Multiple cancers escape from multiple MAPK pathway inhibitors and use DNA replication stress signaling to tolerate aberrant cell cycles

Abstract

Many cancers harbor pro-proliferative mutations of the mitogen-activated protein kinase (MAPK) pathway. In BRAF-driven melanoma cells treated with BRAF inhibitors, subpopulations of cells escape drug-induced quiescence through a nongenetic manner of adaptation and resume slow proliferation. Here, we found that this phenomenon is common to many cancer types driven by EGFR, KRAS, or BRAF mutations in response to multiple, clinically approved MAPK pathway inhibitors. In 2D cultures and 3D spheroid models of various cancer cell lines, a subset of cells escaped drug-induced quiescence within 4 days to resume proliferation. These "escapee" cells exhibited DNA replication deficits, accumulated DNA lesions, and mounted a stress response that depended on the ataxia telangiectasia and RAD3-related (ATR) kinase. We further identified that components of the Fanconi anemia (FA) DNA repair pathway are recruited to sites of mitotic DNA synthesis (MiDAS) in escapee cells, enabling successful completion of cell division. Analysis of patient tumor samples and clinical data correlated disease progression with an increase in DNA replication stress response factors. Our findings suggest that many MAPK pathway-mutant cancers rapidly escape drug action and that suppressing early stress tolerance pathways may achieve more durable clinical responses to MAPK pathway inhibitors.

Fig. 1 |. EGFR-, KRAS-, and BRAF-mutant cancer cells can survive and cycle through high-dose MAPK pathway inhibitors.

(A) MAPK signaling pathway and subsequent cell-cycle progression. Displayed markers of cell-cycle commitment (phospho-Rb) and DNA synthesis (EdU incorporation) are used throughout this study. Common MAPK pathway mutations and inhibitors used are shown. (B) Representative immunofluorescence images of five human MAPK-mutant cancer cell lines 96 hours after indicated treatments, stained for DNA content and phospho-Rb. Images are representative, and value displayed is the percent of surviving cells that are positive for phospho-Rb staining; mean ± std, representative of 2 experiments. Scale bar = 100 μm. (C) Dose-response profiling of MAPK-mutant cancer cell lines to their corresponding inhibitors after 96-hour treatments. The phospho-Rb⁺ cell count is normalized as a percent relative to that of the 96-hour DMSO condition. Shaded bars cover dose ranges commonly used in cell culture models in the literature. Green shaded bar is shared by EGFR inhibitors; blue shaded bars are shared by BRAF inhibitors; orange bars are shared by MEK inhibitors. Reported % value in plot legends corresponds to the measured value remaining at the highest concentration of shaded bar regions. Mean ± std of 3 replicate wells imaged, representative of 2 experiments. (D) Apoptosis measurements of five cancer cell lines in response to effective inhibitor doses, typically > IC₈₀ from panel (C), reported as % of cells positive for propidium iodide and annexin by flow cytometry (see fig. S2 for gating scheme). Mean ± std of 3 replicate cell suspensions, representative of 2 experiments.

Fig. 2 |. Escape from MAPK pathway inhibition is concurrent with MAPK pathway reactivation.

(A) Western blot analysis of phospho-ERK after 96 h of the indicated inhibitors. Histone H3 is used as a loading control. Blots are from one representative experiment (see (29) for additional independent experiments in melanoma cells). (B) Immunofluorescence images of A375 cells stained for phospho-ERK after Encorafenib treatment. White arrowhead points to a cell with high phospho-ERK intensity after 96 hours in drug. Images are representative of 2 experiments. Scalebar = 40 μm. (C) Single-cell violin plots of phospho-ERK intensity over time after MAPK inhibitor treatment. Osim, 100 nM; AMG510, 100 nM; Enc, 1 μM. Red box is drawn at the mean intensity value for untreated cells, and the value shown represents the % of the population above this threshold. Red line connects the mean intensity at each time point. n = 1,000 cells plotted per condition, representative of 2 experiments. (D) Mean mRNA plots for 8 genes downstream of ERK in the indicated cell lines. MAPK inhibitor treatment for each cell line is the same as in (C). Data are representative of 2 experiments. (E) Representative images of cells co-stained for CCND1 mRNA and phospho-Rb. White arrowhead points to an escapee cell with high CCND1 mRNA. Each row corresponds to each cell line, as in (D). Data are representative of 2 experiments. Scale bar = 20 μm. (F) Split violin quantification of CCND1 mRNA puncta in quiescent (p-Rb⁻) and cycling (p-Rb⁺) cells for each indicated treatment condition. Symbol (+) indicates the distribution mode. n > 500 cells per condition, representative of 2 experiments. ***P<0.001, determined by permutation test of non-cycling cells (p-Rb⁻) compared to cycling cells (p-Rb⁺).

Fig. 3 |. Escapee cell cycles display slowed DNA synthesis and increased time spent in G2-M.

(A) Density scatter of DNA content and EdU incorporation in PC9 cells after indicated 96-hour treatments. Red gate marks standard intensity of EdU⁺ control (n > 2000 cells per condition, representative of 2 experiments). (B) Box plots of DNA synthesis rate in all cancer cell lines 96 hours after indicated treatments. Box displays median and quartiles, whiskers display 1^st-99^th percentile; n > 100 cells per condition, representative of 2 experiments; ns = not significant, **P< 0.01, and ***P< 0.001 determined by permutation test of MAPKi-treated cells compared to vehicle-treated cells. (C) PC9, SW837, and A375 single-cell traces of DNA helicase B (DHB)-based CDK2 activity sensor (43) (see fig. S5A for sensor localization dynamics) over 2 to 3 days in treatment. Time spent in active cell-cycle progression is highlighted by shaded bar based on time between CDK2 rise and anaphase. Data are representative of 2 experiments. (D) Time spent in active cell-cycle progression (time from CDK2-rise to anaphase) for untreated proliferating cells and escapees. Box displays median and quartiles, representative of 2 experiments. **P< 0.01 and ****P< 0.0001 determined by unpaired t-test. (E) Percent of cycling cells in G2-M after 96-hour treatments (see fig. S5, C to E for gating scheme). Mean ± std of 3 replicate wells imaged, representative of 2 experiments; ns = not significant, *P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001 determined by unpaired t-test of MAPKi-treated cells compared to vehicle-treated cells. (F) Film strip of H2B-mIFP in A375 cells completing mitosis under DMSO treatment or BRAF inhibition. Images are representative of 2 experiments. Scale bar = 40 μm. (G) Quantification of time spent in mitosis (from nuclear envelope breakdown to the point of anaphase) for indicated cell lines and MAPK inhibition treatments. n = 30 cells per condition, representative of 2 experiments. ****P<0.0001 determined by Mann-Whitney U test. (H) Representative images of A375 cells immunostained for phospho-histone H3 (mitotic marker) and stained for EdU incorporation (DNA synthesis) treated for 4 days with encorafenib. Scale bar = 5 μm. (I) Quantification of Edu foci in mitotic cells treated as indicated. Data are mean ± SEM of 10 mitotic cells per condition, representative of 2 experiments; p-values specified, determined by unpaired t-test.

Fig. 4 |. Cells escaping MAPK inhibition show widespread indications of DNA damage accumulation and DNA replication stress responses.

(A) Representative images of CDKN1A mRNA FISH (green) in A375 cells treated with encorafenib for 0, 3, 24 or 96 hours. Images are representative of 2 experiments. Scale bar = 20 μm. (B) Violin plot quantification of CDKN1A mRNA puncta per cell over time following BRAF inhibition; n > 500 cells per condition, representative of 2 experiments. (C) Representative immunofluorescence images of five human MAPK-mutant cancer cell lines after 72 hours of the indicated treatments, stained for DNA content (Hoechst, blue), EdU incorporation (pink), and γ-H2AX (green). Images are representative of 2 experiments. Scale bar = 10 μm. (D) Quantification of γ-H2AX nuclear fluorescence intensity after 72 hours of the indicated treatments in EdU^– and EdU⁺ cells, displayed as split violins; n > 100 cells per condition, representative of 2 experiments. **P<0.01, and ***P<0.001 determined by permutation test of MAPK-inhibited EdU⁺ cells compared with DMSO-treated EdU⁺ cells. (E) Top, representative immunofluorescence images of phospho-CHK1 (Ser³¹⁷) in A375 cells escaping BRAF inhibition (cycling status displayed by EdU incorporation). Scalebar = 40 μm. Bottom, histograms of phospho-CHK1 (Ser³¹⁷) nuclear intensity in S-phase cells (EdU⁺) in the indicated cell lines after 96 hours of MAPK inhibition. (F) Distribution of EdU intensity in escapees (after 100 nM osimertinib or 1 μM encorafenib) co-treated with ATR inhibition (100 nM AZ20 or 100 nM AZD6738) for 96 hours. Line drawn at the mean value; n = 190 cells plotted per condition, representative of 2 experiments; *P<0.05 and **P<0.01 determined by permutation test. (G) Quantification of cycling cells (determined by phospho-Rb Ser^807/811 immunofluorescence) after 96 hours of the indicated treatment combinations with ATR inhibitors at 100 nM and 1 μM. Data are mean ± std of 3 replicate wells imaged; ns = not significant, *P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001 determined by unpaired t-test.

Fig. 5 |. Escapee cancer cells broadly rely on the recruitment of FANCD2 in late S/G2 to successfully complete mitosis in drug.

(A) Representative immunofluorescence images of cells after 72 hours of 100 nM AMG510 or 1 μM encorafenib (Enc), stained for DNA content, EdU incorporation, and FANCD2. Images are representative of 2 experiments. Scale bar = 10 μm for H358 cells; 20 μm for A375 cells. (B) Scatter plots of DNA content and EdU intensity in A375 cells treated with DMSO or 1 μM dabrafenib for 72 hours. Each data point is color-coded to represent nuclear FANCD2 intensity (a.u., arbitrary units); n ≥ 2200 cells per condition, representative of 2 experiments. (C) Quantification of FANCD2 intensity after 72-hour treatments in EdU⁺ cells, displayed as violin plots. Trametinib (Tra): 10 nM; binimetinib (Bini), gefitinib (Gefit), osimertinib (Osim), ARS1620, and AMG510: 100 nM; dabrafenib (Dab) and encorafenib (Enc): 1 μM. n > 100 cells per condition, representative of 2 experiments; ns = not significant, **P<0.01, and ***P<0.001 determined by permutation test of MAPK-inhibited cells compared with DMSO-treated cells. (D) Representative single-cell traces of CDK2 activity (cytosolic-to-nuclear ratio of DHB-mCherry) and FANCD2-mCitrine nuclear foci in A375 cells. Data are representative of 2 experiments. (E) Quantification of maximum FANCD2 foci number or duration for untreated cells (−24 to 0 hours) or cell cycles spanning early or later periods in 1 μM dabrafenib. n > 12 cells per condition, representative of 2 experiments. *P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001 determined by Mann-Whitney U test. (F) Representative images of escapees immunostained for phospho-histone H3 (mitotic marker), EdU, and FANCD2. Same drug doses as in (C). Scale bar = 3 μm. Data are representative of 2 experiments. (G) Line intensity profile (across orange line) for Hoechst, EdU, and FANCD2 in the cell indicated in (F). (H) Box plots of EdU-FANCD2 colocalization events for untreated mitotic cells and mitotic escapees. Symbol (+) indicates the distribution mean. n = 10 mitotic cells per condition, representative of 2 experiments. *P<0.05 determined by unpaired t-test. (I) Heatmaps of single-cell CDK2 activity traces of A375 cells after transfection with control or FANCD2 siRNA as indicated and 1 μM dabrafenib. Each row represents the CDK2 activity in a single cell over time according to the colormap. Traces cropped due to cell death are marked by black coloring. Data are representative of 2 experiments. (J) Fate summary of cells observed over the imaging period from the experiment in panel (I). (K) Flow cytometric apoptosis assay after the indicated 7-day treatments. Top are representative contour plots of PC9 cells progressing toward apoptosis (quadrant Q2) after EGFR inhibition with 100 nM osimertinib and FANCD2 depletion. Bottom is quantification of Q2 across the indicated cell lines and treatments. Data are mean ± std of 3 replicate cell suspensions, representative of 2 experiments; ns = not significant, *P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001 determined by unpaired t-test.

Fig. 6 |. Matrigel-embedded spheroids and clinical data highlight replication stress tolerance features of escapees.

(A) Spheroid growth of the indicated cells after the indicated treatments. Across all spheroid panels: KRASi, 100 nM AMG510; BRAFi, 1 μM Encorafenib; ATRi, 5 μM AZD6738; FANCi, 250 nM PIP199. Data are mean ± SEM of 10 spheroids per condition, representative of 2 experiments; p-values specified, determined by unpaired t-test of from the final time point. (B) Representative maximum intensity projections of 3D spheroid clusters stained for phospho-Rb and cleaved PARP (c-PARP, apoptotic marker) after 7 days of the indicated treatments. Scale bar = 20 μm. Images are representative of 2 experiments. (C) Quantification of cell count at day 7 in each individual spheroid. Each stacked bar represents one spheroid, where the cell count is broken down by p-Rb status. Combination treatment groups often resulted in spheroids with zero p-Rb-positive cells. n = 5 spheroids per condition, representative of 2 experiments. (D) Percent of c-PARP-positive cells per spheroid in (C). P-values specified, determined by unpaired t-test. (E) Formalin-fixed paraffin-embedded (FFPE) tissue sections longitudinally biopsied before and after treatment from two patients with BRAF^V600E melanoma (see fig. S12A for patient and tumor details). Immunofluorescence of Ki67 (proliferation marker) and γ-H2AX (DNA damage marker) are displayed. Scale bar = 25 μm for MB2282, 50 μm for MB3022. (F) Quantification of γ-H2AX intensity (blue coloring) in all cells and in Ki67-positive cells (red coloring) from data in (E). Line drawn at the mean; n > 235 cells per ungated condition, n > 45 cells per Ki67-gated condition; p-value as indicated or **P<0.01 and ***P<0.001, determined by permutation test. (G) Kaplan-Meier survival analyses based on data generated from the Harmonized Cancer Datasets within the Genomic Data Commons (GDC) research database (as of April 28, 2022; https://portal.gdc.cancer.gov). Regardless of treatment type, all patients with various tumor types harboring ≥ 1 EGFR, KRAS, or BRAF mutation and ≥ 1 TP53 mutation were included; N = 33 to 829 patients in each group. Cohort comparisons were made between tumor subgroups containing wild-type or mutant forms of the indicated genes. Left-most plot examines a cohort that has ≥ 1 mutation among a list of FA-associated genes (fig. S13A). Displayed P-values were determined by log-rank test.

Fig. 7 |. Proposed model of early drug escape, stress tolerance, and cancer cell evolution in extended treatment.

Cells treated with drug have divergent fates within the first few days of treatment, where they will undergo cell death, remain quiescent, or escape treatment and re-enter the cell cycle. These escapee cells will suffer DNA replication stress, relying on the ATR and FA pathways to restore replication fork stability and complete mitosis.