Pan-cancer analysis of co-occurring mutations in RAD52 and the BRCA1-BRCA2-PALB2 axis in human cancers

Abstract

In human cells homologous recombination (HR) is critical for repair of DNA double strand breaks (DSBs) and rescue of stalled or collapsed replication forks. HR is facilitated by RAD51 which is loaded onto DNA by either BRCA2-BRCA1-PALB2 or RAD52. In human culture cells, double-knockdowns of RAD52 and genes in the BRCA1-BRCA2-PALB2 axis are lethal. Mutations in BRCA2, BRCA1 or PALB2 significantly impairs error free HR as RAD51 loading relies on RAD52 which is not as proficient as BRCA2-BRCA1-PALB2. RAD52 also facilitates Single Strand Annealing (SSA) that produces intra-chromosomal deletions. Some RAD52 mutations that affect the SSA function or decrease RAD52 association with DNA can suppress certain BRCA2 associated phenotypes in breast cancers. In this report we did a pan-cancer analysis using data reported on the Catalogue of Somatic Mutations in Cancers (COSMIC) to identify double mutants between RAD52 and BRCA1, BRCA2 or PALB2 that occur in cancer cells. We find that co-occurring mutations are likely in certain cancer tissues but not others. However, all mutations occur in a heterozygous state. Further, using computational and machine learning tools we identified only a handful of pathogenic or driver mutations predicted to significantly affect the function of the proteins. This supports previous findings that co-inactivation of RAD52 with any members of the BRCA2-BRCA1-PALB2 axis is lethal. Molecular modeling also revealed that pathogenic RAD52 mutations co-occurring with mutations in BRCA2-BRCA1-PALB2 axis are either expected to attenuate its SSA function or its interaction with DNA. This study extends previous breast cancer findings to other cancer types and shows that co-occurring mutations likely destabilize HR by similar mechanisms as in breast cancers.

Fig 1. Distribution of RAD52, BRCA2, BRCA1 and PALB2 mutations in cancer cells.

A. Distribution by type of mutation. The COSMIC mutations were partitioned into non-coding (5’ UTR, 3’UTR and intronic) and coding (translated). The coding mutations were further partitioned into missense, non-sense, frameshift, InDels and silent. All frameshift mutations introduce a stop codon and are therefore truncations. B. The distribution of all coding RAD52, BRCA2, BRCA1 and PALB2 among the various cancers reported on COSMIC. The large intestine, breast, prostate and skin are the most represented cancers.

Fig 2. Co-occurring mutations in RAD52 and the BRCA2-BRCA1-PALB2 axis.

A. Percent co-occurring mutations between the four genes using data from Table 1. Total percentage (100%) represents all mutations in Table 1. B. Distribution of co-occurring mutations among the different cancer types. A graphical representation of the data from Table 1. C. Pan-cancer statistical analysis if co-occurring mutations using cBioPortal (www.cbioportal.org). D. Statistical analysis of co-occurring mutations by tissue using cBioPortal.

Fig 3. Significant mutations in RAD52, BRCA2, BRCA1 and PALB2.

Cartoon diagrams of RAD52, BRCA2, BRCA1 and PALB2 showing specific domains in each gene. The diagrams were adapted from [, –64]. Co-occurring mutations from Table 1 are color coded. Highlighted mutations are those with a high probability of being driver or pathogenic.

Fig 4. Location of mutated residues on the structure of RAD52.

A. The structure of RAD52 with single-stranded DNA bound to the inner site (PDB ID: 5XRZ) was used to map the mutated residues, shown in magenta. The DNA is shown in orange with bases in dark blue and potassium ions are shown as purple spheres. RAD52 is shown in gray with a single monomer represented in tan. B. The tan single monomer is shown individually, and side chains of mutated residues shown as magenta sticks.

Fig 5. The effect of R55H, D149E, and G59R on the structure of RAD52.

Structures of point mutations (cyan) were generated using homology modeling and then aligned to known structures of RAD52 (tan). The side chain of the mutated residue is shown in sticks. A. R55H mutant is aligned to a monomer with DNA bound to the inner site (PDB ID: 5XRZ). B. Zoomed in view of the R55H mutation compared to wildtype. C. D149E mutant is aligned to a monomer with DNA bound to the inner site (PDB ID: 5XRZ). D. Zoomed in view of the D149E mutation compared to wildtype. E. G59R mutant is aligned to a monomer with DNA bound to the outer site (PDB ID: 5XS0). F. Zoomed in view of the G59R mutation compared to wildtype.