Exome sequencing identifies breast cancer susceptibility genes and defines the contribution of coding variants to breast cancer risk

Abstract

Linkage and candidate gene studies have identified several breast cancer susceptibility genes, but the overall contribution of coding variation to breast cancer is unclear. To evaluate the role of rare coding variants more comprehensively, we performed a meta-analysis across three large whole-exome sequencing datasets, containing 26,368 female cases and 217,673 female controls. Burden tests were performed for protein-truncating and rare missense variants in 15,616 and 18,601 genes, respectively. Associations between protein-truncating variants and breast cancer were identified for the following six genes at exome-wide significance (P < 2.5 × 10^-6): the five known susceptibility genes ATM, BRCA1, BRCA2, CHEK2 and PALB2, together with MAP3K1. Associations were also observed for LZTR1, ATR and BARD1 with P < 1 × 10^-4. Associations between predicted deleterious rare missense or protein-truncating variants and breast cancer were additionally identified for CDKN2A at exome-wide significance. The overall contribution of coding variants in genes beyond the previously known genes is estimated to be small.

Fig. 1. Manhattan plot of z scores from the meta-analysis assessing the association between protein-truncating variant carriers within genes and breast cancer risk.

The x axis is the chromosomal position, and the y axis is the meta-analyzed z score from testing H₀: β = ln(OR) = 0 in the UK Biobank and BCAC datasets (two-tailed). The blue lines correspond to z = ±3.29, P = 0.001, the red lines correspond to z = ±4.71, P = 2.5 × 10⁻⁶. All labeled genes are those with P < 0.001. All P values are unadjusted for multiple testing.

Fig. 2. Quantile–quantile plot of P values from the meta-analysis assessing the association between protein-truncating variant carriers and breast cancer risk.

P values are from the meta-analyzed z score from testing H₀: β = ln(OR) = 0 in the UKB and BCAC datasets (two-tailed). The x axis is the expected log₁₀ P values from the null hypothesis, the y axis is the observed log₁₀ P values. All highlighted genes have P < 0.0005 and are associated with an increased risk of breast cancer. All P values are unadjusted for multiple testing.

Extended Data Fig. 1. Manhattan plot of Z-scores from the meta-analysis assessing the association between rare missense variant carriers by gene and breast cancer risk.

The x-axis gives chromosomal position, and the y values are the meta-analyzed Z-scores from testing H₀: β = ln (OR)=0 β =ln =0 in the UKB and BCAC datasets (2-tailed). The blue lines correspond to Z=±3.29 ± 3.29, P = 0.001, the red lines correspond to Z=±4.71 ± 4.71, P = 2.5 × 10⁻⁶. The labeled genes are those with P < 0.001. All P-values are unadjusted for multiple testing.

Extended Data Fig. 2. Quantile-quantile plot of P-values from the meta-analysis assessing the association between rare missense variant carriers by gene and breast cancer risk.

P-values are from the meta-analyzed Z-score from testing H₀: β=ln(OR)=0β=ln =0 in the UKB and BCAC datasets (2-tailed). The x-axis gives the expected log₁₀ P-values under the null hypothesis and the y-axis the observed log₁₀ P-values Highlighted genes are those with P < 0.0005. Blue corresponds to an increased risk of breast cancer, and cream corresponds to a decreased risk of breast cancer. All P-values are unadjusted for multiple testing.

Extended Data Fig. 3. Manhattan plot of Z-scores from the meta-analysis assessing the association between PTV or deleterious (CADD > 20) rare missense variant carriers by gene and breast cancer risk.

The x-axis gives the chromosomal position, and the y values are meta-analyzed Z-scores from testing H₀: β = ln (OR) = 0 β = ln =0 in the UKB and BCAC datasets (2-tailed). The blue lines correspond to Z=±3.29 ± 3.29, P = 0.001, the red lines correspond to Z= ± 4.71 ± 4.71, P = 2.5 × 10⁻⁶. All labeled genes are those with P < 0.001. All P-values are unadjusted for multiple testing.

Extended Data Fig. 4. Quantile-quantile plot of P-values from the meta-analysis assessing the association between PTV or deleterious (CADD > 20) rare missense variant carriers by gene and breast cancer risk.

P-values are from the meta-analyzed Z-score from testing H₀: β = ln(OR)=0 β = ln =0 in the UKB and BCAC datasets (2-tailed). The x-axis gives the expected log₁₀ P-values under the null hypothesis and the y-axis the observed log₁₀ P-values Highlighted genes are those with P < 0.0005. Blue corresponds to an increased risk of breast cancer, and cream corresponds to a decreased risk of breast cancer. All P-values are unadjusted for multiple testing.