Discorhabdin N, a South African Natural Compound, for Hsp72 and Hsc70 Allosteric Modulation: Combined Study of Molecular Modeling and Dynamic Residue Network Analysis

Abstract

The human heat shock proteins (Hsps), predominantly Hsp72 and Hsp90, have been strongly implicated in various critical stages of oncogenesis and progression of human cancers. While drug development has extensively focused on Hsp90 as a potential anticancer target, much less effort has been put against Hsp72. This work investigated the therapeutic potential of Hsp72 and its constitutive isoform, Hsc70, via in silico-based screening against the South African Natural Compounds Database (SANCDB). A comparative modeling approach was used to obtain nearly full-length 3D structures of the closed conformation of Hsp72 and Hsc70 proteins. Molecular docking of SANCDB compounds identified one potential allosteric modulator, Discorhabdin N, binding to the allosteric β substrate binding domain (SBDβ) back pocket, with good binding affinities in both cases. This allosteric region was identified in one of our previous studies. Subsequent all-atom molecular dynamics simulations and free energy calculations exhibited promising protein⁻ligand association characteristics, indicative of strong binding qualities. Further, we utilised dynamic residue network analysis (DRN) to highlight protein regions actively involved in cross-domain communication. Most residues identified agreed with known allosteric signal regulators from literature, and were further investigated for the purpose of deducing meaningful insights into the allosteric modulation properties of Discorhabdin N.

Keywords: South African natural compounds; allosteric drugs; heat shock proteins; human cancers; molecular dynamics.

Figure 1

(A) Cartoon representation of Hsp72 and Hsc70 3D structures calculated via homology modelling. The nearly full-length closed confirmation structures of Hsp72 and Hsc70 were modelled from Escherichia coli DnaK PDB entry: 2KHO. Predicted binding mode of SANC00132 docked to the allosteric target site, the SBDβ back pocket, is shown in sticks. Polar contacts are displayed as green dashes while interacting residues are shown in ball and stick. (B) Canonical nucleotide dependent cycle of Hsp70. (C) Schematic illustration of the experimental approach. Set1 comprises the simulation of SANC00132-free and SANC00132-bound systems only. Set2 simulations are for SANC00132-free and SANC00132-bound systems complexed with endogenous modulators (ADP and peptide substrate). Concisely, each protein was subjected to both sets of experiments.

Figure 2

(A) RMSD evolution of back-bone atoms. (B) Average RMSF of residues computed from C_α atoms. (i) Hsp72 Set1, (ii) Hsp72 Set2, (iii) Hsc70 Set1, (iv) Hsc70 Set2. Colour key: Black and red: SANC00132-free proteins. Blue and green: SANC00132-bound proteins.

Figure 3

(A,B) 2D projections of free energy landscape along the first and second principal components. Comparisons between SANC00132-free and SANC00132-bound systems. Surfaces are coloured based on Gibbs free energy levels from maroon (high energy/maxima) to blue (low energy/minima). Each contour represents an increase of the free energy of 1 kJ·mol⁻¹. Conformations sampled are labelled from C1–C20 (Hsp72) and C1–C11 (Hsc70), respectively. Snapshots of representative conformations occupying important free energy wells (minima) are shown as surfaces and coloured domain-wise: Blue: NBD, Green: Linker, Red: SBD. A: Hsp72: (i) Hsp72 apo Run1, (ii) Hsp72-SANC00132 Run1, (iii) Hsp72 endo-apo Run1, (iv) Hsp72 endo-complex Run2. B: Hsc70: (i) Hsc70 apo Run1, (ii) Hsc70-SANC00132 Run1, (iii) Hsc70 endo-apo Run1, (iv) Hsc70 endo-complex Run2. Duplicate trajectory results are shown in the Supplementary Data (Figure S5).

Figure 4

Bar plots showing per residue contribution to total binding free energy. (A) Hsp72 SANC00132-bound complexes. (B) Hsc70 SANC00132-bound complexes. All residues contributing more than +/−5 kJ·mol⁻¹ are listed in Table S7.

Figure 5

Contact network analysis: (A) Average shortest path (L) results. Colour key: Black and red: SANC00132-free, blue and green: SANC00132-bound. The upper threshold values of 9.31 (Hsp72 apo run1), 8.17 (Hsp72 apo run2), 8.20 (Hsp72 endo-apo run1), 7.65 (Hsp72 endo-apo run2), 7.09 (Hsc70 apo run1), 7.35 (Hsc70 apo run2), 7.40 (Hsc70 endo-apo run1), and 7.69 (Hsc70 endo-apo run2) are indicated by the dotted lines. Colour code: Cyan: run1, yellow: run2. (B) Betweenness centrality (BC) results. Colour key: Black: SANC00132-free, red: SANC00132-bound. The lower threshold values of 0.07 (Hsp72 apo run1), 0.04 (Hsp72 endo-apo run1), 0.03 (Hsc70 apo run1), and 0.03 (Hsc70 apo run1) are indicated by the grey dotted lines. Results of duplicate trajectories are shown in Figure S7.

Figure 6

Structural mapping of average L and average BC on SANC00132-free and SANC00132-bound Hsp72 and Hsc70 representative structures. Structures were retrieved from the largest clusters occupying important free-energy basins conformers (A: C1 (73 ns), C11 (82 ns), and B: C2 (56 ns), C7 (96 ns) in relation to Figure 3). Structures were coloured based on normalised average L values from red (lowest value; highly accessible) to blue (highest value; limited accessibility). Residues bearing the highest BC indices are indicated as spheres. First, both Hsp72 and Hsc70 images depict the global effects of ligand binding on structure compaction as described by Rg and interdomain distance results. Binding of SANC00132 to Hsp72 promotes intraprotein reorganisation from a loosely packed (more elongated) ligand-free 3D structure to a more compact complex. The opposite holds true for Hsc70. Second, it is evident that residue communities with significant average L and average BC values are populated at the interdomain linker, nucleotide, and substrate binding sites. SANC00132 is represented in sticks and coloured green.

Figure 7

Effects of ligand binding on residue accessibility and centrality. Graphical mapping of structural regions making significant contribution to total binding free energy (shaded blue) onto plots of average L difference (green) and average BC difference (red). Change in residue accessibility (average L difference) and residue centrality (average BC difference) were obtained from calculations of SANC00132-free less SANC00132-bound values. For each metric, per residue means across both sets (set1 and set2) of experiments was initially calculated. (A) Hsp72, (B) Hsc70.

Figure 8

Graphical illustration of correlation between 1/RMSF, 1/Average L, and Average BC. (A) Hsp72 (B) Hsc70. Pairwise comparisons between 1/BC and 1/L as well as comparisons from duplicate trajectories are found in Supplementary Data (Figure S16 and Figure S17). A: (i) Hsp72 apo run1, (ii) Hsp72-SANC00132 complex run1, (iii) Hsp72 endo-apo run1, (iv) Hsp72 endo-complex run2. B: (i) Hsc70 apo run1, (ii) Hsc70-SANC00132 complex run1, (iii) Hsc70 endo-apo run1, (iv) Hsc70 endo-complex run1.

Figure 9

Structural mapping of residues showing significant changes in centrality index on snapshots of Hsp72-SANC00132 run1 and Hsc70-SANC00132 run1. Regions were coloured from green (negative changes or increased residue centrality) to red (positive changes or decreased residue centrality).

Figure 10

A workflow illustration of methodology. The figure schematically describes the main steps performed and tools employed as indicated in brackets in this work.