Background: CDK4/6 inhibitors (CDK4/6i) in combination with antiestrogens have prolonged survival of patients with ER+ metastatic breast cancer. However, this combination is not curative mainly due to acquired drug resistance. Knowledge about mechanisms of such resistance remains quite incomplete. We report herein a forward-genetics screen to discover a broad spectrum of novel somatic mutations causal to CDK4/6i resistance.

Methods: We used CRISPR/Cas9 to delete the DNA mismatch repair (MMR) gene MSH2 in MCF7 and T47D ER+ breast cancer cells. Deficiency of DNA MMR proteins such as MSH2 results in a high nucleotide substitution rate which, in turn, predisposes cells to acquire drug resistance-associated mutations. MSH2-/- MCF7 and T47D cells were infected with a lentiviral barcode library containing ~1000 unique DNA barcodes. MSH2-/- barcoded cells were expanded for ~25 doublings to allow the accumulation of random mutations. Clones resistant to CDK4/6i were selected in the presence of IC90 of palbociclib (200 nM) or abemaciclib (500 nM) for 4-6 weeks. CDK4/6i resistant clones with unique barcode IDs were subjected to whole-exome sequencing (WES).

Results: Following drug selection, ~73 uniquely barcoded resistant colonies emerged from MCF7 and T47D MSH2-/- clonal lines. As expected, MCF7 and T47D MSH2-/- clones harbored a high mutation burden compared to parental cells. Candidate variants were distilled based on (a) functionality prediction and (b) mutation frequency in Project GENIE. We observed RB1 (5/73 clones; 6.8%) mutations in CDK4/6i resistant clones, providing proof-of-principle that this approach can identify clinically-relevant drug resistant alterations. Overall, we identified non-synonymous alterations in 2,206 genes in T47D palbociclib-resistant, 2,195 genes in T47D abemaciclib-resistant, and 1,312 genes in MCF7 palbociclibresistant lines. A secondary screen of the 10 genes recurrently mutated in all three CDK4/6i resistant groups identified loss of ASXL1 as top hit. ASXL1 encodes a polycomb repressive complex protein that regulates chromatin accessibility. Loss of ASXL1 has been implicated in myeloid transformation through epigenetic reprogramming. WES of 76 CDK4/6i resistant tumor biopsies (DFCI/MBCproject cohort) identified ASXL1 alterations in two and four patients with acquired and primary resistance, respectively (6/76=7.9%). One of the tumors that progressed after an initial response to palbociclib had acquired the same ASXL1 R549C mutation that was identified in our screen. Among 1,769 tumors from patients treated with CDK4/6i (TEMPUS database), 37 exhibited ASXL1 alterations (4 frameshift, 6 truncating, 3 in-frame del, 24 missense mutations). DNAseq of patient-derived organoids established from post-CDK4/6i metastases identified ASXL1 mutations in 2/7 organoids (29%). ASXL1-/- MCF7 and T47D cells were cross-resistant to fulvestrant. GSEA analysis of RNA-seq data showed upregulation of E2F targets in palbociclib-treated cells stably transduced with ASXL1 shRNA but not control shRNA (Enrichment score=0.75, q=1.00E-09). This was associated with maintenance of RB phosphorylation in the presence of CDK4/6i, markedly higher levels of CDK2, CDK6, cyclins E and A, and downregulation of p21 and p27. Finally, siRNAs targeting CDK2 or cyclin A reduced the viability of ASXL1-deficient T47D cells by 50% and 90%, respectively.

Conclusions: An accelerated mutagenesis approach using MMR-deficient ER+ breast cancer cells identified loss of ASXL1 as a novel mechanism of resistance to CDK4/6i. ASXL1 alterations were found in ~8% of tumors from patients with de novo or acquired resistance to CDK4/6i. Knockdown of CDK2 and cyclin A restored sensitivity to CDK4/6i and reduced viability of ASXL1 deficient cells, suggesting CDK2 inhibitors are a treatment approach against these drug-resistant tumors.

VIEW THE POSTER

VIEW THE PUBLICATION