APOBEC3 mutagenesis drives therapy resistance in breast cancer

Abstract

Acquired genetic alterations drive resistance to endocrine and targeted therapies in metastatic breast cancer; however, the underlying processes engendering these alterations are largely uncharacterized. To identify the underlying mutational processes, we utilized a clinically annotated cohort of 3,880 patient samples with tumor-normal sequencing. Mutational signatures associated with apolipoprotein B mRNA-editing enzyme catalytic polypeptide-like 3 (APOBEC3) enzymes were prevalent and enriched in post-treatment hormone receptor-positive cancers. These signatures correlated with shorter progression-free survival on antiestrogen plus CDK4/6 inhibitor therapy in hormone receptor-positive metastatic breast cancer. Whole-genome sequencing of breast cancer models and paired primary-metastatic samples demonstrated that active APOBEC3 mutagenesis promoted therapy resistance through characteristic alterations such as RB1 loss. Evidence of APOBEC3 activity in pretreatment samples illustrated its pervasive role in breast cancer evolution. These studies reveal APOBEC3 mutagenesis to be a frequent mediator of therapy resistance in breast cancer and highlight its potential as a biomarker and target for overcoming resistance.

Fig. 1. APOBEC3 mutational signatures are prevalent in breast cancers.

a, Schematic of analysis pipeline of MSK-IMPACT breast cancer cohort. b, Summary of genomic characteristics of the clinical cohort demonstrating percentage contribution of APOBEC3 mutational signature (first panel), TMB (second panel), SNV change (third panel) and OncoPrint of select genes in samples. c, Barplots displaying the proportion of samples with indicated dominant mutational signature categorized by sample type and receptor status. Groups were compared using the two-tailed Pearsonʼs chi-squared test. d, Violin plots representing TMB in samples categorized by receptor status. Groups were compared with APOBEC3-dominant samples using the two-tailed Wilcoxon test. e, Proportion of TMB-high samples with different dominant mutational signatures categorized by receptor status. f, Proportion of samples categorized by dominant mutational signature and histology. Groups were compared with APOBEC3-dominant samples using the two-tailed Pearson’s chi-squared test. IDC, invasive ductal carcinoma; ILC, invasive lobular carcinoma; MIDLC, mixed invasive ductolobular breast cancer. Panel a created using BioRender.com. Source data

Fig. 2. APOBEC3 enzymes induce APOBEC3 mutational processes.

a, Hematoxylin and eosin staining and immunohistochemical images displaying A3A and A3A/B/G staining in an APOBEC3-dominant patient sample. Scale bars, 50 µm (middle panel) and 20 µm (right panel). In total, n = 130 tissue samples were stained. b, Schematic of experimental design to investigate ongoing mutational processes in cells overexpressing WT A3A, A3B or their catalytic mutant controls. c, Mutational signature contribution of acquired SNVs in the indicated samples. d, Barplots representing number of clusters with acquired regions of kataegis in samples from c. e, Substitution profile of dA3A^WT-5 and dA3A^WT-8 in the kataegis regions. f, Circos plot representing acquired SNVs, indels, CNAs and structural rearrangements in dA3A^WT-5 cells. g, Mutational signature contribution and number of SNVs from WGS of five paired primary/metastatic patient samples. h, Barplots representing number of clusters with acquired regions of kataegis in the indicated metastatic patient samples. i, Circos plots of samples MSK-BR-WGS-05-P and MSK-BR-WGS-05-M. d, daughter; p, parent.

Fig. 3. APOBEC3 mutagenesis promotes therapeutic resistance in breast cancers.

a,b, Kaplan–Meier curves displaying PFS probability of patients with HR⁺/HER2⁻ MBCs treated with ET as single agent (PFS 8.6 versus 15.6 months in APOBEC3-dominant tumors and tumors with Other dominant signatures, respectively; hazard ratio, 1.4; 95% CI, 0.9–2.2; P = 0.12) (a) or in combination with CDK4/6 inhibition (hazard ratio, 1.5; 95% CI, 1.2–1.8; P = 2.4 × 10⁻⁴ for APOBEC3-dominant versus Others and hazard ratio, 1.8; 95% CI, 1.4–2.2; P = 6.4 × 10⁻⁷ for HRD-dominant versus Others) (b). Patients were categorized according to the dominant mutational signatures of the biopsy obtained before start of treatment. Groups were compared using log rank test. c, Growth curves of T47D A3A^WT and A3A^E72Q cells treated with DMSO or fulvestrant (10 nM). Data represent mean ± s.d. of three replicates. The groups were compared using two-way ANOVA test. d, Growth curves of MCF7 A3B^WT and A3B^E255Q cells treated with DMSO or fulvestrant (10 nM). Data are represented as individual replicates (n = 3). Groups were compared using two-tailed Mann–Whitney U test. e, Growth curves of T47D A3A^WT and A3A^E72Q cells treated with DMSO or abemaciclib (500 nM). Data represent mean ± s.d. of three replicates. Groups were compared using two-way ANOVA test. f, Schematic showing the timeline of generation of abemaciclib-resistant T47D A3A^WT (A3A^WT-R) cells (left panel). Mutational signature contribution of acquired SNVs in the samples indicated (right panel). g, Growth curves of BT-474 WT and A3A KO cells treated with DMSO or lapatinib (20 nM). Data represent mean ± s.d. of three replicates. Groups were compared using two-way ANOVA test. h, Mutational signature contribution of acquired SNVs in the samples indicated. i, Growth curves of BT-474 WT and A3A KO cells treated with DMSO or MK2206 (100 nM). Data represent mean ± s.d. of three replicates. Groups were compared using two-way ANOVA test. j, Crystal violet staining of MDA-MB-453 WT and A3A KO cells treated with DMSO (for 6 days) or T-DXd (100 ng ml⁻¹, for 73 days). Images are representative of n = 3 replicates. Scale bar, 200 µm. a.u., arbitrary units.

Fig. 4. APOBEC3-class alterations drive therapeutic resistance in breast cancers.

a, Volcano plot displaying enrichment of genes in metastatic HR⁺/HER2⁻ samples categorized according to the dominant mutational signature. Odds ratios (log₂ transformed) were computed by logistic regression (two-sided) with P values (−log₁₀ transformed) corrected by permutation test. Genes are represented as circles, color coded and sized according to the legend. b, OncoPrint of acquired alterations in HR⁺/HER2⁻ tumor samples categorized according to the dominant mutational signature. c, Barplots representing proportion of shared or acquired SNVs categorized as APOBEC3-context and non-APOBEC3-context SNVs. Groups were compared using the two-sided Fisher’s exact test. d, Volcano plots depicting site-specific enrichment of ESR1 (left panel) and PIK3CA (right panel) categorized according to the dominant mutational signature. Odds ratios (log₂ transformed) were computed by logistic regression with P values (−log₁₀ transformed) corrected for FDR by Benjamini–Hochberg method. Genes are represented as circles, color coded and sized according to the legend. e, Site-specific enrichment of alterations in RB1 gene in HR⁺/HER2⁻ tumor samples. f, Mutation spectrum of acquired SNVs in MSK-BR-WGS-05-M. g, Pathogenic mutations with APOBEC3-context substitutions in samples MSK-BR-WGS-05-P and MSK-BR-WGS-05-M. Asterisk indicates nonsense mutations. h, Likely oncogenic CNAs in samples MSK-BR-WGS-05-P and MSK-BR-WGS-05-M. i, Rainfall plot displaying intermutational distance between SNVs in chromosome 19 of sample MSK-BR-WGS-05-P. j, FACETS plot showing LOH of chromosome 13 in dA3A^WT-5 and acquired SNVs in RB1 after DMSO (A3A^sensitive) or abemaciclib (1 µM) (A3A^resistant) treatment. Asterisk indicates nonsense mutations. k, Immunoblots displaying changes in cell-cycle regulatory proteins in A3A^sensitive and A3A^resistant cells treated with DMSO or indicated doses of abemaciclib for 24 h. Vinculin was used as a loading control. Immunoblots are representative of n = 3 independent experiments. l, Inhibition of proliferation of A3A^sensitive and A3A^resistant cells treated with increasing concentrations of abemaciclib. Data represent mean ± s.d. of three replicates and are normalized to the DMSO control. Source data

Fig. 5. Vulnerabilities of APOBEC3-dominant breast cancers.

a, Barplots displaying the proportion of APOBEC3-dominant and non-APOBEC3-dominant tumors categorized based on the disease control rate (DCR) on 1–2 lines or >2 lines of anti-PD-1 immunotherapy. b, Kaplan–Meier curves displaying PFS probability of patients with HR⁺/HER2⁻ MBCs treated with PI3Kα inhibitor. Patients were categorized according to the dominant mutational signatures of the biopsy obtained before start of treatment. Groups were compared using log rank test. c, Barplots representing proportion of PIK3CA mutations in APOBEC3-dominant and non-APOBEC3-dominant tumors categorized as TMB-high or TMB-low. Groups were compared using the two-sided Fisher’s exact test. d, Sankey plots representing the evolution of early/late paired HR⁺/HER2⁻ patient samples categorized according to the dominant mutational signatures. e, Boxplots displaying the mutational signature contribution in paired earlier samples when the later sample is APOBEC3-dominant (n = 439). Groups were compared using the two-tailed Wilcoxon test. Boxplots display the first quartile (Q1), median and third quartile (Q3), with whiskers extending to the nearest datapoints within Q1 − 1.5 × IQR and Q3 + 1.5 × IQR.

Extended Data Fig. 1. Evaluation of SigMA as a tool to assess dominant APOBEC3 mutational signature.

(a) Schematic of the mutational signature analysis pipeline from publicly available datasets. (b) Barplots displaying the sensitivity, specificity, and accuracy of SigMA to call APOBEC3 as a dominant signature in the indicated datasets. (c) Concordance between the highlighted signature calling tools and SigMA in computing mutational signatures from the TCGA and Bertucci et al. datasets. (d) Barplots representing the proportion of APOBEC3-dominant samples in the indicated datasets categorized as <5 single nucleotide variants (SNVs) or ≥5 SNVs.

Extended Data Fig. 2. Characterization of APOBEC3 mutational signatures in breast cancers.

(a) Box plots showing the contribution of APOBEC3, HRD, Clock and Other mutational signatures in primary and metastasis samples categorized according to the receptor status. The primary and metastasis groups were compared using two-tailed Wilcoxon test. (b) Box plots displaying the fraction of genome altered (FGA) in primary and metastasis samples categorized according to the dominant mutational signature and receptor status. The primary and metastasis groups were compared using two-tailed Wilcoxon test. (c) Violin plots displaying the FGA in primary (top panel) and metastatic (bottom panel) patient samples categorized according to the dominant mutational signature and receptor status. The groups were compared using two-tailed Wilcoxon test. (d) Barplots representing the proportion of primary (top panel) and metastatic (bottom panel) patient samples with whole genome doubling (WGD) categorized according to the receptor status. The groups were compared using two-tailed Wilcoxon test.

Extended Data Fig. 3. A3A and A3B induce APOBEC3 mutagenesis.

(a) Scatter plot displaying the correlation between APOBEC3 signature contribution and immunohistochemistry (IHC) staining (in percentage of positive cells) of anti-UMN-13 (left panel) and anti-5210-87-13 (right panel) antibodies in HR+/HER2- breast cancer samples categorized according to the dominant mutational signatures. (b) Immunoblots of T47D cells stably transduced with doxycycline-inducible HA-tagged A3A^WT, A3A^E72Q, A3B^WT or A3B^E255Q (top panel). Vinculin was used as a loading control. Single-stranded DNA (ssDNA) deaminase activity of T47D A3A^WT, A3A^E72Q, A3B^WT or A3B^E255Q cells (lower panel). The cells were treated with or without 0.1 µg/mL doxycycline for 48 h before harvesting. (c) Western blot (top panel) and ssDNA deaminase activity (lower panel) of the indicated T47D daughter clones of A3A^WT, A3A^E72Q, A3B^WT or A3B^E255Q cells. Vinculin was used as a loading control. The cells were treated with 0.1 µg/mL doxycycline for the indicated number of days. (d) Pie charts representing the contribution of mutational signatures in sample dA3A^WT-5 calculated from the WGS using Signal or the simulated MSK-IMPACT panel using SigMA. (e) Tumor mutational burden (TMB) in the indicated samples (median TMB 5.25 for A3A^WT vs 2.75 for A3A^E72Q, 4.67 for A3B^WT vs 3.36 for A3B^E255Q). (f) Rainfall plots displaying non-clustered (grey) and clustered (colored) mutations in samples dA3A^WT-5 and dA3A^E72Q-5. (g) Schematic of treatment timeline of MSK-BR-WGS-03. Source data

Extended Data Fig. 4. APOBEC3 activity drives therapeutic resistance in breast cancers.

(a-g) Kaplan-Meier curves displaying progression-free survival probability of patients with (a) HR+/HER2- metastatic breast cancers treated with everolimus and endocrine therapy, (b) HR+/HER2- metastatic breast cancers treated with chemotherapy in 1^st-2^nd line, (c) HR+/HER2- metastatic breast cancers treated with chemotherapy in 3^rd-4^th line, (d) HR+/HER2- metastatic breast cancers treated with chemotherapy in >4^th line, (e) TNBC metastatic breast cancers treated with chemotherapy in 1^st-2^nd line, (f) TNBC metastatic breast cancers treated with chemotherapy in 3^rd-4^th line or (g) TNBC metastatic breast cancers treated with chemotherapy in >4^th line. The patients are categorized according to the dominant mutational signatures of the biopsy obtained prior to treatment start. The groups were compared using two-sided log-rank test. (h) Growth curves of T47D A3A^WT and A3A^E72Q cells treated with DMSO or palbociclib (500 nM). Data are represented as mean ± SD of three replicates. The groups were compared using two-way ANOVA test. AU is abstract unit. (i, j) Growth curves of MDA-MB-453 WT and A3A KO cells treated with DMSO or (i) lapatinib (2 µM) or (j) neratinib (100 nM). Data are represented as mean ± SD of three replicates. The groups were compared using two-way ANOVA test. (k) Crystal violet staining of MDA-MB-453 WT and A3A KO cells treated with DMSO (for 6 days) or neratinib (100 nM, for 24 days). Images are representative of n = 3 replicates. Scale bar = 200 µm.

Extended Data Fig. 5. Gene enrichment in breast cancer samples.

Volcano plots depicting enrichment of genes in (a) primary HR+/HER2-, (b) primary TNBC or (c) metastatic TNBC samples categorized according to the dominant mutational signature.

Extended Data Fig. 6. Site-specific enrichment in HR+/HER2- breast cancer samples.

Volcano plots depicting site-specific enrichment of the indicated genes categorized according to the dominant mutational signature.

Extended Data Fig. 7. A3A-driven abemaciclib resistance in MDA-MB-453 cells.

(a) Circos plot displaying acquired changes in MDA-MB-453 WT cells upon abemaciclib treatment. (b) Inhibition of proliferation of MDA-MB-453 WT sensitive and resistant cells treated with increasing concentrations of abemaciclib. Data are represented as mean ± SD of three replicates and are normalized to the DMSO control. (c) Immunoblots displaying changes in cell cycle regulatory proteins in MDA-MB-453 WT sensitive and resistant cells treated with DMSO or abemaciclib (100 nM) for 24 h. Vinculin was used as a loading control. (d) Bar plot displaying the fold change in CDK6 mRNA expression in MDA-MB-453 WT sensitive and resistant cells. Data are represented as mean ± SD of three replicates. Source data

Extended Data Fig. 8. Outcome of HR+/HER2- breast cancers to anti-PD-1 immunotherapy.

Kaplan-Meier curves displaying progression-free survival probability of patients with HR+/HER2- metastatic breast cancers treated with anti-PD-1 immunotherapy. The patients are categorized according to the dominant mutational signatures of the biopsy obtained prior to treatment start and the line of treatment. The groups were compared using the two-sided log-rank test.