A missense variant effect map for the human tumor-suppressor protein CHK2

Abstract

The tumor suppressor CHEK2 encodes the serine/threonine protein kinase CHK2 which, upon DNA damage, is important for pausing the cell cycle, initiating DNA repair, and inducing apoptosis. CHK2 phosphorylation of the tumor suppressor BRCA1 is also important for mitotic spindle assembly and chromosomal stability. Consistent with its cell-cycle checkpoint role, both germline and somatic variants in CHEK2 have been linked to breast and other cancers. Over 90% of clinical germline CHEK2 missense variants are classified as variants of uncertain significance, complicating diagnosis of CHK2-dependent cancer. We therefore sought to test the functional impact of all possible missense variants in CHK2. Using a scalable multiplexed assay based on the ability of human CHK2 to complement DNA sensitivity of Saccharomyces cerevisiae cells lacking the CHEK2 ortholog, RAD53, we generated a systematic "missense variant effect map" for CHEK2 missense variation. The map reflects known biochemical features of CHK2 while offering new biological insights. It also provides strong evidence toward pathogenicity for some clinical missense variants and supporting evidence toward benignity for others. Overall, this comprehensive missense variant effect map contributes to understanding of both known and yet-to-be-observed CHK2 variants.

Keywords: CHEK2; CHK2; DNA damage repair; Saccharomyces cerevisiae; breast cancer; deep mutational scanning; tumor suppressor; variant effect mapping; yeast.

Figure 1

Assay validation and generation of the CHK2 variant effect map (A) Overview of the TileSeq framework that was followed to produce the CHK2 variant effect map. (B) Yeast-based functional complementation assay for CHK2 domains labeled with tested variants. The sml1Δ rad53Δ yeast strain was transformed with the expression vector pADH1-Leu carrying wild-type (WT) CHEK2 or the empty pADH1-Leu expression vector (empty). The selective condition was 0.007% MMS, and the non-selective condition was 2% DMSO. Yeast growth was assessed by spotting serial dilutions of yeast cells on selective media and incubation for 3 days. P, LP, and VUS indicate pathogenic, likely pathogenic, and variant of uncertain significance, respectively. We note LP^∗ for p.Glu161del (p.E161del) because it has been annotated variously as LP, P, and VUS, and note VUS^∗ for p.His143Tyr (p.H143Y) because one of six annotations in the ClinVar database was LP, while the others were VUS. The CHK2 protein domain graphic is based on an image from Wang et al. and information from Boonen et al.

Figure 2

Experimental variant effect maps of CHK2 (A) Functional scores for synonymous (green), nonsense (blue), and missense (gray) variants from positions 2 to 489 in the original map were plotted as a histogram with scores on the x axis and density on the y axis. The dashed vertical blue and green lines overlaid on the missense distribution mark the median nonsense and median synonymous score, respectively. (B) A preview of the full-length original CHK2 variant effect map with magnified views of segments of the FHA domain (positions 116–186) and kinase domain (positions 346–426). An enlarged version of the map can be found in Figure S5. Each heatmap shows functional scores for every possible amino acid substitution and nonsense mutation (bottom row). Above each heatmap is a consensus summary of the distribution of functional scores at each position. As shown in the legend (right), a functional score of 0 (blue) corresponds to the median score of nonsense variants; a score of 1 (white) corresponds to the median of synonymous variants; scores above 1 (red) are considered hyper-complementing; gray indicates missing data; and the length of the diagonal line on each block indicates estimated each score’s estimated standard error.

Figure 3

CHK2 map scores agree with expected effects of conservation and solvent accessibility (A) Median CHK2 functional scores were calculated and stratified based on each position’s conservation score. Conservation scores were derived from ConSurf multiple sequence alignments based on an AlphaFold predicted PDB file for CHK2. Residues labeled “Not Conserved” have ConSurf scores above 0.91, while “Conserved” positions are below −0.47. The p value was calculated via a Wilcoxon rank-sum test comparing functional scores at “Not Conserved” positions to “Conserved” positions (B) Solvent accessibility was determined using FreeSASA software (v.2.02, October 22, 2017) on the AlphaFold (v.2022-11-01, Monomer v2.0 pipeline; UniProt: O96017) CHK2 structure. Residues with less than 20% accessible surface area (ASA) for their side chain were considered “inaccessible” while those with above 50% side-chain ASA are “accessible.” Results were separated into the FHA domain (positions 92–205) and kinase domain (positions 212–501). p values comparing functional scores at solvent accessible and inaccessible positions were calculated via a Wilcoxon rank-sum test for positions in the FHA domain or the kinase domain. (C) Crystal structure of the CHK2 dimer with colored residues based on the median functional score of each position. (D) Crystal structure of the CHK2 dimer with blue residues for positions with a dimer interface burial greater than 0%.

Figure 4

Regions with functional but not (predicted) protein stability impacts highlight positions in the FHA domain important for dimerization Stability scores (estimated ΔΔG folding energy values) were generated via FoldX software using the AlphaFold predicted protein structure of CHK2. At each position, ΔΔG was calculated for every substitution and the mean value was plotted against the average positional CHK2 functional scores in a ten-amino-acid-wide moving window analysis (with single-residue step size). Variants at positions with a ΔΔG near 0 are predicted to be typically stable, while ΔΔG values greater than 0 are predicted to be less stable. Variants with functional scores of 1 should be considered “wild-type-like,” while scores near 0 should be considered to have profound loss of function. The dashed horizontal black line indicates the damaging threshold functional score of 0.5 and the dashed horizontal blue line a stable ΔΔG score of 0. Spearman's R and p value were calculated comparing the ten-amino-acid-wide moving windows of functional scores and ΔΔG stability scores.

Figure 5

Known mutational hotspots in CHK2 are largely intolerant to missense variants Mutational hotspots identified in Hudson et al. were separated into distinct kinase activation motifs (APE, DFG, GxGxxG, HRD, and VAIK) and plotted against CHK2 functional scores for each variant at these positions. Boxplots depict the median functional score for each motif as well as the 25th and 75th percentiles with vertical lines extending to 1.5 times the interquartile range above and below 25th and 75th percentiles. All motifs were conserved, matching the consensus sequence observed in other kinases. Double asterisks indicate significance by Mann-Whitney U test of less than 1e−5 as compared to non-hotspot positions. The absolute delta medians for motifs APE, DFG, GxGxxG, HRD, and VAIK, as compared to non-hotspot positions, are 1.07, 1.08, 1.04, 0.96, and 0.81, respectively.

Figure 6

CHK2 variant effect map scores agree with mammalian and yeast functional assays Scatterplot relating CHK2 variant effect map scores (x axis, indicating map score threshold of 0.5 for reference) to: (A) functional complementation scores for 120 CHK2 missense variants in yeast in the presence of MMS (y axis, indicating map score threshold of 0.5 for reference), indicating previously assigned categories of non-functional, semi-functional, and functional; and (B) measurements of pKAP1 phosphorylation after ionizing radiation in mouse embryonic stem cells (y axis), indicating previously assigned categories of damaging, intermediate, and functional scores.

Figure 7

CHK2 functional scores correctly classify clinical variants with known annotations and provide evidence for VUS reclassification (A) Using CHK2 map scores after confidence-interval filtering and a known set of clinically annotated pathogenic or benign CHEK2 variants from Invitae, we evaluated balanced precision—defined at each score threshold by the fraction of variants that are pathogenic given a balanced (50% prior probability of pathogenicity) test set—vs. recall (fraction of pathogenic variants captured at this threshold). The horizontal dashed line indicates R90BP with the numerical AUBPRC and R90BP listed in the bottom-right legend. LB/B indicates likely benign and benign, while LP/P means likely pathogenic and pathogenic. (B) CHK2 functional score ranges were converted into ACMG evidence strengths using log-likelihood ratios (LLRs) and matched against all VUS in ClinVar (as of June 2023). Evidence toward pathogenicity or benignity are split into discrete categories where V. Str, Str., Mod., and Sup. mean very strong, strong, moderate, and supporting, respectively. Sections in blue represent evidence toward pathogenicity and green toward benignity, while gray are VUS where no new evidence is provided. The numbers below each label show how many variants fall into each category. (C) LLRs of pathogenicity were calculated by comparing the probability distributions of scores for known pathogenic and benign variants in the reference set. The log ratio between likelihood of observing a score in the positive pathogenic reference set (blue) compared to the negative benign reference set (green) was calibrated to ACMG evidence strengths. Probability distributions are overlaid on the gray histogram of CHK2 missense variant scores, with the top panel showing which score ranges correspond to each ACMG evidence strength. (D) Functional scores were matched with subjects with breast cancer from the BRIDGES and CARRIERS studies, and odds ratios were calculated for individuals without a CHK2 variant, individuals with variants found to be tolerated (score above 0.5) in our functional assay, and individuals with variants that were found to be damaging (score below 0.5). The asterisk indicates a significant difference (p < 0.05) according to Fisher’s exact test.