Dynamics of preferential substrate recognition in HIV-1 protease: redefining the substrate envelope

Abstract

Human immunodeficiency virus type 1 (HIV-1) protease (PR) permits viral maturation by processing the gag and gag-pro-pol polyproteins. HIV-1 PR inhibitors (PIs) are used in combination antiviral therapy but the emergence of drug resistance has limited their efficacy. The rapid evolution of HIV-1 necessitates consideration of drug resistance in novel drug design. Drug-resistant HIV-1 PR variants no longer inhibited efficiently, continue to hydrolyze the natural viral substrates. Though highly diverse in sequence, the HIV-1 PR substrates bind in a conserved three-dimensional shape we termed the substrate envelope. Earlier, we showed that resistance mutations arise where PIs protrude beyond the substrate envelope, because these regions are crucial for drug binding but not for substrate recognition. We extend this model by considering the role of protein dynamics in the interaction of HIV-1 PR with its substrates. We simulated the molecular dynamics of seven PR-substrate complexes to estimate the conformational flexibility of the bound substrates. Interdependence of substrate-protease interactions might compensate for variations in cleavage-site sequences and explain how a diverse set of sequences are recognized as substrates by the same enzyme. This diversity might be essential for regulating sequential processing of substrates. We define a dynamic substrate envelope as a more accurate representation of PR-substrate interactions. This dynamic substrate envelope, described by a probability distribution function, is a powerful tool for drug design efforts targeting ensembles of resistant HIV-1 PR variants with the aim of developing drugs that are less susceptible to resistance.

Figure 1. Contact potentials between the protease and seven substrates are well represented by crystal structures

For each protease residue contacting at least one substrate, its potential energy of vdW contact is averaged over seven PR-substrate crystal structures (black bars), and their respective MD trajectories (light gray bars). Each panel shows one monomer of the protease. Stars represent PR drug-resistance mutation sites according to the Stanford Database of HIV Drug Resistance. Most of these sites do not make extensive contacts with the natural substrates, suggesting that these residues are more important for inhibitor binding than substrate recognition. A few of these sites contact natural substrates and have been reported to co-evolve with natural substrates, i.e., D30N/N88D with p1-p6, I50V with both NC-p1 and p1-p6, V82A with NC-p1.

Figure 2. Protease structures have similar interaction profiles although they are in complex with non-homologous substrates

(A) Unprimed monomer (B) Primed monomer.

Figure 3. Substrates maintain a consensus vdw interaction potential with the protease independent of sequence

(a) Despite low sequence homology, substrates have an optimal total vdW interaction with the PR. (b) The interaction potential between the PR and substrate differs by the individual contribution of substrate residues P4 to P4’.

Figure 4. Fluctuations in the vdW interaction potential within/among substrate residues are negatively correlated, revealing their interdependency

The vdW contact potential of each substrate residue with the PR was calculated over time. Fluctuations in this potential with respect to every other residue in the same substrate were estimated by computing Pearson correlation coefficients. The correlation coefficients are contoured for each substrate and color-coded from red (fully correlated) to blue (fully anti-correlated). (A) MA-CA, (B) CA-p2, (C) p2-NC, (D) NC-p1, (E) p1-p6, (F) RT-RH, (G) RH-IN.

Figure 5. Backbone hydrogen bonds are more conserved across varied substrates and more stable than the side-chain hydrogen bonds

Frequency of occurrence of hydrogen bonds during the simulations between any protease atom and substrate backbone (A) and substrate side-chains (C), and the hydrogen bonds observed in the crystal structures between any protease atom and substrate backbone (B) and substrate side-chains (D). The side-chain hydrogen bonds are shown by residue (B, D) for the crystal structures and simulations, respectively. In panels A and C, the hydrogen bonds are color-coded with respect to the time percentage of occurrence throughout the simulations with red being 100%. Only the hydrogen bonds that existed more than 50% of the time were shown, below 50% occurring bonds are shown in white. In panels B and D, gray indicates a hydrogen bond observed in the crystal structure and white shows that the bond does not exist in that structure.

Figure 6. Interdependency within substrate residues revealed by cross-correlations of mean-square fluctuations between protease and substrate residues

Atomic positional fluctuations are calculated for all residues in the PR-substrate complex. How one PR residue fluctuates with respect to a substrate residue was estimated by computing cross-correlations (see Methods). The correlation coefficients are contoured and color-coded from red (fully correlated) to blue (fully anti-correlated). (A) MA-CA (B) CA-p2 (C) p2-NC (D) NC-p1 (E) p1-p6 (F) RT-RH (G) RH-IN.

Figure 7. Unique intrinsic dynamic coupling revealed by cross-correlations of the substrate mean square fluctuations

The same cross-correlation analysis in Figure 6 was repeated for the substrate residues. The correlation coefficients are contoured and color-coded from red (fully correlated) to blue (fully anti-correlated). (A) MA-CA, (B) CA-p2, (C) p2-NC, (D) NC-p1, (E) p1-p6, (F) RT-RH, (G) RH-IN.

Figure 8. The static and dynamic substrate envelopes share the same overall trend in V_out, i.e., the volume lying outside the envelope

The absolute values, however, vary for static and dynamic cases, most likely due to the dynamic substrate envelope being more heterogeneous than the static envelope. The dynamic envelope is denser in the interior region, which consists of substrate backbones. Towards its surface, the dynamic envelope is less well-defined than its interior. This contrast is not as pronounced in the static substrate envelope. Hence, $V_{out}^{stat}$ is smaller than $V_{out}^{dyn}$ for all substrates.

Figure 9. Distributions of V_out, V_in, and V_tot values throughout the MD simulations are unimodal for each substrate

Mean of data was shown as a red line in each histogram.

Figure 10

Substrate size appears to determine how well the substrate fits within the substrate envelope, except for CA-p2, NC-p1, and p1-p6.

Figure 11. The overall dynamics of the substrate in the active site is dominated by side-chain fluctuations

Mean-square fluctuations of the center of mass for each substrate are plotted for the backbone, side-chain, and the entire substrate residue. Two substrates, NC-p1 and CA-p2, which protrude beyond the substrate envelope more than their total volume, appear to have highly dynamic centers of mass.

Figure 12. Intrinsic flexibility appears to play an important role in substrate fit within the substrate envelope for substrates MA-CA, CA-p2, NC-p1, and p1-p6

Correlation coefficients between center of mass fluctuations and V_out for each substrate are plotted using the backbone, side-chains, and entire substrate residues.

Figure 13. The dynamic substrate envelope gives a probabilistic consensus volume, which is easier and more accurate to incorporate into structure-based drug design protocols than the static envelope, which is essentially a step function

(A) Side and (B) top views of static substrate envelope, (C–D) respective views of dynamic substrate envelope. The grid cell centers are color-coded from high occupancy (red) to low occupancy (blue). Both envelopes are visualized as superposed onto the wild-type PR-CA-p2 structure. The dynamic substrate envelope is denser in the interior where the backbone is more rigid than the side-chains. The static substrate envelope can be considered homogeneous compared to the dynamic envelope. Individual substrate volumes for both the crystal structures and conformational ensembles are also shown in Figure S3.

Figure 14. Residue-based V_out values for (A) MA-CA, (B) CA-p2, (C) p2-NC, (D) NC-p1, (E) p1-p6, (F) RT-RH, and (G) RH-IN

The amino acids with both flexible and bulky side-chains, i.e., Arg and Lys, tend to protrude beyond the substrate envelope more than other amino acids with fewer degrees of freedom and/or smaller side-chains, i.e., Met, and/or Val, Ala. The P4 and P4’ positions usually have high V_out values, likely due to the less well-defined characteristic of the substrate envelope in the more solvent-exposed regions of these substrate end residues. Some substrates, however, have higher V_out values even in their more buried positions, i.e., P1’ in NC-p1, P1 in p1-p6. According to our hypothesis, these substrates co-evolve with drug-resistant PR because their greater protrusion beyond the substrate envelope makes them more sensitive to changes in the PR.