CDK4/6 inhibitors induce replication stress to cause long-term cell cycle withdrawal

Abstract

CDK4/6 inhibitors arrest the cell cycle in G1-phase. They are approved to treat breast cancer and are also undergoing clinical trials against a range of other tumour types. To facilitate these efforts, it is important to understand why a cytostatic arrest in G1 causes long-lasting effects on tumour growth. Here, we demonstrate that a prolonged G1 arrest following CDK4/6 inhibition downregulates replisome components and impairs origin licencing. Upon release from that arrest, many cells fail to complete DNA replication and exit the cell cycle in a p53-dependent manner. If cells fail to withdraw from the cell cycle following DNA replication problems, they enter mitosis and missegregate chromosomes causing excessive DNA damage, which further limits their proliferative potential. These effects are observed in a range of tumour types, including breast cancer, implying that genotoxic stress is a common outcome of CDK4/6 inhibition. This unanticipated ability of CDK4/6 inhibitors to induce DNA damage now provides a rationale to better predict responsive tumour types and effective combination therapies, as demonstrated by the fact that CDK4/6 inhibition induces sensitivity to chemotherapeutics that also cause replication stress.

Keywords: CDK6; Palbociclib; cyclin-dependent kinase; replication stress; senescence.

Figure 1. A prolonged G1 arrest following CDK4/6 inhibition in RPE1 cells causes defects in the next cell cycle

1. Top panel displays structure of each CDK4/6 inhibitor tested. Bottom panel shows dose–response curves with these inhibitors displaying percentage of G1‐arrested RPE1‐FUCCI cells. To obtain dose–response curves, the number of mKO‐Cdt1‐positive (G1‐arrested) cells was calculated following 24 h drug addition (dark blue solid lines) or 24 h after subsequent drug washout (light blue dotted lines). C _max values observed in patients (taken from (He et al, ; Klein et al, 2018)) are represented on each graph with red dotted lines. Graphs display mean data ± SEM from three experiments, with at least 500 cells counted per condition per experiment.

2. Percentage of G1‐arrested RPE1‐FUCCI cells, calculated as in panel (A), but using a fixed concentration of CDK4/6 inhibitor for different durations of time, as indicated. Note, this is a fixed assay that quantifies the percentage of G1‐arrested cells prior to, or 24 h following, CDK4/6 inhibitor washout. Each bar displays mean data + SEM from three experiments, with at least 500 cells counted per condition per experiment.

3. Cell cycle profile of individual RPE1‐FUCCI cells (each bar represents one cell) after washout from 1 (top panel), 4 (middle panel) or 7 (bottom panel) days of treatment with CDK4/6 inhibitor, at the indicated doses (same concentration used in panel (B)). STLC (10 μM) was added to prevent progression past the first mitosis. Fifty cells were analysed at random for each repeat and three experimental repeats are displayed (150 cells total). Underneath the single‐cell profiles are quantifications of the observed cell cycle defects and G1 durations. Note, G1 length is estimated by mKO‐Cdt1 expression, and G1 lengths of less than 6 h could not be quantified since movies were only started after the 6 h drug washout period. These are indicated on the graphs as ≤ 6 h. Bar graphs show mean + SD. In the violin plots, horizontal lines display the median, and error bars show 95% confidence intervals.

4. Representative images and quantification of colony forming assays of RPE1 cells treated with CDK4/6 inhibitor for 1, 4 or 7 days and then grown at low density without inhibitor for 10 days. Bar displays mean data + SEM from three experiments.

Figure EV1. Effect of drug washout protocol or inhibiting CDK4/6 for 48 h on the reversibility of the G1 arrest in RPE1 cells

1. Schematic showing different washout protocols tested to ensure washout of high‐dose (10 μM) palbociclib. RPE1‐FUCCI cells were treated for 1 h with 10 μM palbociclib and subsequently washed out 1–6 times, with 1 h equilibration periods interspersed between washes. STLC (10 μM) was then added to arrest cells in mitosis before quantifying the amount of mKO‐Cdt1‐positive, G1‐arrested cells 24 h later.

2. Quantification of the G1‐arrested cells following the washout protocol described in (A). Graphs display the mean data ± SD from at least 500 cells counted per condition per experiments for two experimental repeats. Note, the points represent five different positions that were imaged per condition per experiment and the different coloured dots represent the two separate experiments.

3. Percentage of G1‐arrested RPE1‐FUCCI cells after treatment for 48 h with different CDK4/6 inhibitors (dark green solid lines) or 24 h after subsequent drug washout (light green dotted lines). The data are overlaid with 24 h arrest data from Fig 1A (blue lines) to allow comparison. Vertical red dotted lines indicate C _max values observed in patients (taken from (He et al, ; Klein et al, 2018)). Graphs display mean data ± SEM from three experiments, with at least 500 cells counted per condition per experiment.

Figure 2. p53 loss restores cell cycle progression and enhances long‐term proliferation following prolonged CDK4/6 inhibition in RPE1 cells

1. Dose–response curves displaying the percentage of G1‐arrested p53‐WT (blue) or KO (green) RPE1‐FUCCI cells following 24 h incubation with palbociclib (dark solid lines) or 24 h after subsequent washout (light dotted lines). Graphs display mean data ± SEM from three experiments, with at least 500 cells counted per condition per experiment.

2. Cell cycle profile of individual p53‐WT or KO RPE1‐FUCCI cells (each bar represents one cell) after washout from 1, 4 or 7 days of treatment with palbociclib (1.25 μM). STLC (10 μM) was added to prevent progression past the first mitosis. Fifty cells were analysed at random for each repeat and three experimental repeats are displayed (150 cells total). The right‐side panels show quantifications of cell cycle defects and G1 lengths from the displayed single‐cell profile plots. Note, G1 length is estimated by mKO‐Cdt1 expression, and G1 lengths of less than 6 h could not be quantified since movies were only stated after the 6 h drug washout period. These are indicated on the graphs as < 6 h. Bar graphs show mean + SD. In the violin plots, horizontal lines display the median, and error bars show 95% confidence intervals.

3. Representative images and quantifications of colony forming assays in p53‐WT or KO RPE1 cells treated with palbociclib (1.25 μM) for 1, 4 or 7 days and then grown at low density without inhibitor for 10 days. Each bar displays mean data + SEM from three experiments.

4. Cell cycle profile of individual p53‐WT or KO RPE1‐FUCCI cells to analyse multiple rounds of division following washout from 1, 4 or 7 days of treatment with palbociclib (1.25 μM). FUCCI profiles show 50 cells analysed at random from one experiment, which is representative of three experimental repeats. The right‐side panels show quantifications of cell cycle defects and G1 lengths from the three experimental repeats. Note, G1 length is estimated by mKO‐Cdt1 expression, and G1 lengths of less than 6 h could not be quantified since movies were only stated after the 6 h drug washout period. These are indicated on the graphs as < 6 h. Bar graphs show mean + SD. In the violin plots, horizontal lines display the median, and error bars show 95% confidence intervals.

5. Quantification of cell cycle profiles from cells treated as in (D). Graph shows the mean percentage of cells + SD that divide at each round of division, with 150 cells analysed in total from three experimental repeats.

Figure 3. Prolonged CDK4/6 inhibition in RPE1 cells induces replication stress and p53‐dependent cell cycle withdrawal

1. A

   Representative immunofluorescence images of p21 levels in p53‐WT or KO RPE1 cells, 48 h after release from 1, 4 or 7 days palbociclib (1.25 μM) treatment. Zoom inserts are 3× magnification of the indicated regions. Scale bars = 250 μM.

2. B

   Quantification of p21 intensities in cells treated as in panel (A). At least 100 cells were analysed per experiment and graph shows data from three experimental repeats. Violin plots display the variation in intensities between individual cells. Horizontal lines display the median, and error bars show 95% confidence intervals.

3. C

   Immunofluorescence images of DAPI and γH2AX staining in p53‐WT or KO RPE1 cells either before or 48 h after release from a 7‐day treatment with palbociclib (1.25 μM). Scale bar = 250 μM, zoom inserts = 3× magnification of highlighted regions.

4. D, E

   Quantification of nuclear morphologies (D) and γH2AX‐positive DNA damage foci (E) following palbociclib (1.25 μM) treatment in p53‐WT and KO RPE1 cells. Cells were treated for 1, 4 or 7 days and then analysed before or after drug washout for 48 h. A total of 100 cells (nuclear morphology) or 50 cells (γH2AX foci) were scored per condition per experiment, and bar graphs represent mean data + SEM from six experiments.

5. F

   Immunofluorescence images showing symmetrical 53BP1 staining following mitotic exit in p53‐KO cells after release from 7 days of palbociclib arrest. Three separate examples are displayed. Scale bar = 25 μM.

6. G–I

   Analysis of chromosome segregation errors and DNA damage during the first mitosis in GFP‐53BP1/H2B‐RFP RPE1 cells after release of from a 1‐ or 7‐day palbociclib (1.25 μM) arrest. Quantified from the same movies are nuclear morphology after mitosis (G), chromosome segregation defects during mitosis (H) and appearance of 53BP1 foci after mitosis (I). A total of 54 cells (1 day) or 80 cells (7 days) were analysed in total from two experiments. Errors bars display SD.

7. J, K

   Representative immunofluorescence images (J) and quantifications (K) of mitotic DNA replication assays (MiDAS) in p53‐KO RPE1 cells released from 7 days of palbociclib (1.25 μM) treatment or following 0.4uM aphidicolin treatment for 40 h. EdU foci were quantified in nocodazole‐arrested cells. Scale bar = 5 μM, zoom inserts = 3× magnification of highlighted areas. Ten cells were analysed per experiment and the bar chart shows the mean + SEM from three experimental repeats.

Figure EV2. Effect of ATR inhibition on the response of RPE1 cells to CDK4/6 inhibition

1. A

   Single‐cell cycle profiles quantified immediately following washout from 7 days 1.25 µM Palbociclib treatment. After washout, cells were cultured in the absence or presence of the ATR inhibitor, VE‐821 (5 µM). STLC was also added at drug washout to allow analysis of just the first cell cycle. Each bar represents an individual cell, and graphs show the data from three experimental repeats (150 cells analysed in total).

2. B, C

   Quantifications of the cell cycle defects (B) and G1 or S/G2 durations (C) from the single‐cell profiles shown in (A). Note, G1 length is estimated by the duration of mKO‐Cdt1 expression, and S/G2 by the time from AG‐Geminin expression until mitotic entry. Only cells that re‐enter the cell cycle were included in this quantification (G1‐arrested cells excluded). Bars graphs in (B) display mean + SD from three experimental repeats, violin plots in (C) display the variation in G1 and S/G2 length between individual cells, with horizontal lines indicating the median, and error bars representing 95% confidence intervals.

3. D

   Quantification of the nuclear morphologies following palbociclib (1.25 μM) treatment in p53‐WT and KO RPE1 cells. Cells were treated for 0, 1 or 7 days and before washout for 48 h (± ATR inhibition with 5 µM VE‐821). Nuclear morphologies of 100 cells were counted per condition and per experiment, and bar graphs represent mean data + SEM from three experiments.

Figure 4. Prolonged G1 arrest following palbociclib treatment in RPE1 cells downregulates replisome components and impairs origin licencing

1. Volcano plot of proteins up or downregulated following prolonged palbociclib (1.25 μM) treatment in RPE1 cells. The top 10 significantly upregulated and downregulated proteins are shown in blue and red respectively.

2. The top up‐ or downregulated Gene Ontology (GO) terms following 7‐day palbociclib (1.25 μM) treatment relative to 2 days of treatment.

3. Quantification of relative change in protein levels of selected replisome components between 2‐day (blue bars) and 7‐day (orange bars) palbociclib (1.25 μM) treatment. Graphs display mean + SD from 2–3 experimental repeats.

4. Representative western blots of whole‐cell lysates from RPE1‐WT cells treated with palbociclib (1.25 μM) for 1, 4 or 7 days, or treated identically, and then washed out for the indicated times to reflect when the majority of cells are in S‐phase (see Fig 1C).

5. Analysis of adjusted relative density from three independent western blot experiments. Bars display mean values ± SD. Significance determined by unpaired Student's t‐test comparing treated target protein to asynchronous target control (*< 0.01, ** < 0.001, *** < 0.0001).

6. Quantification of loaded MCM in p53‐WT and KO RPE1 cells treated with palbociclib (1.25uM) for 1 or 7 days followed by drug washout for 8 h after 1 day of arrest or 24 h after 7 days of arrest to capture cells in early S‐phase. To quantify loaded MCM, soluble MCM was pre‐extracted from cells and the amount of the remaining DNA‐loaded MCM was analysed by flow cytometry (see Appendix Fig S4 for representative FACS profiles). DNA content was measured with DAPI, and DNA synthesis was measured using a 30‐min EdU pulse. The amount of DNA‐loaded MCM in early S‐phase cells was compared to untreated control cells. The measured fluorescent intensity of each sample was divided by the background intensity of an identically treated but unstained control. The resulting ratios were normalized to WT control cells. Graphs display mean data ± SD from three experimental repeats. Significance determined by one‐way ANOVA followed by Tukey's multiple comparisons test (**P = 0.001, ****P < 0.0001).

Figure EV3. Western blots to characterize the level of replisome proteins in p53‐KO RPE1 cells treated with palbociclib

1. Representative western blots of whole‐cell lysates from p53‐KO RPE1 cells treated with palbociclib (1.25 μM) for 1, 4 or 7 days, or treated identically, and then washed out for the indicated times to reflect when the majority of cells are in S‐phase (see Fig 1C).

2. Analysis of adjusted relative density from three independent western blot experiments. Bars display mean values ± SD. Significance determined by unpaired Student's t‐test comparing treated target protein to asynchronous target control (*< 0.01, **< 0.001, ***< 0.0001).

Figure 5. CDK4/6 inhibition sensitizes RPE1 cells to genotoxic chemotherapeutics

1. A

   Cell cycle profile of individual RPE1‐FUCCI cells (each bar represents one cell) after release from 4d treatment with palbociclib (1.25 μM) or DMSO.

2. B–E

   Cell cycle profile of individual RPE1‐FUCCI cells treated as in panel (A), but additionally treated after drug washout with aphidicolin (B), olaparib (C), doxorubicin (D) or camptothecin (E), at indicated concentrations.

3. F, G

   Representative images (F) and quantifications (G) of colony forming assays with RPE1 cells treated with palbociclib (1.25 μM) for indicated times, and then grown at low density without palbociclib for 10 days. DMSO (control) or indicated genotoxic drugs were applied for the first 24 h after palbociclib washout. Each bar displays mean data + SD from four experiments.

Data information: All experiments in panels A‐F were run at the same time to allow comparison to the same control (panel (A)). STLC (10 μM) was added in all movies to prevent progression past the first mitosis. In all FUCCI graphs (A–E), 50 cells were analysed at random for each repeat and three experimental repeats are displayed (150 cells total). Bar graphs in (A–E) show mean + SD.

Figure EV4. EdU staining to quantify cell cycle arrest in different tumour lines treated with palbociclib or nutlin

1. A–H

   Quantification of the percentage of cells undergoing S‐phase during the final 24 h of a 7‐day arrest with 1 µM palbociclib in the tumour lines indicated (A–C and E–H), or the final 24 h of a 3‐day arrest with 5 µM Nutlin in the breast cancer lines indicated (D). Cells were treated with DMSO (untreated) or palbociclib (1 μM) for 7 days, or Nutlin for 3 days, with EdU (10 μM) pulsed in for the last 24 h of treatment. Data show mean + SD from two experiments, with at least 100 cells quantified per experiment.

Figure 6. Prolonged G1 arrest following palbociclib treatment in breast cancer cells downregulates replisome components, causes DNA damage and reduces long‐term proliferation

1. A

   Representative western blots of whole‐cell lysates from MCF7 cells treated with palbociclib (1 μM) for 1, 4 or 7 days, or treated identically, and then washed out for the indicated times to allow cells to enter S‐phase.

2. B

   Analysis of adjusted relative density from three independent western blot experiments. Bars display mean values ± SD. Significance determined by unpaired Student's t‐test comparing treated target protein to asynchronous target control (*< 0.01, **< 0.001, ***< 0.0001).

3. C, D

   Quantification of the nuclear morphologies (C) or γH2AX‐positive DNA damage foci (D) from MCF7, MCF7 p53 KO or T47D cells that were treated with palbociclib (1 μM) for 0, 1, 4 or 7 days and then analysed 72 h after drug washout. Either 100 cells (nuclear morphology) or 50 cells (γH2AX) were scored per condition per experiment and bar graphs represent mean data + SEM from three experiments.

4. E

   Cumulative mitotic entry of cells following washout from 0, 1, 4 or 7 days treated of palbociclib (1 μM). A total of 50 cells were quantified per experiments and graphs display mean ± SEM from three experiments.

5. F

   Quantification of colony forming assays of MCF7, MCF7 p53 KO and T47D cells treated with CDK4/6 inhibitor for 0, 1, 4 or 7 days and then grown at a low density without palbociclib for 14–21 days. Bar graphs display mean data + SEM from 4 to 5 experiments.

Figure 7. Constitutive palbociclib treatment in various tumour lines causes cell cycle delays and DNA damage

1. A

   Quantification of cumulative mitotic entry over time immediately following 1 μM palbociclib treatment in a range of tumour cell lines, as indicated. Mitotic entry of 50 cells per condition per experiment were analysed immediately following drug addition. Graph shows cumulative mean ± SD from three experiments (150 cells total).

2. B, C

   Quantification of the nuclear morphologies (B) or γH2AX‐positive DNA damage foci (C) from different cancer lines, as indicated, treated continuously with palbociclib (1 μM) for 0, 1, 2 or 3 weeks. Either 100 cells (nuclear morphology) or 50 cells (γH2AX) were scored per condition per experiment and bar graphs represent mean data + SEM from three experiments.

3. D

   Quantification of weekly proliferation rate in cells treated as in (B). A total of 60,000 cells of each cell line were cultured in 10 cm dishes with 1 μM palbociclib for a total of 3 weeks. Every 7 days, cells were trypsinized and total cell counts were determined. The 0‐week time point shows the 7‐day fold increase of untreated cells. The data points show fold increase in cell count over each 7‐day period, and bars represent the mean from two experiments.