Structure and energy based quantitative missense variant effect analysis provides insights into drug resistance mechanisms of anaplastic lymphoma kinase mutations

Abstract

Anaplastic lymphoma kinase (ALK) is considered as a validated molecular target in multiple malignancies, such as non-small cell lung cancer (NSCLC). However, the effectiveness of molecularly targeted therapies using ALK inhibitors is almost universally limited by drug resistance. Drug resistance to molecularly targeted therapies has now become a major obstacle to effective cancer treatment and personalized medicine. It is of particular importance to provide an improved understanding on the mechanisms of resistance of ALK inhibitors, thus rational new therapeutic strategies can be developed to combat resistance. We used state-of-the-art computational approaches to systematically explore the mutational effects of ALK mutations on drug resistance properties. We found the activation of ALK was increased by substitution with destabilizing mutations, creating the capacity to confer drug resistance to inhibitors. In addition, results implied that evolutionary constraints might affect the drug resistance properties. Moreover, an extensive profile of drugs against ALK mutations was constructed to give better understanding of the mechanism of drug resistance based on structural transitions and energetic variation. Our work hopes to provide an up-to-date mechanistic framework for understanding the mechanisms of drug resistance induced by ALK mutations, thus tailor treatment decisions after the emergence of resistance in ALK-dependent diseases.

Figure 1

Effect of mutations on ALK kinase domain stability. All 19 possible mutations along with the synonymous mutations that the residue mutates to itself at each position in ALK kinase domain are colored on a vertical bar in terms of their stability relative to wide-type ALK. The values of ΔΔG_fold are binned into seven categories: highly stabilizing (ΔΔG_fold < −1.84 kcal·mol⁻¹) and highly destabilizing (ΔΔG_fold > 1.84 kcal·mol⁻¹); stabilizing (−1.84 kcal·mol⁻¹ < ΔΔG_fold < −0.92 kcal·mol⁻¹) and destabilizing (0.92 kcal·mol⁻¹ < ΔΔG_fold < 1.84 kcal·mol⁻¹); slightly stabilizing (−0.92 kcal·mol⁻¹ < ΔΔG_foldd < −0.46 kcal·mol⁻¹) and slightly destabilizing (0.92 kcal·mol⁻¹ < ΔΔG_fold < 1.84 kcal·mol⁻¹); and neutral (−0.46 kcal kcal·mol⁻¹ < ΔΔG_fold < 0.46 kcal·mol⁻¹).

Figure 2

Scatter plot of the correlation between solvent accessible surface area (SASA) and difference in stability (ΔΔG_fold) of ALK mutations.

Figure 3

Distribution of stability effects of all possible simulated mutations and those observed in clinic. The distribution of stability changes arising from mutations observed in tumor cell (dashed line) stands in contrast to that of all possible simulated mutations (solid line). The probability distributions shown here are obtained by kernel smoothing of the original data

Figure 4

The distribution of mutational effects (ΔE) of ALK mutations. (a) The distribution of mutation effect scores of all possible mutations and observed mutation calculated by Evmutation software. (c) Scatter plot of mutation effects of pathogenic mutations. The dash line indicates the median value of these scores (ΔE = −4.86). The mutation effects shown here are obtained by bubble and color mapped of the original data.

Figure 5

Inhibition constants of ALK Inhibitors with Wild-type and mutant ALK kinase. Predicted FR energies (ΔΔG_FR predicted) vs experimental FR energies (ΔΔG_FR experimental) for inhibitors with ALK and mutations. (a) MM-GBSA (0.601); (b) MM-PBSA (0.635); (c) AM1 (0.874); (d) RM1 (0.842); (d) PM6 (0.588).

Figure 6

Range of binding free energies of 10 ALK inhibitors and a substrate ATP against 21 reported drug resistance mutations occurring in ALK kinase domain.

Figure 7

Heat map of the mutation energy profile of 10 ALK inhibitors as well as a substrate ATP against 21 reported drug resistance mutations occurring in ALK kinase domain.

Figure 8

Structural Basis for drug resistance to crizotinib. (a) The location of ALK drug resistance mutations. AS refers to activation segment. The co-crystal structures of crizotinib bound to the wild type (panel b) and Arg1202 mutant (panel c).

Figure 9

RMSD with respect to the initial structure as a function of time for the various simulations over 50 ns.

Figure 10

The histogram indicates the variance proportion of each principal component from PC1 to PC5.

Figure 11

PCA scatter plot along first two principal components, eigenvector 1 (PC1) and eigenvector 2 (PC2) showing difference between the wild type and mutant systems.

Figure 12

Alignment of structures of wild-type ALK and the non-active sites mutations. (a) Comparison of structures of wild-type ALK and C1156Y mutation. C1156Y results in marked conformational changes in activation segment (AS), αC helix and loop 1122–1130. (b) Comparison of activation segment structures of wild-type ALK (white) and different mutations (light magenta).