Continuous CDK4/6i treatment suppresses the growth of drug-resistant cells.

(A) Schematic illustrating the establishment of CDK4/6i-resistant cells, either maintained in drug or withdrawn from treatment, for use in subsequent experiments. (B) Growth curves of parental and drug-resistant cells. Palbociclib (1 µM) was either withdrawn or maintained in drug-resistant cells. Data represent mean ± SD (n = 3 biological replicates). Statistical significance was determined by two-way ANOVA with Tukey’s post-hoc analysis (**** P < 0.0001). (C) Schematic of live-cell reporters for CDK4/6 and CDK2 activities (top) and cell-cycle phase (bottom). (D, E) Representative single-cell traces showing CDK4/6 and CDK2 activities together with Cdt1-degron intensity in parental (D) and drug-resistant (E) MCF-7 cells. Resistant cells were maintained in palbociclib (1 µM). (F, G) Heatmaps of single-cell traces showing CDK4/6 (left) and CDK2 (middle) activities, and Cdt1-degron intensity (right) in drug-resistant cells without (F) or with (G) continuous palbociclib (1 µM). Proliferating cells were identified by sustained CDK2 activity (>1 for >2 hr during 30–48 hr). (H, I) Single-cell traces of CDK2 activity (left) and Cdt1-degron intensity (right), aligned to the end of mitosis (anaphase) in drug-resistant cells without (H) or with (I) continuous palbociclib (1 µM). Based on CDK2 activity (black line), cells were classified as proliferating (red) or quiescent (blue).

CDK4/6i maintenance extends G1-phase progression.

(A) Representative single-cell traces showing CDK4/6 and CDK2 activities (top) and Cdt1-degron intensity (bottom) in parental MCF-7 cells. (B–E) Violin plots showing intermitotic time (n > 600 cells/condition) (B), G1-phase duration (n > 2,000 cells/condition) (C), S-phase duration (n > 400 cells/condition) (D), and G2/M-phase duration (n > 250 cells/condition) (E) in MCF-7 (left) and CAMA-1 (right) cells. Solid and dashed yellow lines indicate mean and median, respectively.

CDK4/6i maintenance induces an ineffective Rb inactivation pathway.

(A) Immunoblot of phosphorylated Rb (S807/811; p-Rb), total Rb (t-Rb), and β-actin protein levels in MCF-7 cells. Drug-resistant cells were harvested two weeks after drug withdrawal. (B) Singe-cell traces of CDK2 activity aligned to mitosis in proliferating (red) and quiescent (blue) MCF-7 cells. Circles indicate the time of fixation and staining (n = 200 cells). (C) Scatter plot of EdU intensity versus time since mitosis in parental cells. The red dotted line marks the threshold for S phase (n = 2,000 cells). (D, E) Percentage of S-phase cells as a function of time since mitosis in proliferating (D) and quiescent (E) cells. Data represent mean ± SD (n = 2 biological replicates). (F) Representative immunostaining of Hoechst, p-Rb, t-Rb, and E2F1 mRNA FISH (top). Processed images showing nucleus segmentation, p-Rb classification, and mRNA puncta detection (bottom). (G) Histogram of cells with p-Rb normalized to t-Rb. (H, I) Percentage of cells with p-Rb (H) and E2F1 mRNA levels (I) as a function of time since mitosis. Data represent mean ± SD (H, n = 3 biological replicates) or mean ± 95% confidence intervals (I, n > 2,500 cells/condition). (J, K) E2F1 mRNA (J) and EdU (K) levels as a function of CDK2 activity. Data represent mean ± 95% confidence intervals (n > 2,500 cells/condition). (L, M) Averaged traces of CDK4/6 (left) and CDK2 (middle) activities and Cdt1-degron intensity (right) (L) and percentage of S-phase cells (M). MCF-7 cells were released after synchronizing with nocodazole (250 nM, 14 hr).

CDK4/6i maintenance suppresses the growth of drug-resistant tumors.

(A) Schematic representation of experimental design. Once tumors reached a volume of 100 mm3, mice were treated with palbociclib. Following the development of resistance, mice were randomly assigned to one of four treatment groups: treatment discontinuation, palbociclib maintenance, switch to ribociclib, or switch to abemaciclib for 32 days. (B) Tumor growth curves showing the establishment of resistance to palbociclib. Horizontal red dotted lines (145−155 mm3) indicate the point at which mice were assigned to second CDK4/6i treatment (n = 20 mice). (C, D) Averaged (C) and individual (D) tumor growth traces following second CDK4/6i treatment. Data represent mean ± SEM (n = 5 mice/condition). Statistical significance was determined using two-way ANOVA with Tukey’s post-hoc analysis (**** P < 0.0001). (E, F) Box plots showing tumor mass (E) and the percentage of Rb-positive cells (F) after second CDK4/6i treatment (n = 5 mice/condition). Statistical significance was determined using one-way ANOVA with Tukey’s post-hoc analysis (** P < 0.05, *** P < 0.01, **** P < 0.0001).

Maintaining CDK4/6i and ET synergistically suppresses the growth of drug-resistant cells.

(A) IC50 values of fulvestrant. Data represent mean ± SD (n = 3 biological replicates). Statistical significance was determined by one-way ANOVA with Tukey’s post-hoc analysis (*** P < 0.001). Growth curves of MCF-7 (left) and CAMA-1 (right) cells resistant to palbociclib and fulvestrant. Cells were maintained with fulvestrant (500 nM), palbociclib (1 µM), or the combination. Data represent mean ± SD (n = 3 biological replicates). Statistical significance was determined by two-way ANOVA with Tukey’s post-hoc analysis (**** P < 0.0001). (C) Violin plots of c-Myc levels measured by immunofluorescence in resistant cells resistant treated with the indicated drugs for one week. Solid and dashed yellow lines represent mean and median, respectively (n > 2,000 cells/condition). (D, E) Heatmaps showing single-cell traces for CDK4/6 (left) and CDK2 (middle) activities and Cdt1-degron intensity (right) in drug-resistant cells maintained with palbociclib (1 µM) and fulvestrant (500 nM) (D) or fulvestrant alone (E). Proliferating cells were identified based on CDK2 activity (>1 for more than 2 hr between 30 and 48 hr). (F) Averaged traces of CDK4/6 (left) and CDK2 (right) activities aligned by mitosis in drug-resistant cells treated with the indicated drugs. Data represent mean ± 95% confidence intervals (n > 1,000 cells/condition). (G) Single-cell traces of Cdt1-degron intensity aligned by mitosis in drug-resistant cells treated with palbociclib (1 µM) (left) or palbociclib + fulvestrant (500 nM) (right). (H, I) Violin plots showing intermitotic time (n > 70 cells/condition) (H), and G1-phase duration (n > 630 cells/condition) (I) in MCF-7 (left) and CAMA-1 (right) cells. Solid and dashed yellow lines indicate mean and median, respectively.

Therapeutic benefit of combining CDK2i with CDK4/6i and ET.

(A, B) Growth curves of MCF-7 (left) and CAMA-1 (right) cells resistant to palbociclib and fulvestrant under various treatment conditions: drug discontinuation, fulvestrant (500 nM) alone, or in combination with palbociclib (1 µM) and/or INX-315 (100 nM). Data represent mean ± SD (n = 3 biological replicates). P values were calculated with two-way ANOVA with Tukey’s post-hoc analysis (**** P < 0.0001). (C) Cell numbers 20 days after drug treatment. Data represent mean ± SD (n = 3 biological replicates). P values were calculated using an unpaired t-test (* P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001). (D) Heatmaps of single-cell traces for CDK4/6 (left) and CDK2 (middle) activities, and Cdt1-degron intensity (right) in palbociclib/fulvestrant resistant MCF-7 cells treated with the triple combination of palbociclib (1 µM), fulvestrant (500 nM), and INX-315 (100 nM) for one week before imaging. Proliferating cells were identified based on CDK2 activity (>1 for more than 2 hr between 30 and 48 hr). (E) Violin plots showing intermitotic time (n > 200 cells/condition) (left) and G1-phase duration (n > 900 cells/condition) (right). Solid and dashed yellow lines indicate mean and median, respectively. (F) Averaged traces of CDK4/6 (left) and CDK2 (right) activities aligned by mitosis in drug-resistant cells treated with the indicated drugs. Data represent mean ± 95% confidence intervals (n > 1,000 cells/condition). (G) Loewe synergy score calculated from percent-inhibition data generated by dual titration of palbociclib (0–10µM) and INX-315 (0–30µM) in MCF-7 cells following 48 hr treatment (n = 3 biological replicates). (H) Heatmaps comparing gene expression profiles for hallmark E2F (left) and MYC (right) targets in drug-resistant MCF-7 cells treated with the indicated drugs for 20 days. Samples were collected as biological duplicates.

Overexpression of cyclin E and A facilitates resistance to the combination of CDK2i and CDK4/6i.

(A, B) Averaged traces of CDK4/6 (left) and CDK2 (right) activities in MCF-7 cells with or without cyclin E1 or A2 overexpression. Cells were treated with doxycycline (500 nM) 6 hr before the addition of INX-315 (1 µM) alone (A) or in combination with palbociclib (1 µM) (B). Data represent mean ± 95% confidence intervals (n > 900 cells/condition). (C, D) Heatmaps of single-cell traces for CDK4/6 (left) and CDK2 (middle) activities, and Geminin-degron intensity (right) in drug-naïve MCF-7 cells without (C) or with (D) cyclin E1 overexpression. Cells were treated with palbociclib (1 µM), INX-315 (100 nM), and doxycycline (500 nM). Persister cells were identified based on CDK2 activity (>1 for more than 2 hr between 30 and 48 hr). (E) Averaged traces of CDK4/6 (left) and CDK2 (middle) activities and Geminin degron intensity (right) in persister cells tolerant to combination palbociclib (1 µM), and INX-315 (100 nM) without or with cyclin E1 or A2 overexpression. Data represent mean ± 95% confidence interval. n > 2500 cells/condition. (F) Growth curves of palbociclib-resistant MCF-7 cells without or with cyclin E1 or A2 overexpression. Cells were treated with palbociclib (1 µM), INX-315 (100 nM), and doxycycline (500 nM). Data represent mean ± SD (n = 3 biological replicates). Statistical significance was determined using two-way ANOVA with Tukey’s post-hoc analysis (**** P < 0.0001). (G) Summary schematic illustrating the mechanisms underlying the benefits of continued CDK4/6i and ET therapies and the introduction of CDK2i in drug-resistant cells.

Validation of drug resistance and visualization of cell-cycle progression.

(A) Dose-response curves of palbociclib showing its effects on the percentage of S-phase cells. Data represent mean ± SEM (n = 3 biological replicates). Solid lines represent sigmoidal best-fit curves. (B) Growth curves of drug-naïve and drug-resistant cells. Palbociclib (1 µM) was either withdrawn or maintained in drug-resistant cells. Data represent mean ± SD (n = 3 biological replicates). Statistical significance was determined using two-way ANOVA with Tukey’s post-hoc analysis (**** P < 0.0001). (C–F) Heatmaps of single-cell traces for CDK4/6 (left) and CDK2 (middle) activities, and Cdt1-degron intensity (right) in various conditions: drug-naïve MCF-7 (C) and MDA-MB-231 (D) cells, and drug-resistant MDA-MB-231 cells without (E) or with (F) continuous palbociclib (1µM) treatment. Proliferating cells were identified based on CDK2 activity (>1 for more than 2 hr between 30 and 48 hr).

Slow cell-cycle progression in drug-resistant cells continuously treated with CDK4/6i.

(A–C) Single-cell traces of CDK4/6 (left) and CDK2 (middle) activities and Cdt1-degron intensity (right) aligned by mitosis in drug-naïve MCF-7 (A) and MDA-MB-231 (B) cells and drug-resistant MDA-MB-231 cells without CDK4/6i treatment (C). The time of mitosis is marked in gray. (D, E) Single-cell trace of CDK4/6 activity corresponding to Figure 1G (D) and 1H (E). (F) Single-cell traces of CDK4/6 (left) and CDK2 (middle) activities and Cdt1-degron intensity (right) aligned by mitosis in drug-resistant MDA-MB-231 cells with continuous palbociclib (1 µM) treatment.

Potential non-canonical role of CDK6 in promoting CDK4/6i resistance.

(A, B) Immunoblot showing CDK6 and β-actin expression in drug-naïve and palbociclib-resistant cells (A) and WT and CDK6-KO cells (B). The numerical values represent the intensity of the CDK6 band, normalized against the intensity of the β-actin band. (C–E) Heatmaps of single-cell traces for CDK4/6 (left) and CDK2 (middle) activities, and Cdt1 degron intensity (right) in drug-naïve (C) and drug-resistant WT (D) and CDK6-KO (E) cells with continued palbociclib (1 µM) treatment. Proliferating cells were identified based on CDK2 activity (>1 for >2 hr during 30–48 hr). (F) Averaged traces of CDK4/6 (left) and CDK2 (right) activities. Data represent mean ± 95% confidence intervals (n > 7,000 cells/condition).

Incomplete Rb loss mediates the extended G1-phase progression.

(A) Immunoblot showing total Rb and β-actin in WT, Rb-KO, and palbociclib-resistant cells. (B) G1-phase duration in drug-resistant WT and Rb-KO cells with continued palbociclib (1 µM) treatment. Solid and dashed yellow lines represent mean and median, respectively (n > 4,500 cells). (C) Averaged CDK4/6 (left) and CDK2 (right) activities aligned by mitosis in drug-resistant WT and Rb-KO cells with continued palbociclib (1 µM) treatment (n > 7,500 cells).

Slow and heterogeneous G1/S transition in drug-resistant cells maintained with CDK4/6i treatment.

(A) Single-cell traces showing CDK2 activity aligned by mitosis in proliferating (red) and quiescent (blue) drug-naïve MCF-7 cells. Circles indicate the time of fixation and staining (n = 200 cells). (B, C) Scatterplot of EdU intensity against time since mitosis in drug-resistant cells without (B) and with (C) continuous palbociclib (1 µM) treatment. Red dotted line indicates the S-phase threshold (n = 2,000 cells/condition). (D, E) Scatterplot of EdU intensity (D) and E2F1 mRNA levels (E) against CDK2 activity (n = 2,000 cells/condition).

c-Myc overexpression facilitates CDK4/6i resistance by accelerating cell-cycle progression.

(A) Growth curves of drug-resistant cells with treatment discontinuation or continuous palbociclib (1 µM) treatment, without or with c-Myc overexpression. Data represent mean ± SD (n = 3 biological replicates). Statistical significance was determined using two-way ANOVA with Tukey’s post-hoc analysis (**** P < 0.0001). (B) Intermitotic time of drug-resistant cells without and with c-Myc overexpression. Solid and dashed yellow lines represent mean and median, respectively (n > 4,500 cells/condition). (C, D) Heatmaps of single-cell traces for CDK4/6 (left) and CDK2 (middle) activities, and Cdt1-degron intensity (right) in drug-resistant cells undergoing continuous palbociclib (1 µM) treatment without (C) and with (D) c-Myc overexpression. Proliferating cells were identified based on CDK2 activity (>1 for more than 2 hr between 30 and 48 hr).

Maintaining CDK4/6i and ET synergistically suppresses the growth of drug-resistant cells.

(A) Growth curves of MCF-7 (left) and CAMA-1 (right) cells resistant to palbociclib and fulvestrant withdrawn from drug treatment. Data represent the mean ± SD (n = 3 biological replicates). (B) Cell numbers 20 days after treatment with DMSO, palbociclib (1 µM), fulvestrant (500 nM), or their combination. Data represent mean ± SD (n = 3 biological replicates). Statistical significance was determined with an unpaired t-test (** P < 0.01, *** P < 0.001, **** P < 0.0001). (C) Heatmaps of single-cell traces for CDK4/6 (left) and CDK2 (middle) activities, and Cdt1-degron intensity (right) in combination drug-resistant cells maintained with palbociclib (1 µM). Proliferating cells were identified based on CDK2 activity (>1 for more than 2 hr between 30 and 48 hr). (D) Single-cell traces of CDK2 activity aligned by mitosis in combination drug-resistant cells treated with continuous fulvestrant (500 nM) (left), palbociclib (1 µM) (middle), or their combination (right). Based on CDK2 activity, cells were classified into proliferation (red) or quiescence (blue). The time of mitosis is marked in gray. (E) Single-cell traces of Cdt1-degron intensity aligned by mitosis in combination drug-resistant cells treated with continuous fulvestrant (500 nM) alone. (F, G) Violin plots showing S-phase duration (n > 200 cells/condition) (F) and G2/M-phase duration (n > 20 cells/condition) (G) in MCF-7 (left) and CAMA-1 (right) cells. Solid and dashed yellow lines indicate mean and median, respectively.

Combining CDK2i with CDK4/6i and ET effectively suppresses the growth of drug-resistant cells.

(A–D) Heatmaps of single-cell traces for CDK4/6 (left) and CDK2 (middle) activities, and Cdt1-degron intensity (right) in drug-resistant cells under various conditions: treatment discontinuation (A), fulvestrant (500 nM) alone (B), fulvestrant + palbociclib (1 µM) (C), or fulvestrant + INX-315 (100 nM) (D) over one week prior to imaging. Proliferating cells were identified based on CDK2 activity (>1 for more than 2 hr between 30 and 48 hr). (E) Violin plots showing S-phase duration (n > 400 cells/condition) (left) and G2/M-phase duration (n > 150 cells/condition) (right) in drug-resistant MCF-7 cells treated with the indicated drug. Solid and dashed yellow lines indicate mean and median, respectively. (F) Two-dimensional titration of palbociclib (0–10 µM) and INX-315 (0–30 µM) in MCF-7 cells for 48 hr (n = 3 biological replicates). (G) GSEA plots for hallmark E2F (left) and MYC (right) target genes in palbociclib/fulvestrant-resistant MCF-7 cells treated with the triple combination, compared to other drug conditions.

Cyclin E or A overexpression attenuates full CDK2 inhibition by CDK4/6i and CDK2i combination.

(A) Immunoblot showing the expression of GAPDH with cyclin A (left) or cyclin E (right) in MCF-7 cells with or without cyclin E1 or A2 overexpression. Cells were treated with doxycycline (500 nM) for 24 hr. (B–G) Averaged traces of CDK4/6 (left) and CDK2 (right) activities in MCF-7 cells with or without cyclin E1 and A2 overexpression. Cells were treated with indicated doxycycline (500 nM) 6 hr before the addition of indicated drugs. Data represent mean ± 95% confidence intervals (n > 850 cells/condition).

Cyclin E overexpression facilitates adaptation to CDK4/6i and CDK2i combination.

Heatmaps of single-cell traces for CDK4/6 (left) and CDK2 (middle) activities, and Geminin-degron intensity (right) in drug-naïve MCF-7 cells overexpressing cyclin A and treated with palbociclib (1 µM) and INX-315 (100 nM). Proliferating cells were identified based on CDK2 activity (>1 for more than 2 hr between 30 and 48 hr).