Genomic characterization of intrinsic and acquired resistance to cetuximab in colorectal cancer patients

Abstract

Anti-EGFR antibodies are effective in therapies for late-stage colorectal cancer (CRC); however, many tumours are unresponsive or develop resistance. We performed genomic analysis of intrinsic and acquired resistance to anti-EGFR therapy in prospectively collected tumour samples from 25 CRC patients receiving cetuximab (an EGFR inhibitor). Of 25 CRC patients, 13 displayed intrinsic resistance to cetuximab; 12 were intrinsically sensitive. We obtained six re-biopsy samples at acquired resistance from the intrinsically sensitive patients. NCOA4-RET and LMNA-NTRK1 fusions and NRG1 and GNAS amplifications were found in intrinsic-resistant patients. In cetuximab-sensitive patients, we found KRAS K117N and A146T mutations in addition to BRAF V600E, AKT1 E17K, PIK3CA E542K, and FGFR1 or ERBB2 amplifications. The comparison between baseline and acquired-resistant tumours revealed an extreme shift in variant allele frequency of somatic variants, suggesting that cetuximab exposure dramatically selected for rare resistant subclones that were initially undetectable. There was also an increase in epithelial-to-mesenchymal transition at acquired resistance, with a reduction in the immune infiltrate. Furthermore, characterization of an acquired-resistant, patient-derived cell line showed that PI3K/mTOR inhibition could rescue cetuximab resistance. Thus, we uncovered novel genomic alterations that elucidate the mechanisms of sensitivity and resistance to anti-EGFR therapy in metastatic CRC patients.

Figure 1

Clinical response to cetuximab. (a) Representative computed tomography scans from cetuximab intrinsic-sensitive and acquired-resistant colorectal cancer (CRC) patients. (b) Horizontal bar plots represent time (months) for which patients were on cetuximab treatment until progressive disease (PD) (black dots) for first-line cetuximab-based chemotherapy (n = 14) (top) or salvage cetuximab/irinotecan chemotherapy (n = 11) (bottom). Orange, blue, and grey bars indicate partial response (PR), stable disease (SD), and PD, respectively. Blue stars indicate successful re-biopsy in patients who achieved PR and then developed acquired resistance. Vertical waterfall bar plots in the right panels show the percent change in tumour size (y-axis) from baseline during cetuximab treatment. As defined by RECIST criteria, patients who achieved a shrinkage in tumour size of >30% were classified as PR and ‘intrinsic sensitive’ for our study.

Figure 2

Landscape of somatic variants and gene set expression signatures in sensitive, resistant, and acquired-resistant tumours. (a,b) Somatic variants identified in genes grouped by molecular pathways. MUT, non-synonymous single nucleotide or small indel mutation; FUS, gene fusion; AMP, copy number amplification (CN ≥4). The top bar chart shows the tumour mutation burden of non-synonymous somatic variants. Amino acid change is shown for a subset of genes in (b). (c) Hierarchical clustering of gene set expression signatures for a panel of published molecular signatures (see Methods). The Z score represents both the magnitude and relative direction of a signature’s expression. Baseline tumour (T), acquired-resistant tumour (RT).

Figure 3

Characterization of NCOA4–RET and LMNA–NTRK1 gene fusions. (a,e) Schematic depiction of the gene fusion detected by RNA-Seq, indicating the genomic position, resulting fusion mRNA, and predicted protein consequences (with key protein domains indicated). (b,f) RT-PCR of RNA from the tumour or adjacent normal tissue with primers upstream and downstream of the gene fusion breakpoint. NTC, no template control. Red arrows indicate the size of the expected PCR product. (c,g) Overexpression of the 3′ fusion gene partner in the tumour relative to the adjacent normal tissue. RPM, reads per million. (d,h) Sanger sequencing of the RT-PCR product from (b) and (f). The red arrowheads indicate the position of the fusion junction and in-frame amino acid sequence.

Figure 4

Variant and copy number changes in sensitive and acquired-resistant tumour pairs. (a) Variant allele frequency (VAF) of somatic mutations detected in baseline (T) versus acquired-resistant (RT) tumours. Variants were grouped as non-silent (blue) and silent (black) based on the predicted impact on protein sequence. (b) B-allele frequency (BAF) plots show regions of genomic imbalance from copy number variants for each chromosome (1–22) of T and RT.

Figure 5

Gene expression signatures in sensitive and acquired-resistant tumour pairs. Analysis of RNA expression signatures and enrichment scores in sensitive versus acquired-resistant tumour pairs. Baseline tumour (T), acquired-resistant tumour (RT). (a) Pathway enrichment Z-scores plotted for each molecular pathway expression signature. The green line represents a smooth local regression LOESS curve, and grey shading highlights the 95% confidence interval around the curve. Pathways with the highest enrichment Z-scores in each panel are labelled. Red arrows highlight pathways of interest. (b) Expression (log2) of individual genes that comprise the epithelial (E.EMT) and mesenchymal (M.EMT) gene expression signature. A diagonal line is plotted for comparison. The EMT signature genes showing >4-fold change difference of expression between baseline tumour and acquired-resistant tumour are labelled. (c) Gene expression signatures for stromal and immune cell tumour infiltration were quantified using the ESTIMATE tool (see Methods). Orange bars represent scores for immune cell infiltration and grey bars, scores for stromal cell infiltration.

Figure 6

Sensitivity of a patient-derived cell line from patient #4612 to PI3K/mTOR inhibition. (a) Computed tomography scans for patient #4612 at baseline and following 3 and 12 cycles of cetuximab treatment. A patient-derived cell (PDC) line was cultured from the acquired-resistant tumour (malignant ascites, which developed following 12 cycles). (b) Dose-dependent inhibition of PDC cell proliferation by 5-day treatments with cetuximab (EGFR), erlotinib (EGFR), GDC-0994 (Erk1/2), BEZ-235 (dual PI3K/mTOR), AZD-8055 (mTOR), and BKM-120 (pan-PI3K). Results from four independent experiments are shown. The IC₅₀ values of AZD-8055, BEZ-235, and BKM-120 are 0.02, 0.047, and 0.971 µM, respectively. (c) Western blot expression of basal and 24 h treatments of PDC. Cetuximab concentration, 75 µg/mL; other compounds, 1 µM. The experiments were repeated thrice.