subdomain identity in the panel.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Perturbation response analysis identifies the highly influential and sensitive residues that likely propagate allosteric signals due to the methylation in meR83 and meR83,152.

PRS maps ((A) meR83 and (B) meR83,152) indicate the strongest perturbation response sites as shown by the color map (see the scale on right). The peaks in the curves along the axes indicate the effectors (left ordinate) and sensors (upper abscissa). DNA polymerase β subdomains are color-coded along the lower abscissa highlighting lyase (blue), thumb (orange), palm (red), and finger subdomains (green), as mentioned in Fig 1A.

https://doi.org/10.1371/journal.pone.0318614.g006

The results show that the perturbation response of the DNA pol β remains relatively unchanged due to the methylation (meR83, and meR83,152), but specific regions show significant change upon methylation. Both effector and sensor residues signal enhanced compared to WT (Fig S8 in SI). In the case of meR83, there is a strong enhancement in sensitivity for E247 (C-subdomain) and T304 (N-subdomain), but the signal in other domains is relatively weak. However, methylating both R83 and R152, the sensitivity of E247 is suppressed and a new signal emerges in the Lyase domains (see Fig 6B). Upon methylation, the most affected sensors are clustered in the N-subdomain and Lyase domain, while for WT, the most affected sensors clustered in the C-subdomain. The effectors are clustered in the Lyase domain for meR83, while methylating both a few new signals emerge in the D-subdomain (L100) and N-subdomain (I319). Our results demonstrate that the methylation has a long-range effect, R83 lies on the Lyase domain and R152 likes in the D-subdomain but the strong effector and sensor signal emerges in the C-subdomain and N-subdomain compared to the WT.