Disentangling the mutational effects on protein stability and interaction of human MLH1

Abstract

Missense mutations can have diverse effects on proteins, depending on their location within the protein and the specific amino acid substitution. Mutations in the DNA mismatch repair gene MLH1 are associated with Lynch syndrome, yet the underlying mechanism of most disease-causing mutations remains elusive. To address this gap, we aim to disentangle the mutational effects on two essential properties for MLH1 function: protein stability and protein-protein interaction. We systematically examine the cellular abundance and interaction with PMS2 of 4839 (94%) MLH1 variants in the C-terminal domain. Our combined data shows that most MLH1 variants lose interaction with PMS2 due to reduced cellular abundance. However, substitutions to charged residues in the canonical interface lead to reduced interaction with PMS2. Unexpectedly, we also identify a distal region in the C-terminal domain of MLH1 where substitutions cause both decreased and increased binding with PMS2, and propose a region in PMS2 as the binding site. Our data correlate with clinical classifications of benign and pathogenic MLH1 variants and align with thermodynamic stability predictions and evolutionary conservation. This work provides mechanistic insights into variant consequences and may help interpret MLH1 variants.

Fig 1. Yeast-based abundance and interaction assays to measure variant effects in MLH1.

(A) The MutLα complex (MLH1:PMS2) participates in the DNA mismatch repair (MMR) pathway together with several other components to recognize and correct mismatches during DNA replication. MLH1 (purple) and PMS2 (orange) fold into an N-terminal domain (NTD) and a C-terminal domain (CTD), separated by a disordered linker. The two proteins interact through their C-terminal domains. The C-terminal MutLα complex was predicted using AlphaFold2-Multimer, whereas the N-terminal domain of MLH1 is PDB: 4P7A and the N-terminal domain of PMS2 is PDB: 1H7S. The disordered linkers between the domains were hand-drawn for illustrative purposes. (B) Schematic illustration of the dihydrofolate reductase protein-fragment complementation assay (DHFR-PCA) used to measure the cellular abundance of MLH1 variants. (C) Schematic illustration of the yeast two-hybrid (Y2H) assay used to measure the interaction of MLH1 variants with PMS2. Panels B and C of Fig 1 and panel B of Fig 2 were created using BioRender.com.

Fig 2. Screening of site-saturation mutagenesis libraries in the C-terminal domain of MLH1.

(A) Small-scale yeast growth assays to test the DHFR-PCA (left) and the Y2H assay (right). Yeast cells expressed either a vector, wild-type MLH1, or the indicated MLH1 variant. For the DHFR-PCA the vector was an empty vector, while for the Y2H assay the vector was the activation domain (AD) expressed without MLH1. For the DHFR-PCA, cells were grown on medium with or without MTX at 30 °C. For the Y2H assay, cells co-expressed PMS2 as bait fused to the DNA binding domain and were grown on medium with or without histidine at 30 °C. Q542L and L550P are in the C-terminal domain, S406N is in the disordered linker, and I219V and T117M are in the N-terminal domain. (B) Overview of the experimental design to screen libraries of MLH1 variants for abundance and interaction with PMS2. Parts of this figure were created using BioRender.com. (C) Histograms showing the score distributions of MLH1 variants in the DHFR-PCA (left) and the Y2H assay (right). A score of 0 corresponds to the wild-type in both assays, while a negative or positive score indicates reduced or increased abundance or interaction, respectively. As a reference, the abundance and interaction score of the wild-type MLH1 (black) and the L550P variant (red) are highlighted with dashed lines. The abundance scores were rescaled to range from −1 to 1 (see Materials and Methods).

Fig 3. MLH1 abundance and interaction heatmaps.

(A) Heatmaps showing the abundance scores of MLH1 variants. The scores range from low abundance (red), through wild-type-like abundance (white), to high abundance (blue). The wild-type residue at each position is marked in yellow, while missing variants are marked in gray. Substitutions that lead to a stop codon (nonsense mutations) are indicted by ‘Ter’. The position in the full-length MLH1 protein is shown below the heatmap. A linear representation of the secondary structure of MLH1 is displayed above the heatmap, colored by AlphaFold2 pLDDT as a measure of disorder. (B) The per-position median abundance scores are mapped to the C-terminal domain of MLH1 in the AlphaFold2-Multimer predicted C-terminal MutLα complex. The color scale corresponds to that shown in Fig 3A. The C-terminal domain of PMS2 is shown in orange. The N- and C-terminus of the C-terminal domain of MLH1 are annotated with the position in the full-length MLH1 protein. (C) Same as in (A), but for interaction scores. (D) Same as in (B), but for interaction scores.

Fig 4. Experimental effects correlate with computational predictions and clinical classifications.

(A) Scatter plots showing the correlation between the predicted GEMME ΔΔE score and predicted Rosetta ΔΔG score. A negative ΔΔE score suggests that the variant may not be compatible with MLH1 function as encoded in the sequence family. A positive ΔΔG score suggests that the variant is less stable than the wild-type MLH1. The MLH1 variants are colored by the abundance score and the color scale corresponds to that shown in Fig 3A. (B) Same as in (A), but for interaction scores. (C) Scatter plots showing the correlation between the abundance score and the gnomAD allele frequency (log scale). The MLH1 variants are colored based on their classification in ClinVar as either benign (dark green), likely benign (light green), likely pathogenic (yellow), pathogenic (red) or uncertain significance (gray). Variants observed in patients but not in gnomAD are indicated to the left of the broken x-axis. These variants were retrieved from Simple ClinVar [34] (accessed July 25, 2024). (D) Same as in (C), but for interaction scores.