Elicit: CDK4/6 Inhibition and G1 to S Arrest

Skip to main content

CDK4/6 Inhibition and G1 to S Arrest

Research reportView only

Create alertChat

May 5, 2026

CDK4/6 inhibition cell-cycle arrest G1 to S palbociclib

Palbociclib consistently induces G1 to S cell cycle arrest through CDK4/6 inhibition and Rb pathway blockade, achieving complete arrest in 87-90% of Rb-proficient hormone receptor-positive breast cancers, though the arrest is cytostatic and requires continuous therapy to maintain.

Abstract

Palbociclib consistently induces G1 to S phase cell cycle arrest across multiple cancer types through inhibition of CDK4/6, preventing retinoblastoma (Rb) phosphorylation and suppressing E2F-target gene expression. In neoadjuvant trials of hormone receptor-positive breast cancer, complete cell cycle arrest (Ki-67 ≤2.7%) was achieved in 87-90% of patients when palbociclib was combined with endocrine therapy, representing a threefold increase over endocrine therapy alone. This profound antiproliferative effect translated to clinically meaningful progression-free survival benefits in advanced disease, with median PFS of 9.2 months versus 3.8 months for combination therapy versus endocrine therapy alone (HR 0.42). However, the arrest is cytostatic rather than cytotoxic—concurrent suppression of apoptosis limits tumor regression, and continuous therapy is necessary to maintain cell cycle arrest as discontinuation allows Ki-67 rebound. Mechanistically, response depends on intact Rb function, with RB1 loss conferring complete resistance, while high cyclin E levels and elevated CDK2 activity enable G1/S bypass. Sustained mTORC1 activity during palbociclib treatment converts reversible arrest to irreversible senescence, and autophagy induction serves as an adaptive resistance mechanism that can be targeted through combination strategies.

The clinical benefit of palbociclib-induced G1 arrest is biomarker-dependent and context-specific. Tumors with intact Rb, absent or low cyclin E (particularly low-molecular-weight isoforms), and adequate p27 levels demonstrate maximal sensitivity, with progression-free survival exceeding 400 months in optimally selected populations. Grade 3/4 neutropenia occurs in 51-62% of patients but is largely asymptomatic, with febrile neutropenia rates of only 0.6% and treatment discontinuation rates of 2.6%. Combination strategies targeting autophagy, PI3K/MEK signaling, or mTORC1 can overcome resistance mechanisms in biomarker-selected populations. The evidence demonstrates that palbociclib reliably induces G1 arrest in Rb-proficient cells, but durable clinical benefit requires continuous dosing, appropriate biomarker selection, and in some contexts, rational combination therapy to prevent adaptive resistance.

Methods

We analyzed 10 sources from an initial pool of 200, using 8 screening criteria. Each paper was reviewed for 8 key aspects that mattered most to the research question. More on methods

Records from Elicit search

n = 200

Papers screened using: Palbociclib Focus, Cell Cycle Measurement, Study Model Type, Control Groups, Study Type, Measurable Outcomes, Publication Type, Sample Size

n = 200

Papers screened out

n = 190

Papers included for extraction

n = 10

Press enter or space to select a node.You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.

Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.

Paper search

We performed a semantic search across over 138 million academic papers from the Elicit search engine, which includes all of Semantic Scholar and OpenAlex.

We ran this query: “CDK4/6 inhibition cell-cycle arrest G1 to S palbociclib”

The search returned 200 total results from Elicit.

We retrieved 200 papers most relevant to the query for screening.

Screening

We screened in sources based on their abstracts that met these criteria:

We considered all screening questions together and made a holistic judgement about whether to screen in each paper.

Data extraction

We asked a large language model to extract each data column below from each paper. We gave the model the extraction instructions shown below for each column.

Extract study design and setting for CDK4/6 inhibition research, including:

Extract complete details about CDK4/6 inhibitor treatment regimen, including:

Extract all measurements of G1 to S cell cycle arrest and related markers, including:

Extract mechanistic findings related to CDK4/6 inhibition and G1/S arrest, including:

Extract clinical efficacy outcomes specifically related to CDK4/6 inhibitor treatment, including:

Extract findings about cellular consequences of CDK4/6-induced G1 arrest, including:

Extract factors affecting response to CDK4/6 inhibition and G1/S arrest, including:

Extract safety and tolerability data for CDK4/6 inhibitor treatment, including:

Results

Characteristics of Included Studies

The systematic review included 10 studies examining CDK4/6 inhibition by palbociclib and its effects on G1 to S phase cell cycle arrest. Studies comprised 6 clinical trials and 4 preclinical investigations. Full text was available for 6 studies, while 4 were assessed from abstracts only.

Study

Full Text Retrieved?

Study Type

Cancer Type and Subtype

Sample Size

Setting

Ma et al., 2017

Yes

Phase II neoadjuvant clinical trial

ER+/HER2- breast cancer

50 patients (18 premenopausal, 32 postmenopausal)

Clinical stage II-III patients

Johnston et al., 2019

Yes

Phase II randomized trial

ER+/HER2- breast cancer

307 patients

Postmenopausal women with primary tumors ≥2.0 cm

Arnedos et al., 2018

No

Randomized clinical trial

Early breast cancer (93% HR+, 8% HER2+)

74 palbociclib, 26 control

Early breast cancer patients

Maskey et al., 2020

Yes

Preclinical, in vitro

ER+ breast cancer

Multiple cell lines (MCF7, T47D, CAMA1)

ER+ breast cancer cell lines

Turner et al., 2015

Yes

Phase 3 clinical trial

Advanced HR+/HER2- breast cancer

521 patients

Pre- and postmenopausal women with relapsed/progressed disease

Vijayaraghavan et al., 2017

Yes

Preclinical (in vitro, in vivo) and clinical cohort

ER+ breast cancer and other solid tumors

109 patients in clinical cohort; multiple cell lines and xenografts

Cell lines (MCF7, T47D, ZR75-1), xenografts, advanced ER+ breast cancer patients

DeMichele et al., 2014

No

Phase II clinical trial

Advanced breast cancer (84% HR+/HER2-, 5% HR+/HER2+, 11% HR-/HER2-)

37 patients

Metastatic breast cancer, Rb+ with measurable disease

Tien & Sadar, 2021

No

Preclinical

Castration-resistant prostate cancer

Not mentioned

Human xenografts and cultured cells

Asghar et al., 2017

Yes

Preclinical

Triple-negative breast cancer (TNBC), LAR subtype

Multiple cell lines and xenografts

TNBC cell lines and MDA-MB-453 LAR xenografts

Kumarasamy et al., 2020

No

Preclinical

ER+ breast cancer and pancreatic cancer

Not mentioned

ER+ xenografts and pancreatic cancer PDX models

toof

Pageof

Clinical trials predominantly focused on hormone receptor-positive breast cancer, with sample sizes ranging from 37 to 521 patients. The two largest trials were phase 3 (Turner et al., 521 patients) and phase 2 (Johnston et al., 307 patients). Neoadjuvant trials examined palbociclib in combination with endocrine therapy, while the advanced disease trial combined palbociclib with fulvestrant. Preclinical studies utilized established cell lines and xenograft models to investigate mechanistic aspects of CDK4/6 inhibition.

Treatment Regimens

Palbociclib dosing across studies followed established protocols, with clinical trials consistently using 125 mg daily on a 21-days-on, 7-days-off schedule, while preclinical studies utilized concentrations ranging from 500 nM to 1 μM in vitro.

Study

Palbociclib Dose and Schedule

Duration

Combination Therapy

Route

Ma et al., 2017

125 mg daily, days 1-21 of 28-day cycle

Four 28-day cycles, plus optional 10-12 day cycle 5

Anastrozole 1 mg daily; goserelin if premenopausal

Oral

Johnston et al., 2019

125 mg/day, 21-days-on, 7-days-off

14 weeks

Letrozole

Oral

Arnedos et al., 2018

125 mg daily

14 days until day before surgery

None mentioned

Oral

Maskey et al., 2020

500 nM

48 hours and various durations

Approved with endocrine therapies clinically, not specified in study

Added to cell culture medium

Turner et al., 2015

Not mentioned

Not mentioned

Fulvestrant; goserelin for pre/perimenopausal women

Not mentioned

Vijayaraghavan et al., 2017

In vitro: ≤1 μM for 6 days; In vivo: 25 mg/kg/day for 7 days

In vitro: 6 days treatment + 4 days recovery; In vivo: 7 days

Autophagy inhibitor HCQ

Oral gavage (in vivo)

DeMichele et al., 2014

125 mg orally, days 1-21 of 28-day cycle

Not mentioned

Not mentioned

Oral

Tien & Sadar, 2021

Not mentioned

Not mentioned

EPI-7170 (sequential or concomitant)

Not mentioned

Asghar et al., 2017

In vitro: 500 nM; In vivo: 50 mg/kg daily

In vitro: ≥2 weeks; In vivo: 21 consecutive days

Pictilisib, taselisib, AZD2014

Oral (in vivo)

Kumarasamy et al., 2020

Not mentioned

Not mentioned

MEK inhibitor in pancreatic cancer PDX models

Not mentioned

toof

Pageof

The neoadjuvant studies demonstrated the importance of treatment duration and continuity. Ma et al. showed that Ki67 levels rebounded at surgery following palbociclib washout, but this rebound was suppressed by an additional cycle 5 of palbociclib immediately before surgery, suggesting continuous therapy may be necessary to maintain antiproliferative effects.

Cell Cycle Arrest and Proliferation Markers

Palbociclib consistently induced profound suppression of cellular proliferation across studies, with Ki-67 serving as the primary biomarker for cell cycle arrest.

Study

Ki-67 Baseline

Ki-67 Post-Treatment

% Change/CCCA Rate

Time Points

Other Markers

Ma et al., 2017

C0D1

C1D1 (anastrozole alone), C1D15 (palbociclib added)

CCCA rate: 26% (C1D1) vs 87% (C1D15)

Baseline, C1D1, C1D15, surgery

PAM50 11-gene proliferation score

Johnston et al., 2019

Baseline

14 weeks

Median log-fold change: -4.1 (palbociclib+letrozole) vs -2.2 (letrozole); Geometric mean: -97.4% vs -88.5%; CCCA: 90% vs 59%

Baseline, 2 weeks, 14 weeks

None mentioned

Arnedos et al., 2018

Day 1

Day 15

Antiproliferative response: 58% (palbociclib) vs 12% (control)

Day 15

Phospho-Rb

Maskey et al., 2020

Not mentioned

48 hours

Not mentioned

48 hours for DNA content

Cell cycle phase distribution measured using DAPI staining

Turner et al., 2015

Not mentioned

Not mentioned

Not mentioned

Not mentioned

Not mentioned

Vijayaraghavan et al., 2017

Not mentioned

After treatment and recovery phases

Palbociclib induced G1 arrest; CCCA achieved in 90% at appropriate doses

Not mentioned

BrdU (decreased in palbociclib-treated cells); Cell cycle phase distribution by propidium iodide

DeMichele et al., 2014

Not mentioned

Not mentioned

Not mentioned

Not mentioned

Not mentioned

Tien & Sadar, 2021

Not mentioned

Not mentioned

Doubling time increased to >63 hours vs 25 hours (control)

Not mentioned

Cell cycle phase distribution: combination prevented G1 and G2-M progression, caused S-phase arrest

Asghar et al., 2017

Not mentioned

48 hours post-treatment

Cell cycle length: 20 hours (vehicle) vs 38 hours (palbociclib) in CDK2 high cells

2 hours post-cytokinesis, 48 hours post-treatment

CDK2 activity reporter

Kumarasamy et al., 2020

Not mentioned

Not mentioned

Not mentioned

Not mentioned

Not mentioned

toof

Pageof

The neoadjuvant trials provided the most detailed quantitative assessment of proliferative suppression. In the NeoPalAna trial, the complete cell cycle arrest rate (defined as Ki67 ≤2.7%) increased dramatically from 26% with anastrozole monotherapy to 87% after adding palbociclib. Similarly, Johnston et al. demonstrated that 90% of patients achieved complete cell cycle arrest with palbociclib plus letrozole compared to only 59% with letrozole alone. The median log-fold change in Ki-67 was significantly greater with combination therapy (-4.1 vs -2.2), corresponding to near-complete suppression of proliferation (-97.4% vs -88.5% geometric mean change). These findings indicate that CDK4/6 inhibition substantially enhances the antiproliferative effects of endocrine therapy beyond what can be achieved with endocrine therapy alone.

Mechanistic Foundations of Cell Cycle Arrest

Multiple studies investigated the molecular mechanisms by which palbociclib induces G1 arrest, with particular focus on the Rb pathway and downstream effectors.

Rb Pathway Modulation

Retinoblastoma (Rb) phosphorylation emerged as a central mechanistic determinant. Arnedos et al. demonstrated that palbociclib treatment led to significantly greater decreases in phospho-Rb compared to control, and changes in Ki67 correlated with changes in phospho-Rb (Spearman r=0.41). This relationship between Rb dephosphorylation and antiproliferative response suggests that early decreases in Rb phosphorylation could potentially identify patients with primary resistance. Kumarasamy et al. confirmed that activation of RB and inhibition of CDK2 activity emerged as determinants of sensitivity to CDK4/6 inhibition, with RB loss rendering cells completely independent of CDK4 and CDK6.

Ma et al. found that resistance to palbociclib was associated with persistent E2F-target gene expression, indicating ongoing Rb pathway activity despite CDK4/6 inhibition. Specifically, nonluminal subtypes demonstrated persistent elevation of CCND3, CCNE1, and CDKN2D, markers of continued E2F activity that bypass palbociclib’s mechanism.

Cell Cycle Checkpoint Proteins

The p27 protein emerged as a critical modulator of sensitivity. Kumarasamy et al. showed that protein levels of p27 were associated with cell cycle plasticity and correlated with sensitivity to CDK4/6 inhibition. Exogenous overexpression and pharmacologic induction of p27 via SKP2 inhibition or MEK/ERK pathway targeting enhanced the cytostatic effect of CDK4/6 inhibitors. In ER+ xenograft models, few cells retained RB phosphorylation during palbociclib treatment, which was associated with limited p27 protein levels, suggesting that p27 levels influence the durability of palbociclib’s effects.

Downstream Signaling Pathways

Maskey et al. identified mTORC1 activity as a critical determinant of cell fate during CDK4/6 inhibition. In CAMA1 cells, mTORC1 activity remained elevated during palbociclib treatment, while in MCF7 and T47D cells, mTORC1 was suppressed. Importantly, inhibition of mTORC1 signaling via rapamycin or Raptor knockdown during palbociclib treatment blocked the induction of complete senescence in CAMA1 cells. Genetic depletion of TSC2, a negative regulator of mTORC1, resulted in sustained mTORC1 activity during palbociclib treatment and evoked a complete senescence response in MCF7 cells, demonstrating that persistent mTORC1 signaling can convert reversible to irreversible growth arrest.

Vijayaraghavan et al. discovered that autophagy is induced as a stress response to palbociclib. Palbociclib treatment decreased pRb and total Rb levels, but autophagy activation served as a resistance mechanism by degrading reactive oxygen species and potentially reversing G1 arrest. Combined inhibition of CDK4/6 and autophagy produced synergistic effects, suggesting that autophagy represents an adaptive survival pathway during cell cycle arrest.

Clinical Efficacy Outcomes

Clinical efficacy varied by disease setting, with the most robust data emerging from trials in hormone receptor-positive breast cancer.

Neoadjuvant Setting

Study

Clinical Response Rate

Pathological Response

Time-Based Outcomes

Maintenance of Arrest

Ma et al., 2017

80% (exam), 41% (ultrasound), 52% (mammogram)

No pathologic complete responses; significant reduction in tumor stages

Not mentioned

Continuous therapy necessary to maintain CCCA

Johnston et al., 2019

CR+PR: 54.3% (palbociclib+letrozole) vs 49.5% (letrozole); PD: 3.2% vs 5.4%

Not mentioned

Not mentioned

Not mentioned

Arnedos et al., 2018

Antiproliferative response: 58% (palbociclib) vs 12% (control); Ki67 decrease significant

Not mentioned

Not mentioned

Not mentioned

toof

Pageof

In the neoadjuvant setting, palbociclib demonstrated robust antiproliferative activity but variable impact on tumor size reduction. Johnston et al. found that adding palbociclib to letrozole significantly enhanced Ki-67 suppression but did not increase clinical response rates over 14 weeks, possibly related to concurrent reduction in apoptosis. Ma et al. achieved an 80% clinical response rate by physical examination, though ultrasound and mammogram assessments showed lower rates (41% and 52%, respectively). No pathologic complete responses were observed, but significant reductions in tumor stage occurred.

Advanced Disease Setting

Study

CBR

PFS

Overall Survival

Response Duration

Turner et al., 2015

Not mentioned

Median 9.2 months (palbociclib+fulvestrant) vs 3.8 months (placebo+fulvestrant); HR 0.42

Not mentioned

Not mentioned

Vijayaraghavan et al., 2017

Not mentioned

Palbociclib doubles PFS vs letrozole or fulvestrant alone; Rb+/LMWE- tumors: median 436.5 months with letrozole, 10.7 months with fulvestrant

Not mentioned

Not mentioned

DeMichele et al., 2014

19% overall, 21% in HR+, 29% in HR+/HER2- with ≥2 prior endocrine therapies

Median 3.7 months overall

Not mentioned

Not mentioned

toof

Pageof

The PALOMA-3 trial demonstrated that palbociclib combined with fulvestrant resulted in significantly longer progression-free survival (9.2 vs 3.8 months, HR 0.42) in patients with hormone receptor-positive metastatic breast cancer who had progressed on prior endocrine therapy. Vijayaraghavan et al. reported that palbociclib doubled PFS compared to endocrine therapy alone, with particularly prolonged PFS in Rb-positive, low-molecular-weight cyclin E (LMWE)-negative tumors (median 436.5 months with letrozole). DeMichele et al. observed modest clinical benefit rates of 19% overall, increasing to 29% in HR+/HER2- patients who had failed at least two prior endocrine therapies, with median PFS of 3.7 months.

Preclinical Efficacy

Asghar et al. demonstrated tumor reductions in 7 out of 10 mice treated with palbociclib in the LAR subtype of triple-negative breast cancer xenografts. Cell cycle length was prolonged in CDK2-high cells (20 hours vehicle vs 38 hours palbociclib), though cells showed adaptation to CDK4/6 inhibition over time.

Cellular Consequences and Senescence Phenotypes

Studies revealed heterogeneity in cellular responses to CDK4/6 inhibition, with some cell lines undergoing reversible arrest while others entered irreversible senescence.

Reversibility of Growth Arrest

Maskey et al. demonstrated that ER+ breast cancer cell lines exhibited distinct responses to palbociclib: MCF7 and T47D cells showed reversible G1-phase arrest with an incomplete senescence phenotype, whereas CAMA1 cells underwent irreversible cell cycle arrest and complete senescence. This difference correlated with mTORC1 signaling patterns, with sustained activity promoting complete senescence. Ma et al. confirmed the reversibility in clinical samples, showing that Ki67 levels rebounded at surgery after palbociclib discontinuation, though this rebound was suppressed by continuing palbociclib therapy.

Vijayaraghavan et al. found that reversibility was dose-dependent: low doses of palbociclib resulted in reversible G1 arrest, while higher doses led to irreversible growth inhibition. Palbociclib induced senescence markers including increased SA-β-gal activity and cellular complexity, but critically, did not induce apoptosis as evidenced by lack of cleaved PARP and caspase activity.

Autophagy and Survival Responses

Autophagy emerged as an important adaptive mechanism. Vijayaraghavan et al. showed that palbociclib induced autophagy as a stress response, and combining palbociclib with autophagy inhibitors like hydroxychloroquine (HCQ) led to sustained growth inhibition and irreversible senescence without inducing apoptosis. SA-β-gal staining confirmed senescence induction both in vitro and in vivo.

Apoptosis Markers

Johnston et al. observed suppression of cleaved PARP with palbociclib plus letrozole treatment, indicating reduced rather than increased apoptosis. The median log-fold suppression of cleaved PARP was greater with combination therapy (-0.80 vs -0.42). This concurrent reduction in both proliferation and apoptosis may explain why enhanced antiproliferative effects did not translate to proportional increases in clinical response rates.

Predictive Biomarkers and Resistance Mechanisms

Substantial heterogeneity in treatment response prompted investigation of predictive biomarkers and resistance pathways.

Rb Status as a Core Biomarker

Retinoblastoma status emerged as the most fundamental predictor of response. Ma et al. identified RB1 mutations, particularly frameshift mutations, as associated with resistance to palbociclib in HER2-enriched tumors, though missense RB1 mutations were found in tumors sensitive to treatment. Kumarasamy et al. confirmed that RB loss rendered cells completely independent of CDK4/6 activity, with RB phosphorylation status serving as a marker for treatment failure. Vijayaraghavan et al. demonstrated that Rb-positive status predicted sensitivity, with Rb-positive/LMWE-negative tumors showing the longest progression-free survival. Asghar et al. confirmed that loss of RB1 caused resistance to CDK4/6 inhibition through disruption of the CDK4/6-RB1 axis that controls the restriction point in G1 phase.

Cyclin E and CDK2 Activity

Cyclin E levels, particularly the low-molecular-weight isoform (LMWE), predicted resistance. Ma et al. found that CCNE1 gain was associated with resistance to CDK4/6 inhibition, likely through CDK2 activation. Vijayaraghavan et al. showed that Rb-positive but LMWE-positive tumors had reduced sensitivity, with overexpression of LMWE conferring resistance. Asghar et al. revealed that high cyclin E1 expression activated CDK2 and was dysregulated in resistant cells, with palbociclib-resistant basal-like TNBC cells exiting mitosis directly into a proliferative state with high CDK2 activity, bypassing the need for CDK4/6. Johnston et al. noted that response to palbociclib was correlated with RB1 mutation status but occurred independently of PIK3CA or PTEN mutations.

Molecular Subtypes

Ma et al. demonstrated that luminal subtypes (LumA and LumB) were more responsive to palbociclib, while nonluminal subtypes including basal-like and HER2-enriched tumors showed resistance. Persistent E2F-target gene expression, indicated by elevated CCND3, CCNE1, and CDKN2D, was linked to resistance. Asghar et al. found that the luminal androgen receptor (LAR) subtype of triple-negative breast cancer was highly sensitive to CDK4/6 inhibition, while basal-like subtypes were resistant.

Downstream Signaling Pathways

Kumarasamy et al. identified p27 protein levels as associated with cell cycle plasticity and sensitivity, with targeting of the MEK/ERK pathway and SKP2 inhibition emerging as strategies to enhance response. Maskey et al. showed that sustained mTORC1 activity during palbociclib treatment promoted complete senescence, suggesting mTORC1 inhibition via rapamycin or Raptor knockdown as a combination strategy.

Combination Strategies to Overcome Resistance

Several effective combinations emerged from these studies:

Safety and Tolerability Profile

Hematologic toxicities, particularly neutropenia, dominated the adverse event profile across clinical trials, though these were generally manageable with dose modifications.

Study

Grade 3/4 Neutropenia

Grade 3/4 Other Cytopenias

Dose Modifications

Non-Hematologic Toxicities

Ma et al., 2017

G3: 22%, G4: 4%

Leukopenia mentioned

14% required dose reductions

Fatigue, rash; no G4+ non-hematologic AEs

Johnston et al., 2019

Part of 49.8% G3+ toxicity

Asymptomatic neutropenia primary cause

21.6% interruptions/delays, 2.0% dose reductions

Not specified

Turner et al., 2015

62.0%

Leukopenia 25.2%, anemia 2.6%, thrombocytopenia 2.3%

Discontinuation: 2.6% palbociclib, 1.7% placebo

Fatigue 2.0%; febrile neutropenia 0.6%

Vijayaraghavan et al., 2017

56%

Leukopenia 25.2%

Higher doses (75-150 mg/kg) caused significant weight loss

Not specified; combination with HCQ well tolerated

DeMichele et al., 2014

51%

Anemia 5%, thrombocytopenia 22%

51% dose reductions, 24% interruptions

Not mentioned; cytopenias uncomplicated and easily managed

toof

Pageof

Turner et al. reported the highest rate of grade 3/4 neutropenia at 62.0%, compared to 0.6% in the placebo group, along with leukopenia (25.2%), anemia (2.6%), and thrombocytopenia (2.3%). Critically, febrile neutropenia remained rare at 0.6% in both treatment arms, and discontinuation rates due to adverse events were low (2.6% with palbociclib vs 1.7% with placebo). Johnston et al. found that 49.8% of patients experienced grade 3 or greater toxicity with palbociclib plus letrozole versus 17.0% with letrozole alone, primarily driven by asymptomatic neutropenia, with treatment interruptions or delays in 21.6% and dose reductions in only 2.0%.

DeMichele et al. emphasized that cytopenias were uncomplicated and easily managed with dose reduction, with 51% of patients requiring dose modifications and 24% experiencing treatment interruptions. Ma et al. observed grade 3 neutropenia in 22% and grade 4 in 4% of patients, with 14% requiring dose reductions due to neutropenia, elevated transaminases, or rash. No grade 4 or higher non-hematologic adverse events occurred.

Vijayaraghavan et al. noted in preclinical models that optimization of palbociclib dosing was crucial, as higher doses (75 or 150 mg/kg) caused significant body weight loss, but combination with hydroxychloroquine was well tolerated without changes in body weight or blood counts.

Synthesis

The body of evidence reveals consistent mechanisms but variable clinical outcomes that require reconciliation. While palbociclib uniformly induced G1 cell cycle arrest across studies, the durability of this arrest and translation to clinical benefit varied substantially based on molecular context, treatment duration, and combination strategies.

Context-Dependent Response Patterns

The heterogeneity in clinical benefit rates—ranging from 19% in heavily pretreated advanced disease to 87% complete cell cycle arrest in neoadjuvant settings—reflects differences in disease burden, prior treatment exposure, and measurement endpoints rather than conflicting results. In the neoadjuvant setting with less advanced disease, both Ma et al. and Johnston et al. achieved >85% complete cell cycle arrest rates, demonstrating maximal antiproliferative activity when palbociclib is combined with endocrine therapy in treatment-naive tumors. Conversely, DeMichele et al.’s lower clinical benefit rate of 19% occurred in patients with a median of 2 prior cytotoxic regimens, suggesting that heavily pretreated tumors may harbor additional resistance mechanisms beyond those addressed by CDK4/6 inhibition alone.

The progression-free survival benefit also follows a dose-response pattern by line of therapy. Turner et al. achieved 9.2 months median PFS in endocrine-resistant disease, while Vijayaraghavan et al. reported median PFS exceeding 400 months in optimal biomarker-selected populations (Rb+/LMWE-). Both findings are valid within their respective contexts: the former represents a clinically heterogeneous population without biomarker selection, while the latter represents a molecularly defined subset with intact G1/S checkpoint machinery.

Mechanisms Explaining Divergent Cellular Responses

The contrast between reversible and irreversible growth arrest observed by Maskey et al. and Vijayaraghavan et al. can be mechanistically explained through mTORC1 activity and autophagy induction. CAMA1 cells maintained elevated mTORC1 signaling during palbociclib treatment, leading to irreversible senescence, whereas MCF7 and T47D cells suppressed mTORC1 and experienced reversible arrest. This divergence is not contradictory but rather demonstrates that cellular context—specifically basal mTORC1 activity and the cell’s ability to modulate this pathway—determines whether CDK4/6 inhibition causes cytostatic or senescent outcomes. The finding that TSC2 depletion converted MCF7 cells from reversible to irreversible arrest provides direct mechanistic evidence that sustained mTORC1 activity is sufficient to drive complete senescence.

Similarly, Vijayaraghavan et al.’s observation that autophagy serves as an adaptive resistance mechanism explains why some cells escape permanent growth inhibition. Autophagy activation allows cells to survive metabolic stress during G1 arrest, potentially explaining the reversibility observed in certain contexts. The synergy between palbociclib and autophagy inhibitors validates that blocking this survival pathway converts cytostatic arrest into terminal senescence.

Biomarker-Driven Sensitivity Predictions

The apparent contradiction between studies reporting universal benefit and those identifying biomarker-dependent responses reflects measurement granularity rather than conflicting findings. While Turner et al. found consistent benefits across subgroups in their phase 3 trial, this population-level observation does not negate the molecular determinants identified in more detailed mechanistic studies.

Ma et al., Vijayaraghavan et al., and Asghar et al. converge on a unified model: intact Rb and absent or low cyclin E (particularly LMWE) predict maximal sensitivity, while RB1 loss or LMWE overexpression confers resistance. Kumarasamy et al.’s identification of p27 levels as modulators of sensitivity adds nuance to this model—p27 acts as a rheostat determining the threshold at which cells commit to arrest versus escaping through CDK2 activation. Asghar et al.’s single-cell analysis revealed that sensitive LAR cells exit mitosis with low CDK2 activity and require CDK4/6 for reentry, whereas resistant basal-like cells maintain high CDK2 activity post-mitosis, bypassing the CDK4/6 requirement. Both phenotypes coexist within TNBC but respond differently based on their intrinsic cell cycle dynamics.

Non-Linear Relationships: Proliferation vs. Clinical Response

Johnston et al.’s finding that enhanced Ki-67 suppression did not translate to increased clinical response rates initially appears contradictory but is explained by the concurrent suppression of apoptosis. The reduction in cleaved PARP indicates that while palbociclib profoundly arrests proliferation, it simultaneously protects cells from apoptotic death. This creates a cytostatic rather than cytotoxic effect, where tumor cells remain viable but non-proliferative. Over the 14-week treatment period, this translates to disease stabilization rather than tumor regression, explaining why complete cell cycle arrest rates of 90% do not correspond to 90% clinical responses.

This cytostatic mechanism also explains Ma et al.’s observation that continuous therapy is necessary to maintain antiproliferative effects. When palbociclib is discontinued, arrested cells with intact Rb and no terminal senescence can resume cycling once CDK4/6 activity is restored. The addition of cycle 5 palbociclib immediately before surgery suppressed Ki67 rebound, demonstrating that the arrest is maintained only through continuous CDK4/6 inhibition in populations that have not undergone irreversible senescence.

Combination Strategies Addressing Distinct Resistance Nodes

The diverse combination strategies emerging from these studies—autophagy inhibition, PI3K inhibition, MEK inhibition, and mTORC1 modulation—target mechanistically distinct resistance pathways that become activated during G1 arrest. Autophagy inhibition blocks stress-induced survival responses, PI3K inhibition addresses PIK3CA-mutant tumors where constitutive signaling bypasses G1 arrest requirements, and MEK inhibition upregulates p27 to lower the threshold for CDK4/6 dependence. These are not competing strategies but complementary approaches applicable to different molecular contexts.

The MEK combination is particularly rational given Kumarasamy et al.’s demonstration that p27 induction via MEK/ERK pathway inhibition enhances palbociclib efficacy. This mechanistically addresses the subset of tumors with low basal p27 that maintain cell cycle plasticity. Similarly, Tien & Sadar’s sequential dosing strategy in androgen receptor-positive cancers recognizes that different agents target different cell cycle phases—palbociclib delays G1-S transition while EPI-7170 targets S-phase cells—and sequential administration maximizes the proportion of cells arrested in each vulnerable phase.

Quality Hierarchy Considerations

The phase 3 PALOMA-3 trial by Turner et al. carries the greatest weight for establishing clinical efficacy in advanced HR+ breast cancer, with its randomized design, large sample size (n=521), and definitive progression-free survival benefit (HR 0.42). The neoadjuvant trials by Ma et al. and Johnston et al., though smaller phase 2 studies, provide complementary mechanistic insights through serial biopsies that were not feasible in the advanced disease setting. The preclinical studies by Maskey et al., Vijayaraghavan et al., Asghar et al., and Kumarasamy et al. elucidate mechanisms that inform biomarker development but require clinical validation. The convergence of clinical efficacy data from Turner et al. with mechanistic predictions from preclinical models (particularly regarding Rb status) strengthens confidence in these biomarkers despite coming from studies of varying design rigor.

Resolved Model of Palbociclib Action

Integrating these findings yields a coherent model: palbociclib reliably induces G1 arrest in Rb-proficient cells by preventing Rb hyperphosphorylation and maintaining E2F target suppression. The durability of this arrest depends on three factors: (1) CDK2 activity levels, with high CDK2/cyclin E bypassing the G1 block; (2) mTORC1 activity during arrest, with sustained signaling driving irreversible senescence and suppressed signaling allowing reversible arrest; and (3) autophagy induction as a survival response that can be targeted for combination therapy. Clinical benefit requires both achieving cell cycle arrest and maintaining it, explaining why continuous dosing is necessary and why biomarker selection (Rb+/LMWE-/adequate p27) identifies patients with maximal benefit.

References

Reeja S. Maskey, F. Wang, Elyssa Lehman, Yiqun Wang, N. Emmanuel, and 6 more\ (2020).Sustained mTORC1 activity during palbociclib-induced growth arrest triggers senescence in ER+ breast cancer cells. Cell Cycle

Cynthia X. Ma, F. Gao, Jingqin R Luo, D. Northfelt, M. Goetz, and 29 more\ (2017).NeoPalAna: Neoadjuvant Palbociclib, a Cyclin-Dependent Kinase 4/6 Inhibitor, and Anastrozole for Clinical Stage 2 or 3 Estrogen Receptor–Positive Breast Cancer. Clinical Cancer Research

N. Turner, J. Ro, F. André, S. Loi, S. Verma, and 9 more\ (2015).Palbociclib in Hormone-Receptor-Positive Advanced Breast Cancer. New England Journal of Medicine

S. Vijayaraghavan, Cansu Karakas, I. Doostan, Xian Chen, Tuyen N Bui, and 12 more\ (2017).CDK4/6 and autophagy inhibitors synergistically induce senescence in Rb positive cytoplasmic cyclin E negative cancers. Nature Communications

A. DeMichele, A. Clark, K. Tan, D. Heitjan, Kristi Gramlich, and 13 more\ (2014).CDK 4/6 Inhibitor Palbociclib (PD0332991) in Rb+ Advanced Breast Cancer: Phase II Activity, Safety, and Predictive Biomarker Assessment. Clinical Cancer Research

A. Tien, M. Sadar\ (2021).Cyclin-dependent Kinase 4/6 Inhibitor Palbociclib in Combination with Ralaniten Analogs for the Treatment of Androgen Receptor–positive Prostate and Breast Cancers. Molecular Cancer Therapeutics

U. Asghar, A. Barr, R. Cutts, M. Beaney, I. Babina, and 9 more\ (2017).Single-Cell Dynamics Determines Response to CDK4/6 Inhibition in Triple-Negative Breast Cancer. Clinical Cancer Research

V. Kumarasamy, Paris Vail, Ram Nambiar, A. Witkiewicz, E. Knudsen\ (2020).Functional Determinants of Cell Cycle Plasticity and Sensitivity to CDK4/6 Inhibition. Cancer Research

S. Johnston, S. Puhalla, D. Wheatley, A. Ring, P. Barry, and 32 more\ (2019).Randomized Phase II Study Evaluating Palbociclib in Addition to Letrozole as Neoadjuvant Therapy in Estrogen Receptor-Positive Early Breast Cancer: PALLET Trial. Journal of Clinical Oncology

M. Arnedos, M. Bayar, B. Cheaib, Véronique Scott, I. Bouakka, and 14 more\ (2018).Modulation of Rb phosphorylation and antiproliferative response to palbociclib: the preoperative-palbociclib (POP) randomized clinical trial. Annals of Oncology

Download BIBDownload RISDownload TXT

Report

Status

Gather sources

200 sources found

Details

Screen sources

10 sources included

Details

Extract data

80 data points extracted

Details

Generate report

Save PDF

BIBLaTeX, ZoteroRISZotero, MendeleyTXTAPA BibliographyPDFPDFDOCXMicrosoft Word

Chat

Got some follow-up questions?

Sign up or sign in to chat with this report.

Back

CDK4/6 and autophagy inhibitors synergistically induce senescence in Rb positive cytoplasmic cyclin E negative cancers

S. Vijayaraghavan, Cansu Karakas, I. Doostan, Xian Chen, Tuyen N Bui, M. Yi, A. Raghavendra, Yang Zhao, S. Bashour, N. Ibrahim, M. Karuturi, Jing Wang, J. Winkler, R. Amaravadi, K. Hunt, D. Tripathy, K. Keyomarsi

Nature Communications·

2017·

243 citations

SourceDOI

Plain textPDF

Searching for PDF

Unable to find PDF from source

Back

Study Design

- Study type: Preclinical (in vitro and in vivo) and clinical (patient cohort) - Setting: Cell lines (MCF7, T47D, ZR75-1, MCF10A), xenografts (MCF7-T), patient population (109 patients with advanced ER+ breast cancer) - Cancer type and subtype: Breast cancer (ER+), other solid tumors - Sample size: 109 patients in clinical cohort; multiple cell lines and xenograft models in preclinical studies

CDK4/6 Inhibitor Details

- Specific CDK4/6 inhibitor used: Palbociclib - Dose and schedule: In vitro: 1 mM or less, treated for 6 days, recovered for 4 days; In vivo: 25 mg kg^-1 per day for 7 days - Duration of treatment: In vitro: 6 days treatment + 4 days recovery; In vivo: 7 days - Combination therapies: With autophagy inhibitor HCQ - Administration route and timing: Oral gavage

Cell Cycle Arrest Measurements

- G1 arrest: Palbociclib induces G1 arrest and cellular senescence in vitro and inhibits growth of tumour xenografts in vivo. - Cell cycle phase distribution: Cells were analyzed for cell cycle phase distribution using propidium iodide staining and flow cytometry. - BrdU levels: BrdU analysis showed a decrease in BrdU-positive cells, indicating reduced proliferation. - Proliferation markers: BrdU was used to assess proliferation. - Cell viability and growth inhibition metrics: Palbociclib treatment resulted in reversible G1 arrest and growth inhibition at lower doses and irreversible inhibition at higher doses. - Time points measured: Measurements were taken after treatment and recovery phases.

Mechanism Analysis

- Rb pathway activity: Palbociclib induces G1 arrest and decreases pRb and total Rb levels. - Cyclin D1/CDK4/CDK6 expression and activity: CDK4/CDK6/cyclin D pathway alterations are present in 35% of ER+ breast tumors. - Cell cycle checkpoint proteins: Not explicitly mentioned. - Downstream signaling pathways: Autophagy is induced as a stress response; no mention of mTOR or DNA damage response. - Molecular targets and biomarkers: Rb and LMWE are biomarkers for response to palbociclib.

Clinical Outcomes

- Response rates: Not mentioned - Clinical benefit rate: Not mentioned - Progression-free survival: Palbociclib doubles PFS compared to letrozole or fulvestrant alone. Patients with Rb-positive and LMWE-negative tumors have the longest PFS times (median 436.5 months overall and with letrozole; 10.7 months with fulvestrant). - Pathological response: Not mentioned - Time to progression: Not mentioned - Overall survival: Not mentioned - Response duration and maintenance of cell cycle arrest: Not mentioned

Senescence and Cell Fate

- Senescence markers and phenotype: Increased SA-β-gal activity and cellular complexity. - Reversibility vs irreversibility of arrest: Low doses of palbociclib result in reversible G1 arrest, while higher doses lead to irreversible inhibition. - Apoptosis markers: No apoptosis is induced by palbociclib. - Autophagy activation: Palbociclib induces autophagy as a stress response. - Long-term cell fate outcomes: Sustained growth inhibition and irreversible senescence without apoptosis when combined with autophagy inhibitors.

Resistance and Predictive Factors

- Predictive biomarkers: Rb status, cyclin E levels (LMWE) - Resistance mechanisms: Ablation of Rb, overexpression of LMWE - Factors associated with treatment failure: Adverse events, lack of reliable biomarkers - Molecular subtypes: ER-positive breast cancer - Combination strategies: CDK4/6 inhibition with autophagy inhibition (e.g., HCQ)

Toxicity Profile

- Grade 3/4 adverse events: Neutropenia (56%), leukopenia (25.2%) - Dose modifications, interruptions, and discontinuations: Higher doses (75 or 150 mg kg À 1) led to significant body weight loss, suggesting potential toxicity. - Cytopenias: Neutropenia (56%), leukopenia (25.2%) - Non-hematologic toxicities: Not specifically mentioned - Management strategies for adverse events: Not detailed; emphasis on optimizing doses to minimize toxicity

Deregulation of the cell cycle machinery is a hallmark of cancer. While CDK4/6 inhibitors are FDA approved (palbociclib) for treating advanced estrogen receptor-positive breast cancer, two major clinical challenges remain: (i) adverse events leading to therapy discontinuation and (ii) lack of reliable biomarkers. Here we report that breast cancer cells activate autophagy in response to palbociclib, and that the combination of autophagy and CDK4/6 inhibitors induces irreversible growth inhibition and senescence in vitro, and diminishes growth of cell line and patient-derived xenograft tumours in vivo. Furthermore, intact G1/S transition (Rb-positive and low-molecular-weight isoform of cyclin E (cytoplasmic)-negative) is a reliable prognostic biomarker in ER positive breast cancer patients, and predictive of preclinical sensitivity to this drug combination. Inhibition of CDK4/6 and autophagy is also synergistic in other solid cancers with an intact G1/S checkpoint, providing a novel and promising biomarker-driven combination therapeutic strategy to treat breast and other solid tumours.

D

eregulation of the cell cycle checkpoint proteins, such as cyclin-dependent kinases CDK4 and CDK6, is a key hallmark of cancer, resulting in uncontrolled cellular proliferation and tumorigenesis. Selective CDK4/6 inhibitors, palbociclib, ribociclib and abemaciclib, have shown promising preclinical and clinical activities in numerous solid tumours 1 . In particular, palbociclib (PD-0332991 or Ibrance) has shown clear benefits in Phase II (PALOMA-1) and III (PALOMA-2 and -3) clinical trials in advanced estrogen receptor-positive (ER þ ) breast cancers, doubling the progression-free survival (PFS) compared to letrozole or fulvestrant alone [2][3][4] . Palbociclib was recently approved by the U.S. Food and Drug Administration for both these indications and is currently being evaluated clinically in other solid tumours 5 . Moreover, numerous Phase III clinical trials are currently underway with ribociclib (LEE011) and abemaciclib (LY2835219) in breast and other solid tumours 5 .

Despite these promising clinical advances with CDK4/6 inhibitors, there are three major limitations for this treatment: (i) adverse events (Grade 3&4 neutropenia (56%) and leukopenia (25.2%)), which lead to interruption and/or discontinuation of treatment, possibly attenuating therapeutic benefit 2 , (ii) B16% (ref. 3) of the ER þ cancer patients do not respond to palbociclib or exhibit progression within 24 weeks, and half the patients develop clinical resistance with progression within 25 months, resulting in no overall survival benefit and (iii) lack of reliable biomarkers that can be used as prognostic indicators for advanced ER þ breast cancer. Clinical investigation of prognostic biomarkers to-date, have failed to show a correlation to PFS in patients treated with palbociclib 2,6 .

Mechanistically, palbociclib induces G1 arrest and cellular senescence in vitro and inhibits growth of tumour xenografts in vivo 7,8 . Palbociclib also induces autophagy in fibroblasts and leukemic cells 9,10 . Autophagy is a stress tolerance mechanism in cancer that recycles cellular constituents by engulfing them into a double-membrane vesicle called autophagosome, which eventually fuses with lysosomes to facilitate degradation of the cellular constituents and generation of energy for survival 11 . Autophagy mediates resistance to numerous cancer-targeting agents 12,13 , making it a promising co-target and leading to the recent development of numerous clinical trials with autophagy inhibitors, chloroquine (CQ) and hydroxychloroquine (HCQ) 14 .

In this study we address the aforementioned limitations of palbociclib therapy by interrogating the mechanisms by which the drug's action can be enhanced in breast cancer. Our studies suggest that the selectivity and efficacy of palbociclib is improved through a biomarker-driven combination treatment approach that targets CDK4/6 and autophagy in breast and other solid tumours.

Results

CDK4/6 inhibition induces ROS-mediated senescence and autophagy. Analysis of The Cancer Genome Atlas's (TCGA) ER þ breast tumour cohort revealed alterations in the CDK4/CDK6/cyclin D pathway in about 35% of the patients, making them an ideal population for targeting CDK4 and CDK6 (Supplementary Fig. 1a ). To interrogate the biological effect of downregulating these proteins, we knocked down CDK4 and/or CDK6 in two ER þ breast cancer cell lines, MCF7 and T47D via shRNA and siRNA (Supplementary Fig. 1b,c and f ). Results revealed significant growth inhibition only with combined knockdown of CDK4 and CDK6 (Supplementary Fig. 1d,e and g ). Next, we examined the growth-inhibitory potential of the CDK4/6 inhibitor palbociclib in ER þ cell lines (MCF7, T47D, ZR75-1) and in an immortalized non-transformed human mammary epithelial cell line (MCF10A). While palbociclib inhibited growth of ER þ cells in a time-and dose-dependent manner, MCF10A was more resistant to this drug (Fig. 1a, b and Supplementary Fig. 2a-c ). To examine their ability to recover from palbociclib-mediated growth inhibition, cells were treated for 6 days and cultured in the absence of drug for 4 days (Supplementary Fig. 2d ). Treatment with palbociclib at doses of 1 mM or less resulted in reversible G1 arrest and growth inhibition, while only higher doses (42.5 mM) resulted in irreversible inhibition of growth and cell cycle progression in the ER þ cells but not in MCF10A (Fig. 1c and Supplementary Fig. 2e-h ). Palbociclib did not trigger apoptosis (Supplementary Fig. 3a-d ), but significantly increased senescence-associated -galactosidase (SA-gal) activity and cellular complexity in a dose-dependent manner, both characteristics of cells undergoing senescence (Fig. 1d and Supplementary Fig. 3e, f ). Further, western blot analysis showed a dose-dependent decrease in known palbociclib effectors, pRb, FOXM1 (ref. 8) and total Rb levels, a consequence of G1 arrest, as previously reported in breast cancer 15,16 (Supplementary Fig. 3g ). To confirm the specificity of palbociclib-mediated growth inhibition and examine potential off-target effects, MCF7 and T47D cells with shRNA-or siRNA-mediated CDK4 and/or CDK6 knockdown were treated with palbociclib. Combined knockdown of CDK4 and CDK6 was able to recapitulate the drug effect at 1 mM but not 5 mM, indicating that the irreversible grown inhibition and high levels of senescence observed at 5 mM could be due to off-target effects at higher palbociclib concentrations (Fig. 1e and Supplementary Fig. 3h ).

A stress response signal mediated by cell cycle arrest is autophagy, a catabolic process that can degrade ROS and mediate reversal of G1 arrest and senescence 12,14,17 . Hence, it was hypothesized that CDK4/6 inhibition-mediated cell cycle arrest (on-target effect at 1 mM palbociclib) triggers autophagy as a stress response, which prevents the induction of senescence at these doses. To test this hypothesis, we assessed autophagy by: (i) monodansylcadavarine (MDC) staining 18 , (ii) transmission electron microscopy (TEM) 19 , (iii) GFP-LC3 puncta, (iv) western blot analysis of LC3B-II (autophagosomal surface protein 20 ) and p62 (SQSTM1, an autophagic substrate 21 ) and (v) reverse-phase protein array (RPPA) analysis of key autophagy proteins. siRNA against CDK4/6 or palbociclib treatment (low doses) of MCF7 and T47D cells significantly increased levels of MDC, LC3B-II and other key autophagy proteins (Atg-7, Beclin-1, BNIP3), while decreasing BCl2 (a known inhibitor of autophagy 22 ) and p62 at low doses (Fig. 1f ,g and Supplementary Fig. 4a-d ). Further, TEM and GFP-LC3 assays showed significant accumulation of double-membrane electron-dense autophagosomes and GFP-LC3 puncta, respectively, in Palbociclib (1 mM)-treated cells (Fig. 1h, j and Supplementary Figs 4e, 5d ). We next examined the presence of an intact autophagic flux following palbociclib treatment by: (i) flux ratio (LC3B-II to LC3B-I and LC3B-II to p62), (ii) treatment with lysosomal block CQ 23 , (iii) GFP-LC3 puncta with CQ treatment and (iv) RFP-GFP-LC3 dual-reporter assay 24 . Treatment with low-dose (1 mM) palbociclib, exhibited higher flux ratios (Supplementary Fig. 5a ), elevated MDC and LC3B-II levels with CQ treatment (Fig. 1i and Supplementary Fig. 5b, c ), significantly higher GFP-LC3 puncta with CQ treatment (Fig. 1j and Supplementary Fig. 5d ) and significant increase in RFP þ GFP þ puncta (autophagosomes) and RFP þ puncta (autophagolysosomes) (Fig. 1k and Supplementary Fig. 5e ), all demonstrative of an intact autophagic flux.

Another stress response signal associated with cell cycle arrest, autophagy and senescence is oxidative stress induced by reactive oxygen species (ROS), which also mediates senescence 25,26 . This led us to hypothesize that CDK4/6 inhibition mediated by on-target effects of palbociclib (1 mM) may induce ROS, which in turn triggers autophagy. Consistent with this, siRNA knockdown of CDK4/6 or palbociclib treatment increased cellular ROS levels in a dose-dependent manner in ER þ cells but not in MCF10A as measured by CellROX assay and RPPA (Fig. 1l ,m and Supplementary Fig. 6a-c ). To examine the dependence of autophagy on ROS, we ablated ROS in palbociclib-treated cells by treating with N-acetyl-L-cysteine (NAC) or trolox 27,28 , and observed significant decreases in MDC and LC3B-II expression (Fig. 1n ,o and Supplementary Fig. 6d ), suggesting that palbociclib-induced autophagy is dependent on ROS production. Collectively, these results demonstrate that palbociclib induces ROS, leading to intact autophagy and reversible G1 arrest and growth inhibition at low concentrations (r1 mM; on-target effect) (Supplementary Fig 6e ).

Autophagy inhibition sensitizes breast cancer to CDK4/6 inhibition. To test our hypothesis that it is autophagy that protects ER þ breast cancer cells from palbociclib-induced senescence at low doses, we first downregulated two crucial autophagy genes, Beclin-1 and Atg-5 (refs 29,30) (Fig. 2a and Supplementary Fig. 7a ). This had no effect on cell viability by itself, but significantly increased the sensitivity of MCF7 and T47D cells to palbociclib (Fig. 2b and Supplementary Fig. 7b, c ). Specifically, Beclin-1 or Atg-5 knockdown induced irreversible growth inhibition, irreversible G1 arrest and elevated senescence, even in cells treated with low concentrations of palbociclib (Fig. 2c and Supplementary Fig. 7d-f ), suggesting that ablation of autophagy significantly augments the drug's ability to induce senescence.

We next interrogated whether pharmacological inhibition of autophagy would synergize with palbociclib to induce senescence. Treatment with the autophagy inhibitor HCQ, in combination with low-dose palbociclib induced long-term sustained growth inhibition, irreversible G1 arrest, significant increase in ROS levels and cellular senescence, without inducing apoptosis, when compared to palbociclib alone (Fig. 2d-g and Supplementary Fig. 8a-g ). This suggests that autophagy degrades ROS, preventing the induction of senescence at low concentrations (Supplementary Fig. 8j ). Moreover, in MCF7 and T47D cells with CDK4 and CDK6 knocked down via siRNA and shRNA, HCQ treatment significantly increased growth inhibition compared to CDK4/6 knockdown alone, validating the synergy between CDK4/6 and autophagy inhibition (Fig. 2h and Supplementary Fig. 8h, i ).

Next, we examined whether the other clinically available CDK4/6 inhibitors, abemaciclib and ribociclib (Supplementary Fig. 9a, b ) could also synergize with autophagy inhibition. Results revealed that combined treatment with ribociclib or abemaciclib and HCQ yielded significant sustained growth inhibition (Fig. 2i ,k and Supplementary Fig. 9c, d ). Unlike palbociclib, co-treatment with abemaciclib and HCQ also induced significant apoptosis when compared to abemaciclib alone (Fig. 2j and Supplementary Fig. 9e ). Finally, four additional autophagy-inhibiting drugs, Lys05 (ref. 31), CQ, bafilomycin A1 (ref. 32) and spautin-1 (ref. 33) (Supplementary Fig. 9f ), all synergized with low-dose palbociclib, resulting in irreversible growth inhibition (Fig. 2l -n and Supplementary Fig. 9g-k ).

Taken together, these results demonstrate that inhibition of autophagy via genetic ablation or pharmacological agents induces a synergistic response in combination with a CDK4/6 inhibitor, resulting in sustained growth inhibition and senescence (or apoptosis in the case of abemaciclib).

Palbociclib synergizes with autophagy inhibition in vivo.

To interrogate whether palbociclib induces autophagy in vivo, we treated mice with MCF7-T cell line (MCF7 cells passaged through mice and verified to behave similarly to MCF7 cells; Supplementary Fig. 10a-f ) orthotopic xenografts with increasing concentrations (25-150 mg kg À 1 per day) of palbociclib for 7 days (Supplementary Fig. 10g ). Palbociclib treatment significantly decreased tumour volume (Fig. 3a ,b and Supplementary Fig. 10h ), while increasing ROS (measured by 8-hydroxydeoxyguanosine 34 and 4-hydroxynonenal 35 ) and SA-gal activity, and lowering expression of bromo-2 0 -deoxyuridine (BrdU), pRb and FOXM1 in a dose-dependent manner (Fig. 3c ,e and Supplementary Fig. 10i ). Consistently, RPPA analysis of xenograft tumours showed dose-dependent downregulation of cell cycle proteins and upregulation of senescence proteins (Supplementary Fig. 10l-n ). While treatment with palbociclib at 25 and 50 mg kg À 1 was well tolerated, mice treated with 75 or 150 mg kg À 1 palbociclib exhibited significant loss in body weight, suggesting potential toxicity at these concentrations and emphasizing the need to optimize the utility of lower doses (Supplementary Fig. 10j, k ). Notably, western blot, RPPA and TEM analysis of the tumours treated with the lowest dose (25 mg kg À 1 per day) of palbociclib showed higher levels of LC3B-II, Atg-7 and lower levels of p62, as well as the presence of double-membrane electron-dense autophagosomes (Fig. 3c, d and Supplementary Fig. 10l,o-q ), demonstrating the induction of intact autophagy.

We next interrogated the synergy between autophagy inhibitor and low-dose palbociclib by treating xenograft tumour-bearing mice with (i) vehicle, (ii) HCQ (60 mg kg À 1 per day), (iii) low-dose palbociclib (25 mg kg À 1 per day) or (iv) palbociclib þ HCQ for 21 days (treatment phase) with an additional recovery period of 21 days (recovery phase; Supplementary Fig. 11a ). The mice treated with the combination of palbociclib þ HCQ had significantly smaller tumour volumes, which remained small even after treatment was stopped, while the tumour volumes in the other arms increased during both treatment and recovery phases (Fig. 3f-h and Supplementary Fig. 11b-d ). This suggests that co-treatment with HCQ enabled the induction of sustained tumour growth inhibition even at a palbociclib dose that is one-fifth of that (150 mg kg À 1 ) used in most studies [36][37][38] . Moreover, palbociclib þ HCQ in vivo treatment resulted in significantly higher ROS (8-OHdG, 4-HNE) and SA-gal activity, and decreased BrdU and pRb expression at the end of both treatment and recovery phases, while palbociclib alone exhibited changes only at the end of the treatment phase (Fig. 3i ,k and Supplementary Fig. 11f, g ). RPPA analysis of the tumours demonstrated a significant decrease in cell cycle and an increase in senescence proteins in the tumours of mice that received the combination compared to those that received no treatment or palbociclib only (Fig. 3j and Supplementary Fig. 11e ). Further, the drug combination increased LC3B-II levels with no decrease in p62, compared to palbociclib alone, confirming the inhibition of autophagic flux by HCQ (Fig. 3i ). Finally, mice that received palbociclib þ HCQ showed no significant changes in body weight or blood counts, suggesting that this combination is well tolerated (Supplementary Fig. 11h-k ).

To further confirm the synergy in vivo, we utilized another autophagy inhibitor, Lys05 (ref. 31) (a more potent inhibitor of autophagy compared to HCQ), which showed no significant toxicity as a single agent (Supplementary Fig. 12a-d ). Tumour-bearing mice were treated with vehicle, 10 mg kg per day Lys05, 25 mg kg À 1 per day palbociclib or the combination of palbociclib and Lys05 for 21 days (treatment phase) with a recovery phase of 14 days. Treatment with the combination of palbociclib þ Lys05 significantly decreased tumour volume during both the treatment and recovery phases, resulting in significantly smaller tumours and prolonged survival compared to vehicle or single-treatment controls (Fig. 3l -n and Supplementary Fig. 12e-g ).

Collectively, these results demonstrate that autophagy inhibition synergizes with low doses of palbociclib to induce irreversible tumour growth inhibition in vivo.

G1/S checkpoint predicts response to palbociclib and HCQ. To identify biomarker(s) of response to palbociclib and the palbociclib þ HCQ combination, we examined the gene expression of cell lines known to be sensitive or resistant to palbociclib 7 (Supplementary Fig. 13a, b ). This produced a distinct signature within the Biocarta_Cell_Cycle Pathway gene set, with a striking correlation between palbociclib sensitivity and expression of Rb and cyclin E (Fig. 4a ), leading us to hypothesize that these proteins may be effective biomarkers to predict response to palbociclib and the combination of CDK4/6 and autophagy inhibitors.

Ablation of Rb in MCF7 and T47D cells (Supplementary Fig. 13c, d ) reduced sensitivity to palbociclib, mediating resistance to the induction of growth inhibition, G1 arrest and senescence (Supplementary Fig. 13e-l ). Further, knockdown of Rb resulted in no significant increase in MDC staining or autophagic vesicles with palbociclib, and no growth-inhibitory effect upon treatment with HCQ (Fig. 4b-d and Supplementary Fig. 13m, n ), demonstrating a role for Rb in palbociclib-induced autophagy. Since Rb loss is not common in ER þ breast cancer, and hence may not be sufficient to predict patient response to palbociclib 39 , we also interrogated the effect of cyclin E. We have previously shown that post-translational modification (cleavage) of full-length cyclin E generates an oncogenic form termed low-molecular-weight isoforms of cyclin E (LMWE) 40 , which is uniquely expressed in tumours, hyperphosphorylates Rb, mediates resistance to letrozole and is a strong prognostic indicator in breast cancer [41][42][43][44] . Over-expression of LMWE, but not empty vector or full-length cyclin E (Supplementary Fig. 14a ), significantly reduced sensitivity to palbociclib, and mediated resistance to palbociclib-induced growth inhibition, G1 arrest, and senescence (Fig. 4e and Supplementary Fig. 14b-h ). Moreover, treatment with HCQ had no impact on the growth of palbociclib-treated LMWE-overexpressing cells (Fig. 4f and Supplementary Fig. 14i ). Finally, knockdown of Rb or induction of LMWE abolished the ability of low-dose (1 mM) palbociclib to induce an increase in cellular ROS levels (Fig. 4g, h and Supplementary Fig. 14j-l ). This suggests that intact Rb and cyclin E are required for palbociclib-induced senescence, autophagy and ROS (Supplementary Figs 13o, 14m ).

Since palbociclib is currently used clinically in combination with letrozole, we next interrogated how this compares to our proposed combination of palbociclib þ HCQ. Treatment of MCF7 aromatase-expressing cells with palbociclib þ HCQ had a significantly higher growth-inhibitory effect compared to palbociclib and letrozole alone or in combination (Fig. 4i, j ). Moreover, treatment with palbociclib þ letrozole also induced autophagy (Supplementary Fig. 15a ), making the triple combination of palbociclib, letrozole and HCQ more effective than palbociclib þ letrozole (Fig. 4i, j ). Further, over-expression of LMWE, but not empty vector-mediated resistance to the combination of palbociclib þ letrozole and the triple drug combination (Fig. 4k ,l and Supplementary Fig. 15b-d ).

Finally, we examined the effect of palbociclib and autophagy inhibition in cells made resistant to aromatase inhibitors, letrozole and anastrazole. Cells that are ten-fold more resistant to letrozole and anastrazole (Fig. 4m ) remained sensitive to palbociclib as a single agent (Fig. 4n and Supplementary Fig. 15e, f ), induced autophagy in response to palbociclib, and were responsive to the combination of palbociclib and autophagy inhibition (Fig. 4o and Supplementary Fig. 15g, h ), highlighting the clinical utility of the drug combination in the aromatase inhibitor resistance setting.

Taken together, these results reveal the importance of an intact G1/S transition for palbociclib response and provide a rationale for utilizing Rb in combination with LMWE as biomarkers of response to the combination of CDK4/6, aromatase and autophagy inhibitors in ER þ breast cancer.

Palbociclib synergizes with autophagy inhibition in solid tumours. We next asked whether autophagy inhibition would be synergistic with palbociclib in ER À ve breast cancer and other solid tumours, and if LMWE and Rb can predict drug response. To this end, we first examined seven triple-negative breast cancer (TNBC) cell lines for their expression of Rb and LMWE (Fig. 5a ). The Rb þ and LMWE À cell lines (HCC38, MDA-MB-231, SUM-159) were significantly more sensitive to palbociclib (Fig. 5b and Supplementary Fig. 16a-c ) than the Rb À or LMWE þ cell lines, mediating dose-dependent sustained growth inhibition, G1 arrest and senescence (Supplementary Fig. 16d-h ). Moreover, palbociclib treatment-induced autophagy with an active flux in Rb þ /LMWE À TNBC cells (Fig. 5c and Supplementary Fig. 17a, b ). Treatment with the autophagy inhibitor HCQ further sensitized Rb þ /LMWE À cells to palbociclib, inducing irreversible growth inhibition and elevated senescence even at low doses (Fig. 5d ,e and Supplementary Fig. 17d ). In contrast, cell lines with a deregulated G1/S checkpoint displayed resistance to palbociclib-mediated senescence and autophagy, and the combination of palbociclib and HCQ (Fig. 5a, b, d and Supplementary Figs 16, 17a-d ).

To examine this synergy between palbociclib and autophagy inhibition in vivo, we treated Rb þ LMWE À TNBC patientderived xenograft (PDX) tumours with vehicle, HCQ, palbociclib (25 mg kg À 1 ) or palbociclib þ HCQ (Supplementary Fig. 17e ).

Treatment with the drug combination significantly decreased tumour growth and prolonged survival compared to vehicle or either drug alone (Fig. 5f -h and Supplementary Fig. 17f-h ).

While palbociclib is currently approved only for ER þ /Her2 À breast cancer, numerous studies have shown that CDK4/6 inhibitors have activity in other cancers 5,37,38 combination of palbociclib and autophagy inhibitor is effective in other solid tumours, we examined several ovarian, pancreatic, lung, colorectal and prostate cancer cell lines with varied expression levels of Rb and cyclin E (Fig. 5i ). Only cell lines with intact Rb and no LMWE (HeyA8, 59M, Panc-1, BxPc-3, Calu-1, H358, Colo-205, SW-620 and PC3) exhibited significant dose-dependent growth inhibition in response to palbociclib (Fig. 5j ,k and Supplementary Fig. 18a, b ). Moreover, treatment with the autophagy inhibitor further sensitized the Rb þ /LMWE À cell lines to palbociclib (1 mM) and induced irreversible growth inhibition (Fig. 5l ,m and Supplementary Fig. 18c, d ).

Correlation between the half-maximal inhibitory concentration (IC 50 ) of palbociclib and expression of Rb and LMWE across all cancer cell lines revealed that cell lines with an intact G1/S transition (Rb þ /LMWE À ) were significantly more sensitive to the CDK4/6 inhibitor and that this sensitivity was further increased by autophagy inhibition (Fig. 5n and Supplementary Fig. 18e ). Taken together, these results demonstrate that the combination of CDK4/6 and autophagy inhibitors can be utilized to effectively treat solid tumours, with Rb and cyclin E as biomarkers of response (Fig. 5o ).

Rb and cytoplasmic cyclin E: prognosticators of palbociclib in breast cancer. TCGA analysis revealed that 85, 81, 89, 86, 93 and 96% of breast, ovarian, pancreatic, lung, colorectal and prostate cancers, respectively, are positive for Rb and have low/no expression of cyclin E (Fig. 6a and Supplementary Fig. 19a ). However, these analyses are based on RNA levels, which do not differentiate between full-length cyclin E (nuclear) and LMWE (cytoplasmic), nor can accurately predict Rb protein expression (Supplementary Fig. 19b ). For these studies we evaluated the subcellular localization of cyclin E and scored the samples either as cytoplasmic negative or positive. LMWE positive refers to cytoplasmic staining of cyclin E. Hence, we examined LMWE and Rb protein expression in tissue microarray samples from a cohort of 879 early stage breast cancer patients from the NCI Cancer Diagnosis Program 41,44 . LMWE and Rb staining of formalin-fixed paraffin-embedded slides from this cohort revealed that 33% of all tumour samples were Rb þ and LMWE À , while 40% of the ER þ subtype were Rb þ /LMWE À (Fig. 6b , Supplementary Fig. 19c ). Further, Kaplan-Meier curves showed that the patient groups with deregulated cyclin E (LMWE þ ) had worse overall prognosis, emphasizing the prognostic ability of LMWE in breast cancer (Supplementary Fig. 19d ). These results suggest that over 35% of all breast cancer patients (regardless of ER status) may benefit from treatment with the palbociclib þ HCQ combination, highlighting the clinical utility of this regimen.

Next, to directly test the utility of Rb and LMWE (cytoplasmic cyclin E) protein as prognostic biomarkers in metastatic breast cancer, we utilized a cohort of 109 patients with advanced ER þ breast cancer who were/are currently being treated with the combination of palbociclib and letrozole or fulvestrant at MD Anderson Cancer Center (Supplementary Fig. 20a ). Archival, pretreatment specimens from the primary tumour and/or local or metastatic recurrence were obtained from each patient and subjected to cyclin E and Rb staining and scoring (Fig. 6f and Supplementary Fig. 20d ). The majority (102/109) of the samples were positive for Rb, while only 49.5% (54/109) were positive for LMWE (Supplementary Fig. 20b ). Analysis of clinical and pathologic variables revealed that the patients with LMWE or Rb loss exhibited higher rates of progression at 6 and 12 months (assessed by Kaplan-Meier methods) with the heavily treated fulvestrant group experiencing greater progression rate at 12 months as expected (Supplementary Tables 1, 2 ). Further, Rb and LMWE varied significantly with progression rates for the palbociclib þ letrozole and palbociclib þ fulvestrant treatment groups as single (Rb: P ¼ 0.02, LMWE: P ¼ 0.01 and Rb: P ¼ 0.03, LMWE: P ¼ 0.009, respectively) and combined variables (P ¼ 0.006 and P ¼ 0.009, respectively) (Supplementary Tables 1, 2 ). Univariate Cox proportional hazards model showed that Rb and LMWE (as single and combined variables) were the only factors significantly associated with the progression-free interval with both palbociclib þ letrozole and palbociclib þ fulvetrant, with hazard ratios of 0.2 and 0.09 (Rb), 3.2 and 5.2 (LMWE), and 9.2 and 23.8 (Rb þ LMWE), respectively (Supplementary Table 3 ). Further, 84.2% of Rb þ /LMWE À exhibited stable disease in response to palbociclib þ letrozole and 82.4% with palbociclib þ fulvestrant, compared to 61.4 and 58.3% in the other patient groups, respectively (Fig. 6c and Supplementary Fig. 20c ). Kaplan-Meier PFS plots for the 109-patient cohort separated by letrozole (n ¼ 78) or fulvestrant (n ¼ 31) treatment and Rb/LMWE expression revealed that the patients with Rb þ and LMWE À tumour had the longest PFS time (median 436.5 months overall and with letrozole; 10.7 months with fulvestrant), compared to Rb þ / LMWE þ patients (median ¼ 13.4 months; 17 months with letrozole and 4.7 months with fulvestrant) and Rb À /LMWE þ patients who had the shortest PFS time (median ¼ 4.2 months; 3.5 months with letrozole and 4.2 months with fulvestrant) (Fig. 6d, e ). The concordance indices (C-indices) for the multivariate model also showed substantial gains when Rb and LMWE were included as single or combined variables, compared to the model without Rb and cyclin E. (Fig. 6g and Supplementary Table 4 ). Collectively, these results provide evidence that LMWE and Rb are reliable prognostic biomarkers in advanced ER þ breast cancer patients.

Discussion

We report here that cancer cells activate autophagy in response to palbociclib, and that blockade of autophagy significantly improves the efficacy of CDK4/6 inhibition in vitro and in vivo in cancers with an intact G1/S transition (Supplementary Fig. 21 ). While research has shown opposing roles for autophagy-as a pro-survival and a pro-death mechanism-numerous recent studies have highlighted the importance of autophagy as a mediator of drug resistance, specifically in breast cancer 13,45,46 . These studies have shown an association between high expression of autophagy proteins like LC3B and tumour aggressiveness or residual disease post chemotherapy, thus providing strong rationale for using autophagy inhibitors to combat chemoresistance. Further, a recent study has shown that cyclin D1 can upregulate autophagy, which when downregulated, results in an increase in senescence 47 . Thus, results from our study corroborates these findings and provides strong in vitro and in vivo evidence that autophagy inhibitors can be utilized to combat resistance to cell-cycle-targeted therapies, such as CDK4/6 inhibitors.

Although our results show that CDK4/6 inhibition induces ROS, its molecular mechanism remains unclear. Cyclin D1 has been shown to bind to and phosphorylate Nrf1, a regulator of mitochondrial biogenesis and ROS, in a CDK-dependent manner 48 . Hence, it is possible that CDK4/6-cyclin D1 inhibition via palbociclib increases Nrf1 levels, thus increasing ROS activity. The levels of ROS and the subsequent induction of senescence, in turn, might be controlled by c-jun through a previously elucidated mechanism involving the ROS genes, MnSOD and catalase 49 . Alternatively, the induction of ROS might be mediated directly by the Rb targets FOXM1 and BIRC5 (survivin), which decrease in response to palbociclib and have been shown to negatively regulate oxidative stress 50,51 .

A recent study revealed that palbociclib has kinase targets apart from CDK4 and CDK6, namely PIK3CD and PIK3R4 (ref. 52). PIK3R4 (Vps15) is a class III phosphatidylinositol 3-kinase shown to be required for autophagic clearance of proteins. Defects in Vps15 lead to dysfunctional lysosomes 53,54 , similar to those observed in our study in response to high doses of palbociclib (5 mM or 150 mg kg À 1 ). Hence, it is likely that palbociclib inhibits these secondary targets at higher concentrations, accounting for the disruption of autophagic flux observed at these doses, and the observed off-target effects with siRNA against CDK4/6. This hypothesis might also explain why treatment with other CDK4/6 inhibitors failed to elicit such a response, given that these secondary targets are unique to palbociclib 52 .

Identification of reliable biomarkers for palbociclib has proven challenging. While previous in vitro studies showed that Rb, cyclin D and p16 could predict response to palbociclib [55][56][57] , results from Phase II/III trials showed no significant correlation between drug response and the expression of p16 (ref.

2), Ki67, CCND1 amplification 58 , PIK3CA or ESR1 (ref. 59) mutational status, leaving no established prognostic or predictive biomarkers 6 . Here, we use a dual biomarker strategy and show that Rb and LMWE proteins are reliable prognostic biomarkers in advanced ER þ breast cancers. Future clinical trial investigations in early stage breast cancer patients, in the neoadjuvant setting, where patients are treated with either palbociclb þ letrozole or letrozole alone, would reveal the predictive utility of these proteins for palbociclib treatment. Thus, we propose that a simple immunohistochemical assay for Rb and LMWE can be used clinically to identify patients who are likely to have a response to palbociclib and its combination with autophagy inhibitor.

While previous preclinical studies have shown that a small percentage of TNBC cell lines respond to palbociclib 7,60 , this has been overlooked thus far, possibly because of the lack of a definitive biomarker to select TNBC patients. The in vitro and in vivo data presented here provide a rationale to expand the scope of CDK4/6 inhibitor treatment to TNBC patients, given that Rb and LMWE are used to identify the potentially most responsive patients.

We also propose future clinical trials that are biomarker integrated and utilize HCQ (or other autophagy inhibitors such as Lys05) to potentiate the action of palbociclib. Given that HCQ is well tolerated and is currently in clinical trials to reverse hormonal and cytotoxic drug resistance 14,61 , we first propose a Phase II clinical trial (in the neoadjuvant setting) in postmenopausal advanced ER þ /HER2 À breast cancer patients, treating them with the combination of low-dose palbociclib (75 mg kg À 1 per day), HCQ and letrozole using surrogate biomarkers such as Ki67 staining. Eligible patients will be selected on the basis of Rb and LMWE expression assessed from baseline biopsy specimens. Results from this clinical trial would support initiation of a definitive Phase III trial for the triple combination, possibly even extending this treatment to other tumours. This would also support the utility of palbociclib þ HCQ combination to treat RB þ /LMWE À ER þ breast cancer in the neoadjuvant setting. We predict that the combination of continuous low-dose palbociclib and HCQ would be more beneficial than standard-dose palbociclib (21 days on, 7 days off), allowing us to minimize palbociclib-mediated toxicities, avoid proliferative bursts that occur when palbociclib is stopped 62 and prolong overall patient survival-a goal that has not yet been met with currently approved palbociclib treatment combinations.

In summary, this study addresses the current limitations of CDK4/6 inhibitor treatment and proposes a biomarker-driven clinical trial that evaluates the ability of autophagy inhibitors to improve the selectivity and efficacy of CDK4/6 inhibitors in clinic.

Methods

Cell lines. All cell lines used in this study were obtained from ATCC. MCF7, T47D, ZR75-1, MCF7-T, MDA-MB-231, HCC38, HCC1806, MDA-MB-157, HeyA8, 59M and FUOV1 were maintained in minimum essential medium Eagle alpha modification (alpha MEM) supplemented with 10% fetal bovine serum (FBS), 10 mM HEPES, nonessential amino acids, 2 mM L-glutamine, sodium pyruvate and hydrocortisone. 293T and MDA-MB-468 cells were cultured in Dulbecco modified Eagle medium (DMEM) supplemented with 10% FBS. SUM-159 and BT-549 cells were cultured in a 1:1 mixture of alpha MEM and nutrient F-12 Ham medium supplemented with 10% FBS and 5% insulin. Calu-1, H358, A549, PC3 and Du145 cells were cultured in RPMI-1640 medium supplemented with 10% FBS. The pancreatic cancer cell lines (Panc-1 and BxPC-3) were a gift from the laboratory of Dr Anirban Maitra (MD Anderson Cancer Center) and were cultured in DMEM and RPMI 1640, respectively, supplemented with 10% FBS. The colon cancer cell lines (Colo-205, SW-620 and HCT-116) were obtained from the laboratory of Dr Jae-II Park (MD Anderson Cancer Center) and cultured in DMEM supplemented with 10% FBS. The MCF7 aromatase-expressing cells were maintained in DMEM supplemented with 10% tet-free FBS þ 4 mg ml À 1 blasticidin, 1 mg ml À 1 puromycin (InvivoGen, San Diego, CA) and 400 mg ml À 1 G418. Before experimental set-up, they were estrogen deprived for 4 days using phenol red free IMEM supplemented with 10% charcoal-dextran-treated tet-free FBS. During the experiment, the cells were maintained in estrogen deprivation condition in the presence of 25 nM 4-Androstene-3,17-dione. To induce the expression of empty vector and LMWE, the cells were treated with 5 ng ml À 1 Doxycyclin (Sigma) for 24 h before the start of the experiment. The aromatase resistant cell lines were maintained in phenol red DMEM supplemented with 10% charcoal-dextran-treated FBS þ 25 nM 4-Androstene-3,17-dione þ 400 mg ml À 1 G418. All cell lines were maintained free of mycoplasma contamination and authenticated on a regular basis by karyotype and short tandem repeat analysis at MD Anderson's Characterized Cell Line core facility.

Antibodies and drugs. Antibodies against p-Rb (Ser807/811 -#9308, 1:1,000), FOXM1 (#3948, 1:1,000), total Rb (4H1 -#9309, 1:1,000), PARP (#9542, 1:1,000), caspase-7 (#9492, 1:1,000), LC3B (#2775, 1:1,000), p62/SQSTM1 (#5114, 1:1,000), Beclin-1 (#3738, 1:1,000), Atg-5 (#2630, 1:1,000) and p53 (#9282, 1:1,000) were purchased from Cell Signaling Technology (Danvers, MA). Antibodies against CDK4 (C-22-sc260, 1:1,000), CDK6 (C-21-sc177, 1:1,000), cyclin E (HE-12-sc247,1:200) and Mdm2 (sc812, 1:1,000) were purchased from Santa Cruz Biotechnology (Dallas, TX); antibodies against actin (C-4-MAB1501, 1:5,000) and vinculin (V9131, 1:10,000) were purchased from Millipore and Sigma-Aldrich (St Louis, MO), respectively. Palbociclib was obtained from Pfizer, Inc (San Diego, CA). It was diluted in dimethyl sulfoxide (DMSO) for in vitro use and in 0.5% methylcellulose (Sigma-Aldrich, St Louis, MO) for in vivo use and administered to mice via oral gavage. HCQ sulfate was purchased from Sigma-Aldrich and Selleckem. Lys-05 was provided by Dr Ravi Amaravadi (University of Pennsylvania). HCQ and Lys-05 were diluted in sterile water for in vitro use and in sterile phosphate-buffered saline solution (PBS) for in vivo use and administered to mice via intraperitoneal injection. N-acetyl-L-cysteine (NAC), trolox [(±)-6-hydroxy-2,5,7,8tetramethylchromane-2-carboxylic acid], bafilomycin A1, spautin-1 and CQ diphosphate were all purchased from Sigma-Aldrich. Ribociclib was purchased from Selleckem and abemaciclib from MedChem Express; NAC and CQ were diluted in sterile water. Trolox, bafilomycin A1, spautin-1, ribociclib and abemaciclib were diluted in DMSO.

siRNA knockdown. siRNA knockdown of CDK4 and CDK6 were generated as described previously 63 . ON-TARGETplus SMARTpool siRNA for CDK4 (L-003238-00-0005) and CDK6 (L-003240-00-0005) were purchased from Dharmacon. Briefly, cells were plated on 6-well or 12-well plates and tranfected with siRNA targeting CDK4 and/or CDK6 using JetPRIME transfection reagent (Polyplus transfection, New York, NY), as per manufacturer's protocol. Non-coding siRNA pool (siNT) was used as a negative control for comparison. Cells were collected 72 h post transfection for western blot analysis, MDC analysis, CellROX assay and on days 3 and 6 for cell counting.

shRNA knockdown and cyclin E over-expression. Stable shRNA knockdown cells were generated as described previously 63 . All shRNA constructs were purchased from Thermo Scientific (Open Biosystems, Waltham, MA). For single knockdown of CDK4, CDK6, Beclin-1, Atg-5 or Rb, GIPZ lentiviral shRNA was used [CDK4: V3LHS_641689, V3LHS_641690; CDK6: V2LHS_112906, V3LHS_404083; Beclin-1: V3LHS_349514, V3LHS_349513; Atg-5: V2LHS_67978, V3LHS_301131; Rb: V2LHS_130611, V2LHS_130606, V3LHS_340829]. For dual knockdown of CDK4 and CDK6, TRIPZ inducible shRNA for CDK6 was utilized for better selection of knockdowns. To generate lentivirus-expressing shRNA, HEK 293T cells were transfected with pCMVdeltaR8.2, pMD2.G (produced by the Didier Trono laboratory and made available through the Addgene repository) and pGIPZ vector (scrambled shRNA or shRNA against gene of interest) using LipoD293 (SignaGen) or polyethylenimine transfection reagent according to manufacturer's protocol. After 48 h of transfection, the virus-containing medium was collected, filtered through a 0.45-mm filter and added to the cells of interest in the presence of 8 mg ml À 1 of polybrene (Millipore). GFP or RFP expression was confirmed and the lentivirus-infected cells were selected with 1 mg ml À 1 puromycin (InvivoGen, San Diego, CA). MCF7 and T47D cells overexpressing different isoforms of cyclin E (vector, full-length cyclin E (EL) or LMWE) were generated previously 64 and verified in this study by western blot. Dose-response studies. For dose-response studies, 1,000 to 3,000 cells (depending on plating efficiency of each cell line;) were plated in each well of a 96well plate and treated with increasing concentrations (0.01-12 mM) of palbociclib, ribociclib or abemaciclib for 1, 2, 4, 6 or 8 days. The medium was replaced with drug-containing medium every other day. At completion of drug treatment, cultures were continued in drug-free medium (also replaced every other day) until day 12, after which they were stained with 0.5% crystal violet solution. The plates were then solubilized with a solution of 0.1% sodium citrate in 50% ethanol, and absorbance was measured at 570 nm using the Epoch Microplate Spectrophotometer (BioTek Instruments, Inc, Winooski, VT). Values were normalized to those of their no treatment controls and analysed in GraphPad Prism by non-linear regression to obtain the IC 50 values as used in Fig. 1a .

Cellular proliferation assay. For cell proliferation studies, 7,500 to 15,000 cells (depending on plating efficiency of each cell line; data not shown) were plated in each well of six-well plates and treated with the indicated agents for 6 days and cells were allowed to recover for 4 days in the absence of drug to examine reversibility. The medium was replaced every other day during the course of the experiment, either with drug-containing medium (days 2 and 4) or drug-free medium (days 6 and 8). Cells were then collected and counted using the BioRad TC20 Automated Cell Counter on days 0, 3, 6 and 10.

For clonogenic-/colony-formation assay, 5,000 to 10,000 cells (depending on the plating efficiency of each cell line; data not shown) were plated in each well of six-well plates, treated for 6 days and allowed to recover for 6 days in the absence of drug. Cells were then washed with PBS and stained with a 0.5% crystal violet solution in 25% methanol for 10 min. Plates were then scanned to obtain pictures.

Cell cycle and BrdU analyses. Cells were plated (1 Â 10 5 cells per plate) on 10-cm plates and treated with the indicated agents for 6 days; they were allowed to recover without drug(s) for 4 days to examine reversibility. Medium was replaced every other day during the course of the experiment, either with drug-containing medium (days 2 and 4) or drug-free medium (days 6 and 8). For cell cycle analysis, following treatment or treatment (6 days) and recovery (6 days þ 4 days of recovery), cells were subjected to trypsinization, washed with PBS, and fixed with 3.5 ml ice-cold PBS and 1.5 ml of 95% ethanol. Cells were prepared as described previously 63 and incubated in a solution of propidium iodide (1 mg ml À 1 ) and RNAase (1 mg ml À 1 ) at 4 °C overnight. Samples were then analysed on the Beckman Coulter Gallios Flow Cytometer, and data were analysed with the Kaluza software (Beckman Coulter) after excluding doublet cells.

For BrdU analysis, following treatment or recovery, cells were incubated with 10 mM 5-BrdU (Sigma-Aldrich) for 1 h, after which they were subjected to trypsinization, washed with PBS and fixed with 70% ethanol. Cells were then washed with PBS containing 0.5% bovine serum albumin and denatured with a solution of 2 M HCl þ 0.5% Triton X-100 at room temperature for 20 min. Cells were washed again and incubated in 200 ml 0.1 M sodium borate, pH 8.5, to neutralize any residual acid for 2 min at room temperature. Cells were then stained with fluorescein isothiocyanate-conjugated anti-BrdU (BD Biosciences) for 20 min at room temperature, protected from light. Cells were finally washed and incubated in a solution of 0.5 ml propidium iodide (10 mg ml À 1 in PBS) for 30 min at room temperature, protected from light. Samples were analysed on the Beckman Coulter Gallios Flow Cytometer and data were analysed with Kaluza software to obtain the percentages of BrdU-positive cells.

Measurement of senescence. For in vitro studies, senescence was measured by the SA-gal staining kit (Millipore, Billerica, MA) according to the manufacturer's standard protocol. Briefly, cells were plated at a low density of 2,000 to 4,000 cells (depending on the plating efficiency of the cell line; data not shown) in each well of 12-well plates and treated with the indicated agents for 72 h or 6 days. The medium was replaced every other day during the course of the experiment with drug-containing medium (days 2 and 4). Cells were then washed with PBS, fixed and stained with SA-gal solution overnight. The cells were then photographed using the Evos XL Core cell imaging system (ThermoFischer, Waltham, MA) and senescent cells were quantified by counting 100 cells in three different fields for each replicate. A minimum of three technical and three biological replicates were performed for each condition.

For in vivo studies, mouse tumours tissues were snap-frozen in liquid nitrogen, embedded in OCT and cut into thin sections (5 mm). Senescence was measured by the SA-gal detection kit (Biovision, Milpitas, CA) by following the manufacturer's protocol. Briefly, slides were washed with PBS and incubated in staining solution (prepared according to protocol) overnight. Slides were then washed with PBS and fixed in 70% glycerol. Images of the tissue sections were obtained by a Leica DM light microscope using the  20 and  40 optical lenses. Images were acquired with a SPOT Imaging Solutions camera and SPOT Advanced software.

For quantitation of cellular granularity, cells were plated (1 Â 10 5 cells per plate) on 10-cm plates and treated for 6 days; some of the cells were allowed to recover for 4 days without drug. Following treatment, cells were collected, stained with propidium iodide (1 mg ml À 1 ) and analysed on the Beckman Coulter Gallios Flow Cytometer. The data were then analysed with the FlowJo software to obtain a distribution curve of side-scatter versus normalized cell counts. Annexin V and caspase-3 apoptosis assays. Apoptotic cells were measured by using the Alexa Fluor 488 Annexin V Dead Cell Apoptosis kit (Invitrogen, Waltham, MA) according to the manufacturer's protocol. Briefly, cells were plated (1 Â 10 5 cells per plate) on 10-cm plates and treated with the indicated agents for 6 days and allowed to recover for 4 days. Cells were collected at the end of treatment (6 days) or after treatment þ recovery, washed with 1 Â Annexin binding buffer and stained with a solution of Alexa Fluor 488 Annexin V and propidium iodide (100 mg ml À 1 ) as directed by the protocol. Samples were then analysed on the Beckman Coulter Gallios Flow Cytometer, and data were analysed with Kaluza software to obtained the percentages of Annexin V-positive/propidium iodide-negative (early apoptosis) and Annexin V-positive/propidium iodide-positive (late apoptosis) cells.

For the caspase-3 activity assay, following drug treatment, cells were collected, washed with a solution of PBS with 2% FBS and fixed with 100 ml of the fixation solution from the Cytofix/CytoPerm kit (BD Biosciences). Cells were then washed with Perm/Wash buffer and stained with phycoerythrin-conjugated rabbit active caspase-3 antibody (BD Biosciences) for 30 min on ice. Cells were finally washed, resuspended in the PBS with 2% FBS solution and analysed on the Becton Dickinson FACS Calibur Flow Cytometer to obtain the percentage of caspase-3-positive cells.

Cellular ROS measurement. Cellular ROS levels were measured by using the CellROX Deep Red Flow Cytometry assay kit (ThermoFischer Scientific, Waltham, MA) according to the manufacturer's protocol. Briefly, cells were plated (1 Â 10 5 cells per dish) on 10-cm dishes and treated for 6 days. Cells were then collected and stained with 500 nM CellROX reagent for 1 h at 37 °C. Samples were analysed on the Beckman Coulter Gallios Flow Cytometer using the 635-nm laser; data were analysed with the FlowJo software and mean fluorescence intensity (MFI) was obtained.

Monodansylcadavarine measurement. Cells were seeded at a density of 1 Â 10 5 cells on 10-cm plates and treated for 6 days with various concentrations of palbociclib. At the end of drug treatment, cells were incubated with 50 mM MDC (Sigma-Aldrich) at 37 °C for 45 min. Cells were then collected, washed with PBS and suspended in a solution of PBS with 1% FBS. Samples were analysed on the Becton Dickinson FACS LSR II Flow Cytometer using the 355-nm ultraviolet laser. Data were analysed with the FlowJo software, and percentages of MDC-positive cells and MFI were obtained.

Immunofluorescence. For the GFP-LC3 puncta assay, MCF7, MDA-MB-231 and MDA-MB-468 cells stably expressing GFP-LC3 were plated in six-well plates and treated with drug for 48 h. Following treatment, cells were washed with 1 Â PBS, fixed with 4% paraformaldehyde for 10 min and mounted with Vectashield mounting media with DAPI. Cells were then visualized with Zeiss Confocal microscope LSM880 usign the 488 nm laser (GFP) for the presence of GFP-LC3 puncta. The puncta was quantified using ImgeJ.

For RFP-GFP-LC3 dual reporter assay, cells were transfected with ptf-LC3 vector and analysed as described previously 24 . Briefly, ptfLC3 vector was tranfected using Lipofectamine 2000 as per manufacturer's instructions. Cells were then treated with the drug for 48 h, washed with 1 Â PBS, fixed briefly (for 10 min) with 4% paraformaldehyde and mounted with Vectashield moutning media. They were then visualized using with Zeiss Confocal microscope LSM880 using the 488 and 643 nm channels to image the GFP þ ve and RFP þ ve LC3 puncta respectively. The puncta was quantified using ImgeJ to obtained the number of autophagosomes (yellow

Transmission electron microscopy. For cell lines, B5,000 cells were plated in each well of 12-well plates and treated with the drug for 6 days. For xenografts, tumours were harvested and samples were cut into 1-mm 3 pieces. The cell lines and tumour samples were processed similarly and Electron microscopy was performed at the High Resolution electron microscopy facility at MD Anderson Cancer Center as described previously 65 . Briefly, they were fixed with a solution containing 3% glutaraldehyde plus 2% paraformaldehyde in 0.1 M cacodylate buffer, pH 7.3. Samples were then washed in 0.1 M sodium cacodylate buffer and treated with 0.1% Millipore-filtered cacodylate-buffered tannic acid. They were fixed with 1% buffered osmium tetroxide for 30 min and stained en bloc with 1% Millipore-filtered uranyl acetate. The samples were dehydrated in increasing concentrations of ethanol, filtrated and embedded in LX-112 medium. They were then polymerized in a 60 °C oven for approximately 3 days. Ultrathin sections were cut in a Leica Ultracut microtome (Leica, Deerfield, IL), stained with uranyl acetate and lead citrate in a Leica EM Stainer, and examined in a JEM 1010 TEM (JEOL, USA, Inc., Peabody, MA) at an accelerating voltage of 80 kV. Digital images were obtained at magnifications of  5,000,  25,000 and  50,000 using the AMT Imaging System (Advanced Microscopy Techniques Corp, Danvers, MA).

In vivo xenograft studies. For all xenograft experiments, estrogen pellets (0.72 mg 17-beta estradiol pellet, 90-day release, Innovative Research of America, Sarasota, FL) were implanted subcutaneously into 4-to 6-week-old female nude mice. MCF7-T cells (5 Â 10 6 in a 1:1 ratio with matrigel (BD Biosciences)) were injected into the fifth and tenth inguinal mammary fat pads bilaterally. For the dose-determining experiment, once the tumours reached an average volume of 200 mm 3 , the tumour-bearing mice were randomized into five groups (n ¼ 3 per group) and treated with vehicle (0.5% methylcellulose) or 25 mg kg À 1 , 50 mg kg À 1 , 75 mg kg À 1 or 150 mg kg À 1 palbociclib. Palbociclib was administered daily via oral gavage for 7 consecutive days. For the combination treatment experiment, once the tumours reached an average volume of 250 mm 3 , the tumour-bearing mice were randomized into four groups (n ¼ 9 per group) and treated with vehicle (0.5% methylcellulose and PBS), HCQ (60 mg kg À 1 ), palbociclib (25 mg kg À 1 ) or a combination of palbociclib (25 mg kg À 1 ) and HCQ (60 mg kg À 1 ). Drugs were administered daily via oral gavage (palbociclib) or intraperitoneally (HCQ) for 21 days.

For Lys-05 toxicity experiment, non-tumour-bearing female nude mice were treated with varying concentrations of Lys-05 (1 mg kg À 1 , 5 mg kg À 1 , 10 mg kg À 1 and 20 mg kg À 1 ) everyday for 21 days via I.P. For the Lys-05 combination treatment experiment, once the orthotopic xenograft tumours reached an average volume of 250 mm 3 , the tumour-bearing mice were randomized into four groups (n ¼ 5 per group) and treated with vehicle (0.5% methylcellulose and PBS), Lys-05 (10 mg kg À 1 ), palbociclib (25 mg kg À 1 ) or a combination of palbociclib (25 mg kg À 1 ) and Lys-05 (10 mg kg À 1 ). Drugs were administered daily via oral gavage (palbociclib) or intraperitoneally (Lys-05) for 21 days.

For all xenograft studies, tumour volumes ((L Â W 2 )/2)) were measured twice per week with calipers and mouse weight was measured every day. At the end of the treatment and recovery periods, mice were euthanized and the tumours were collected for further analysis. At the time the mice were euthanized, blood (0.2 ml) was collected by cardiac puncture through the left ventricle. Blood samples were subjected to complete blood count analysis (white blood cells, platelets and red blood cells) by the Siemens Adiva 120 Hematology System (Erlangen, Germany). Nude mice for all experiments were obtained from the Department of Experimental Radiation Oncology at The University of Texas MD Anderson Cancer Center, and mice received care in accordance with the Animal Welfare Act and the institutional guidelines of MD Anderson Cancer Center. The protocol for this study was approved by the Institutional Animal Care and Use Committee (IACUC) at The University of Texas MD Anderson Cancer Center (Houston, TX).

Patient-derived xenograft studies. Breast tumour samples were obtained during routine surgery after informed consent was obtained under protocols approved by the MD Anderson Institutional Review Board. PDX models were developed as previously described 66 . Briefly, fresh primary tumours were collected using sterile technique, and B3-mm 3 fragments of the tissue were transplanted to the fat pad of the fourth pair of mammary glands (both sides) in immunodeficient (severe combined immunodeficient) mice within 1 h of surgical resection. When the primary tumour outgrowths reached 10 mm in diameter, 3-mm 3 fragments of the outgrowths were explanted to new hosts (n ¼ 3 per tumour) as secondary passage. As the tumour tissues can stably grow after two passages with our protocol, we considered the PDX line to have been successfully established at that point. The histology of the patient tumours of origin and the corresponding PDX lines were compared by hematoxylin and eosin staining and had very similar histology (data not shown). For the drug treatment experiment, 3-mm 3 fragments of the PDX line were transplanted into the fat pat of the fourth mammary gland of female nude mice. Once the tumours reached an average volume of 200 mm 3 , mice were randomized into four groups (n ¼ 4 per group) and treated with vehicle (0.5% methylcellulose and PBS), HCQ (60 mg kg À 1 ), palbociclib (25 mg kg À 1 ) or combination of palbociclib (25 mg kg À 1 ) and HCQ (60 mg kg À 1 ). Drugs were administered daily via oral gavage (palbociclib) or intraperitoneally (HCQ) for 21 days. tumour volumes ((L Â W 2 )/2)) were measured twice per week with calipers, and mouse weight was measured every day.

At the end of the treatment period, mice were killed and their tumours were collected for further analysis. All mice were obtained from the Department of Experimental Radiation Oncology at The University of Texas MD Anderson Cancer Center and received care in accordance with the Animal Welfare Act and the institutional guidelines of MD Anderson Cancer Center. The protocol for this study was approved by the IACUC at MD Anderson Cancer Center.

Immunohistochemistry analysis of mouse tumour tissues. For BrdU assessment, mice were administered BrdU solution (5 mg kg À 1 intraperitoneally) 2 h before killing. After death, the tumour was resected and snap-frozen in liquid nitrogen, embedded in OCT, cut and utilized for measurement of senescence by SA-gal staining. The remaining tissue was fixed in formalin and used for assessment of histology and of proliferation by BrdU as described previously 63 . Briefly, 5-mm sections from the formalin-fixed, paraffin-embedded tumour tissues were stained with standard hematoxylin and eosin. Other sections of the tumour blocks were subjected to immunohistochemical (IHC) staining. After paraffin removal, tumour sections were heated in a water bath for 20 min in 10 mM sodium citrate buffer (pH 6.0) at 90 °C to retrieve nuclear antigens. Endogenous peroxidase activity was quenched with a 3% hydrogen peroxide solution. Sections were blocked with 1.5% normal goat serum and incubated overnight at 4 °C with rat monoclonal antibody to BrdU (clone BU1/75 [ICR1]; GeneTex Inc, San Antonio, TX) diluted at 1:500. Slides were developed using the VECTASTAIN Elite ABC kit (PK4004; Vector Laboratories, Burlingame, CA), followed by staining with DAB substrate (Vector Laboratories) and counterstaining with hematoxylin (DAKO), and then were mounted. Staining was evaluated with a Leica DM light microscope using the  40 optical lenses. Images were acquired on a SPOT Imaging Solutions camera with SPOT Advanced software. BrdU was quantified as percentage of BrdU-positive cells, which was calculated as percentage of BrdU-positive nuclei from a total of 300 tumour cells from three fields of view.

For 4HNE and Anti-8OHdG immunostaining; after paraffin removal, heat-induced antigen retrieval was performed for 10 min in 10 mM sodium citrate buffer (pH 6.0) at 98 °C. Endogenous peroxidase activity was blocked with 3% hydrogen peroxidase for 15 min. Slides were then incubated for 1 h with diluted rabbit blocking serum. The sections were incubated for overnight at 4 °C with Anti-8OHdG mouse monoclonal antibody (clone N45.1, Genox, Baltimore, MD, 1:100 dilution) and Anti-4-HNE mouse monoclonal antibody (clone HNEJ-2, Genox, Baltimore, MD, 1:50 dilution). The slides were incubated for 30 min with diluted biotinylated secondary antibody and 30 min with Vectastain Elite ABC kit (Vector Laboratories, United States). For the evaluation of 4HNE and Anti-8OHdG antibody expression, slides were scored separately for percentage and intensity of the cells. Percentage positivity was graded using 0-4 scale, where 0 represented no stained cells, 1 was 1 to 5% stained cells, 2 was 6 to 30% stained cells, 3 was 31 to 70% stained cells and 4 was 71 to 100% stained cells. Staining intensity was scored as follows: 0, no staining; 1, weak positive; 2, intermediate positive; and 3, strong positive. H score calculated by multiplying the percentage of positive cells and intensity of staining.

Western blot analysis. Western blot analysis was performed as described previously 63 . Briefly, cells were seeded at a density of 1 Â 10 5 cells on 10-cm plates and treated for 6 days with the indicated drugs. Following treatment, cells were collected and subjected to lysis with RIPA buffer (150 mM NaCl, 10 mM Tris, pH 7.3, 0.1% sodium dodecyl sulfate (SDS), 1% Triton X-100, 1% deoxycholate and 5 mM ethylene-diaminetetraacetic acid) containing protease inhibitors. For mouse tissues, the tumours were minced into small pieces on dry ice and immersed in RIPA buffer for lysis. Lysates were then subjected to centrifugation at 45,000 r.p.m. for 45 min at 4 °C to obtain the protein lysates in the supernatant. Protein concentration was determined by Bradford Protein Assay dye (Bio-Rad), and 50 mg of protein per sample was resolved by SDS-polyacrylamide gel electrophoresis as described previously. Blots were blocked with Blotto milk for 1 h at room temperature and incubated with primary antibody overnight at 4 °C. They were then incubated with goat anti-rabbit or goat anti-mouse immunoglobulinhorseradish peroxidase conjugates (Pierce, Rockford, IL) at a dilution of 1:5,000 in Blotto for 1 h. Blots were then washed and developed using a Renaissance chemiluminescence system (Perkin Elmer Life Sciences, Inc.) by following the manufacturer's instructions. The developed and scanned blots were then analysed by the ImageJ software to obtain densitometry values.

RPPA and data analysis. The mouse tumour tissues were dissected on dry ice and subjected to lysis in RIPA buffer (1% Triton X-100, 50 mM HEPES, pH 7.4, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 100 mM NaF, 10 mM Na pyrophosphate, 1 mM Na 3 VO 4 , 10% glycerol, and freshly added protease and phosphatase inhibitors; Roche Applied Science, Indianapolis, IN). Tumour lysates were then subjected to centrifugation and protein concentration was determined by using the Bradford reagent. Protein concentration was adjusted to 1.5 mg ml À 1 and mixed with 4 Â SDS Sample Buffer containing 40% glycerol, 8% SDS, 0.25 M Tris-HCL and 2-mercapto-ethanol at pH 6.8. Samples were boiled for 5 min, and the RPPA analysis was performed by the Functional Proteomics core facility at MD Anderson Cancer Center. The slide images were quantified using MicroVigene 4.0 (Vigene-Tech, Carlisle, MA). The spot level raw data was processed with the R package SuperCurve (https://r-forge.r-project.org/ projects/supercurve), which returns the estimated protein concentration (raw concentration) and a quality control score for each slide. The raw concentration data was then normalized by median-centring for each sample across all the proteins, to correct for loading bias.

For RPPA data analysis, one-way (analysis of variance) ANOVA was used to identify proteins that are differentially expressed between treatment groups. To adjust for multiple comparisons, Benjamini-Hochberg procedure was used to estimate false discovery rate (FDR). Tukey HSD tests were used for post hoc pairwise comparisons. To compare the different Palbociclib doses treatments in vivo (Vehicle, 25 mg kg À 1 , 75 mg kg À 1 and 150 mg kg À 1 ), proteins with 20% FDR in ANOVA, Tukey Po0.05 and fold change 4 ±1.2 was identified as significantly differentially expressed between groups. A heat map was then drawn using significant proteins from the ANOVA analysis, and ordered based on KEGG cell cycle, senescence and autophagy/catabolism pathways. Pearson distance metric and Ward's minimum variance was used to cluster the samples in the heat map. For comparing the combination treatments (Vehicle, HCQ, Palbociclib and Palbociclib þ HCQ) in vivo, proteins with 15% FDR in ANOVA, Tukey Po0.05 and fold change 4 ± 1.2 were identified as significantly differentially expressed in pairwise analysis. Pathway score for cell cycle and senescence pathways was calculated using mean expression level (in log 2 scale) of the selected proteins from ANOVA analysis.

Bioinformatics and TCGA analysis. Alterations in CDK4, CDK6, CCND1, Rb1 and CCNE1 were obtained using cBio Portal 67 from the TCGA RNA seq data for breast, ovarian, lung, pancreatic, colon and prostate cancer. For biomarker analysis, gene expression data for the 23 cell lines under study was obtained from Kao et al. 68 , and Gene Set Enrichment Analysis (GSEA) was performed against the BioCarta gene sets 69 to obtain a heat map of the cell cycle genes.

Breast cancer patient samples, clinical data and statistical analysis. The Department of Medical Oncology at MD Anderson maintains a prospective curated database of patients from 1997 onward. This database was searched for all patients with breast cancer who had received palbociclib, and this search was supplemented with a manual search. Key demographic, clinical and pathologic data, cancer treatment details and response to therapy were abstracted into a working database for analysis under an IRB-approved protocol. Outcomes of interest that were recorded included response, stable disease and progression (based on the physician's determination, but not always based on RECIST criteria) and associated durations of disease response and stability. To demonstrate the applicability of our staining procedure, blocks with the largest available tumour specimen was chosen for each patient (obtained after informed consent) and retrieved from the archives of the Department of Pathology at MD Anderson under an IRB-approved protocol; this included specimens from the primary tumour and/or local and/or metastatic recurrences. Our working database includes results of pathology reports, including tumour histology and grade and standard IHC analyses for ER and progesterone receptors for primary breast cancer biopsies and surgical specimens as well as local and distant recurrences that were subject to biopsy or surgical excision.

Patient, tumour and treatment characteristics were evaluated and compared between patients who were or were not free from progression; 6-month and 12-month progression rates were calculated for each of factor using the Kaplan-Meier method, and differences were examined using the log-rank test (Supplementary Tables 1 and 2 ). Univariable and multivariable Cox model analyses were used to determine the influence of patient, tumour and treatment factors of known or potential prognostic value on PFS (Supplementary Tables 3 and 4 ). All factors with Pr0.1 in the univariable analyses were entered into a full model, and the final model was selected by using a backwards elimination procedure. Model performance was quantified using Harrell's concordance index 70 . The discriminative ability of the model was assessed using the C-index for comparative purposes with the literature, as well as the concordance probability estimate due to the high degree of censoring in the data 70 . The C-index can range from perfect concordance (1.0) to random predictions (0.5). Similar to the area under the receiver operating characteristic curve, concordance probability estimate can range from perfect concordance (1.0) to perfect discordance (0.0). In addition, Akaike's information criterion (AIC) was calculated. The AIC takes into account how well the model fits the data as well as the complexity of a model, thereby reducing the risk of overfitting. After comparisons, the final model with the lowest AIC value and the highest C-index was reported. All statistical analyses were performed using R 3.3.2 (http://www.r-project.org/). All P values were two-tailed, and Pr0.05 was considered significant, and adjustments for multiple factors were not made.

Immunohistochemical staining of patient samples. For the 109 breast cancer samples recovered, two serial 5-mm sections were cut and mounted on Superfrost Plus glass slides. The antigen retrieval and washing steps were performed as described for BrdU immunohistochemistry analysis. For cyclin E and Rb IHC, two commercially available primary antibodies were used: rabbit polyclonal antibody to cyclin E (clone C19; Santa Cruz Biotechnology, Santa Cruz, CA) diluted 1:1,000 and Rb mouse monoclonal antibody (Clone 4H1; Cell Signaling Technology, Denvers, MA) diluted 1:100. Slides were developed using the VECTASTAIN Elite ABC kit (PK6101 and PK6102; Vector Laboratories, Burlingame, CA) followed by staining with DAB substrate (Vector) and counterstaining with hematoxylin (DAKO), and then were mounted. Tumour cell blocks known to express high levels of LMWE and Rb were included in each batch as positive controls, and negative controls were prepared by replacing the primary antibody with PBS buffer. Staining was evaluated with a Leica DM light microscope using the  20 and  40 optical lenses. Images were acquired by a SPOT Imaging Solutions camera and SPOT Advanced software. Pathologists at MD Anderson (C.K. and P.O.S.) evaluated and scored Rb and LMWE staining.

Immunohistochemical scoring of Rb and cyclin E. Cyclin E staining was scored by two pathologists blinded to patient outcomes (data not shown). Scores (0 ¼ negative, 1 ¼ weak staining, 2 ¼ moderate staining and 3 ¼ strong staining) were assigned for nuclear and cytoplasmic staining according to percentage of cells stained and intensity of staining, as described previously 41 . Each tumour sample was scored separately for nuclear and cytoplasmic cyclin E expression, and LMWE status was assigned as follows: LMWE negative was defined as no staining or nuclear staining only; LMWE positive was defined as nuclear þ cytoplasmic staining or cytoplasmic staining only.

For Rb staining, the intensity of staining and percentage of positive cells were evaluated separately. Staining intensity was scored as follows: 0, no staining; 1, weak positive (faint yellow staining); 2, intermediate positive; and 3, strong positive (brown staining). The number of positive cells was visually evaluated and stratified as follows: o1%, 0 (negative); 1 to o5% positive cells, 1 (weak); 5-50% positive cells, 2 (moderate); 450% positive cells, 3 (strong). The sum of the staining intensity and percentage of positive cell scores was used to determine the staining index for each section, with a minimum score of 0 and maximum score of 6; scores 41 were defined as Rb positivity. Using this cut-off, we compared Rb-positive to Rb-negative tumours.

IHC analysis for Rb and cyclin E for the NCI TMA tumour samples were performed and scored as described above for the Palbociclib-treated patient samples.

Statistical analysis.

All experiments were performed with a minimum of three technical and three biological replicates, and values reported are the mean of the three biological replicates, unless otherwise indicated. Error bars represent the s.d. from the mean, unless otherwise indicated. Pairwise comparisons were analysed using multiple t-tests (one unpaired t-test per row), with corrections applied (Holm-Sidak method) for multiple comparisons. When comparing data from experiments with multiple groups, a regular one-way ANOVA (no matching) was used with the Benjamini-Hochberg procedure to adjust for multiple comparisons. Tukey HSD tests were used for post hoc analysis. Kaplan-Meier survival analysis was performed by using the log-rank (Mantel-Cox) test. For all tests, differences were considered statistically significant at a P value of 0.05 or less. For all figures, NS: P40.05; *Po0.05; **Po0.01; ***Po0.001; ****Po0.0001. All statistical analyses were performed using the GraphPad Prism software and R.

availability

Data availability. All data are available within the Article and Supplementary Files, or available from the authors upon request.

Author contributions

S.V. and K.K. designed experiments, interpreted data and wrote the manuscript. S.V. conducted experiments, acquired and analysed the data. C.K. performed the immunohistochemical analysis and scoring of mice and human tissues. I.D. developed the aromatase resistant cell lines. X.C. and I.D. helped perform the mouse experiments. T.B. helped perform the western blot analysis. M.Y. performed the statistical analysis for patient samples. A.S.R maintained the patient database. Y.Z. and J.W. performed the statistical and data analysis for RPPA. S.I.B, N.K.I, M.K. and D.T. helped maintained the patient database and contributed to the accrual of patient samples. R.K.A and J.D.W. provided expert guidance and reagents for the Lys05 experiments. K.K.H. and D.T. provided expert guidance and helped write the manuscript. All authors read and provided input on the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/ naturecommunications Competing interests: The authors declare no competing financial interests. Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Acknowledgements

AcknowledgementsWe thank Dr. Maria Koehler, VP, Oncology Strategy, Innovation and Collaboration at Pfizer Pharmaceuticals Oncology for discussion and providing us with palbociclib for this study.We also thank Anirban Maitra (MD Anderson) and Jae-II Park (MD Anderson) for providing cells lines, Robert Bast Jr (MD Anderson) for RFP-GFP-LC3 (ptfLC3) plasmid and Gordon B. Mills for MDA-MB-231 and MDA-MB-468 cells expressing GFP-LC3; Jason P.W. Carey and Natalie A. Jabbour-Leung for their scientific input; Kapil Sirohi for scientific assistance with GFP-LC3 and RFP-GFP-LC3 experiments; Nalini Patel and Wendy Schober for assistance with flow cytometry experiments; Pathologist Sergio Pina Oviedo for scoring the palbociclib-treated patient samples for Rb.This research was performed in partial fulfillment of the requirements for the Ph.D. degree from The University of Texas Graduate School of Biomedical Sciences at Houston; The University of Texas MD Anderson Cancer Center, Houston, Texas 77030.This study was supported by the National Institutes of Health (NIH) through a Cancer Center Support Grant (CA016672) to MD Anderson and NIH grants CA87548 and CA1522218 and Cancer Prevention Institute of Texas (CPRIT) RP170079 grants, to K. Keyomarsi, the Susan G. Komen for the Cure grant KG100521 to K.K. Hunt and the CPRIT RP140106 and CPRIT RP170076 Training Grant Awards to S. Vijayaraghavan.The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Failed to load PDF:

StripeM-Inner

Paper sources

Abstract screening pilot

Abstract screening results

Extraction pilot

Extraction results

Research report

Modify setup