Clonal replacement and heterogeneity in breast tumors treated with neoadjuvant HER2-targeted therapy

Genomic changes observed across treatment may result from either clonal evolution or geographically disparate sampling of heterogeneous tumors. Here we use computational modeling based on analysis of fifteen primary breast tumors and find that apparent clonal change between two tumor samples can frequently be explained by pre-treatment heterogeneity, such that at least two regions are necessary to detect treatment-induced clonal shifts. To assess for clonal replacement, we devise a summary statistic based on whole-exome sequencing of a pre-treatment biopsy and multi-region sampling of the post-treatment surgical specimen and apply this measure to five breast tumors treated with neoadjuvant HER2-targeted therapy. Two tumors underwent clonal replacement with treatment, and mathematical modeling indicates these two tumors had resistant subclones prior to treatment and rates of resistance-related genomic changes that were substantially larger than previous estimates. Our results provide a needed framework to incorporate primary tumor heterogeneity in investigating the evolution of resistance.

. Percentage of mutations identified as clonal in the first region that were later determined to be rare in additional tumor regions Supplementary Figure 1. Measures of intra-tumor heterogeneity (a) HFR (high-frequency regional). Clonal mutations C are defined as those with CCF > 0.5 in both samples; regional mutations R1 and R2 are those mutations present at CCF > 0.5 in one sample and CCF < 0.1 in the other sample. (b) tHFR (temporal high-frequency regional) is calculated across time, with one sample from one timepoint (e.g. pre-treatment) and one or more samples from another timepoint (e.g. posttreatment). Red diamonds represent mutations with CCF > 0.5 in all samples from the second timepoint and CCF < 0.1 in the sample from the first timepoint, while green diamonds represent mutations with CCF > 0.5 in all samples.

Supplementary Figure 2. Inference of primary breast tumor growth parameters (a) Heatmap
showing, for each patient tumor, the posterior probability distribution of the selection coefficient inferred via Approximate Bayesian Computation. Nineteen of twenty tumors, across breast cancer subtypes and both treated and untreated, were inferred to have grown under strong selection (s > 0.05). b. Heatmap showing, for each patient tumor, the distribution of deme sizes inferred using the simulated tumors. A large deme size (50,000 cells) was rarely inferred, but deme sizes 1,000-10,000 were plausible. Source data are provided as a Source Data file.

Supplementary Figure 3. Intra-tumor heterogeneity across simulated deme sizes
Simulated tumors that grew with larger deme sizes tended to exhibit lower heterogeneity as measured with HFR (high-frequency regional) than those that grew with smaller deme sizes. This effect was greatest at a deme size of 50,000, which was statistically unlikely to be consistent with the primary breast tumors based on Approximate Bayesian Computation. Source data are provided as a Source Data file. Simulations under different non-neutral selection coefficients were combined given similarity of results. Great variability was seen between tumors. Source data are provided as a Source Data file.

Supplementary Figure 5. Flowchart describing the selection of high-quality tumors for analysis
Note that of the 9 tumors excluded based on poor post-treatment cellularity (steps 2 and 3 in the flowchart), 2 would also have been excluded based on poor pre-treatment cellularity. Figure 6. Intra-tumor heterogeneity in untreated vs treated primary breast tumors HFR = high-frequency regional, BC = breast cancer, TN = triple-negative. Source data are provided as a Source Data file.

Supplementary Figure 7. Inferred mutation clusters across diagnostic and surgical samples
PyClone was used to define mutational clusters and assess changes in cluster frequencies across samples. P1-P5 were treated between the diagnostic timepoint (T1) and the surgical timepoint (T2), while P6 was not. In P1 and P5, a cluster of high frequency mutations was observed in all posttreatment regions that was absent or rare in the pre-treatment region. Multiple other mutation clusters were seen to be present in only one region at low or high frequency, or to be absent in only one region. This pattern of heterogeneity was seen in both the treated and untreated tumors. Source data are provided as a Source Data file. Figure 9. Information about precise sampling scheme can improve ability to assess for clonal evolution In P5, the two pairs of post-treatment samples came from opposite quadrants of the tumor. We used this sampling scheme to sample from virtual (simulated) tumors in order to generate a more specific distribution of tHFR values for this case. Here we plot the tHFR distributions for the general sampling scheme with three samples from the same octant and for the more precise sampling scheme outlined in Supplementary Figure 8B. The blue dotted lines correspond to the tHFR values for P5 computed with three (all possible subsamples of three regions) and then four post-treatment regions. These findings demonstrate that the degree of clonal change seen in P5 was statistically unlikely to have resulted from pre-existing heterogeneity (tHFR > 99 th percentile) under the four-sample precise sampling scheme. Source data are provided as a Source Data file. Figure 10. Increased post-treatment copy number similarity after clonal replacement The pairwise copy number distances inferred with MEDICC were higher in pairs of pretreatmentpost-treatment samples than in pairs of post-treatmentpost-treatment samples in P1 and P5 (which underwent clonal replacement), but were similar pre-post to post-post in P2, P3, and P4 (which did not). Source data are provided as a Source Data file. Figure 11. Somatic mutational signatures across mutation groupings We defined non-synonymous mutations as truncal, subclonal, or post-only (see Methods) and compared the COSMIC mutational signatures present in each category. Truncal mutations appeared to result from a different set of mutational processes than subclonal and post-treatment mutations, while subclonal and post-treatment mutations shared a similar set of mutational signatures. Source data are provided as a Source Data file.