Purpose:

We conducted research on CDK4/6 inhibitors (CDK4/6i) simultaneously in the preclinical and clinical spaces to gain a deeper understanding of how senescence influences tumor growth in humans.

Patients and Methods:

We coordinated a first-in-kind phase II clinical trial of the CDK4/6i abemaciclib for patients with progressive dedifferentiated liposarcoma (DDLS) with cellular studies interrogating the molecular basis of geroconversion.

Results:

Thirty patients with progressing DDLS enrolled and were treated with 200 mg of abemaciclib twice daily. The median progression-free survival was 33 weeks at the time of the data lock, with 23 of 30 progression-free at 12 weeks (76.7%, two-sided 95% CI, 57.7%–90.1%). No new safety signals were identified. Concurrent preclinical work in liposarcoma cell lines identified ANGPTL4 as a necessary late regulator of geroconversion, the pathway from reversible cell-cycle exit to a stably arrested inflammation-provoking senescent cell. Using this insight, we were able to identify patients in which abemaciclib induced tumor cell senescence. Senescence correlated with increased leukocyte infiltration, primarily CD4-positive cells, within a month of therapy. However, those individuals with both senescence and increased TILs were also more likely to acquire resistance later in therapy. These suggest that combining senolytics with abemaciclib in a subset of patients may improve the duration of response.

Conclusions:

Abemaciclib was well tolerated and showed promising activity in DDLS. The discovery of ANGPTL4 as a late regulator of geroconversion helped to define how CDK4/6i-induced cellular senescence modulates the immune tumor microenvironment and contributes to both positive and negative clinical outcomes.

See related commentary by Weiss et al., p. 649

This article is featured in Highlights of This Issue, p. 647

Translational Relevance

We coordinated preclinical work on the molecular basis of geroconversion with a phase II trial of the CDK4/6-inhibitor abemaciclib in dedifferentiated liposarcoma. This highly translational approach identified ANGPTL4 as a late marker of geroconversion and provided the ability to identify senescent cells in on-treatment biopsies. Abemaciclib is a promising treatment for dedifferentiated liposarcoma, which can promote cellular senescence in tumor cells and the expression of an immune provoking senescent-associated secretory paradigm as marked by the recruitment of CD4+ T cells into the tumor. This complex process may be beneficial in the first few months of therapy, but then detrimental for some. This work provides novel insight into the mechanisms of action of CDK4/6 inhibitors in humans and subsequent effects of senescence in the tumor microenvironment, advancing our understanding of how to study, apply, and augment this therapy in cancer care.

Advanced progressive dedifferentiated liposarcoma (DDLS) is a rare and deadly malignancy of adipocytes. These tumors are characterized by copy-number alterations, including dual 12q amplicons involving CDK4 and MDM2, two genes critical for the pathogenesis and propagation of this disease (1). Palbociclib, a CDK4/6 inhibitor, is NCCN compendium listed for DDLS after two phase II trials, evaluating different doses and schedules, each showed a modest but meaningful benefit in progression-free survival (2, 3).

Palbociclib can induce senescence in DDLS cell lines (4–6). In some of these cell lines, RB-dependent drug-induced cell-cycle exit occurs with concomitant downregulation of MDM2 and the accumulation of senescence-specific ATRX foci. This depends on CDH18 sequestering PDLIM7 away from MDM2, allowing MDM2 turnover to occur. Ultimately, ATRX suppresses residual expression from the HRAS locus, and the cell evolves into a stably arrested SASP producing senescent cell (4–6). Both MDM2 loss and pretreatment expression of CDH18 are associated with increased duration of PFS in DDLS patients treated with palbociclib (4, 6). Palbociclib therapy induced senescence (TIS) is p53-independent (6), which is consistent with the finding that combining CDK4/6i with an MDM2i that liberates p53 had little to no additive effect over the CDK4/6i alone (7). How MDM2 loss contributes to geroconversion, the transitional pathway from reversible to irreversible cell-cycle exit, however, remains to be determined.

A number of different mechanisms in various types of cancer have been postulated to account for the anticancer effects of CDK4/6 inhibitors (8–10). These range from cancer cell autonomous to nonautonomous effects of the drug, for example, enforcing a reversible arrest (quiescence) and then, from the quiescent state, an irreversible arrest with the development of a pro-inflammatory transcriptional and secretory program (senescence), versus drug-induced enhancement of antigen presentation in tumor cells and direct modulation of immune cells that make up the tumor microenvironment. These cellular effects may not be mutually exclusive; for example, senescent cells can express and secrete a number of extracellular matrix-remodeling enzymes and pro-inflammatory cytokines, called the senescence-associated secretory phenotype (SASP; refs. 11–13), which may drive the accumulation of tumor-infiltrating leukocytes (TIL; ref. 14). In addition, without promoting senescence, CDK4/6i might enhance antigen presentation in tumor cells and directly suppress the proliferation of Tregs, promoting cytotoxic T-cell-mediated clearance in the tumor microenvironment (15, 16). Improving upon CDK4/6i therapy may depend on understanding how these drugs act in each disease.

The contextual heterogeneity of the transcriptional and secretory programs of each type of senescent cell induced by different signals (17) might explain why the immunologic response to such cells is context specific. In mice, macrophages, natural killer (NK) cells, invariant NK cells, neutrophils, CD4+ cells, and CD8+ cytotoxic T cells have all been documented to be recruited to senescent cells to eliminate them (18–26).

The heterogeneous SASP also contains some cytokines that can reinforce the stability of the senescence-associated growth arrest, both in cell autonomous and in paracrine manners (11, 27, 28), and others that can alter the tumor microenvironment and create a more permissive environment for tumor growth (29, 30). This has led investigators to coin the phrase that senescence can be a “double-edged sword.” Whether early growth-inhibitory effects of TIS are lost as senescence becomes chronic is not known (31). TIS studies in mice, which largely defined what happens in an acute state, are difficult to fully do in DDLS as there are no immunologically competent mouse models. Conversely, to our knowledge, no human studies in any malignancy have been performed to establish how senescence and the immune response might co-evolve in people. In part, this is due to an inability to accurately define a SASP-producing senescent cell in a human tumor. Consequently, a methodology to identify a senescent cell in a person (or a mouse) is needed to understand how senescence contributes to human disease, aging, or cancer therapy (31–36).

Here we report the first-of-its-kind phase II study of abemaciclib in DDLS (JCO.2019.37.15_suppl. 11004). Correlative biopsies from this trial were used to establish that senescence is a common outcome of treatment with abemaciclib, leading to an increase in TIL accumulation, and potential resistance for some after a promising initial response.

Experimental design

This was an investigator-initialed, single-center nonrandomized open-label phase II study. The trial was intended to evaluate the biological activity of abemaciclib. Eligible patients were adults with advanced dedifferentiated liposarcoma (DDLS), ECOG performance status of 0 or 1, and progression of disease by RECIST v1.1 in the 6 months prior to study entry. CDK4 amplification was not required at study entry (because it is nearly ubiquitous in this disease), but it was confirmed retrospectively using next-generation sequencing. Any number of prior systemic therapies were allowed, including none. Patients were treated with 200 mg of abemaciclib by mouth twice daily in 4-week cycles. Dose reductions as necessary to alleviate toxicities were undertaken similar to experiences in patients with HR+ metastatic breast cancer who received abemaciclib. PFS at 12 weeks was the primary endpoint, defined as time from initiating abemaciclib treatment until disease progression according to RECIST v1.1. PFS at 12 weeks of >60% was considered promising and a PFS of <35% was considered not promising. With a one-stage design based on the exact bionomial test, the study would be positive if 15 of 30 patients were progression-free at 12 weeks, with type I error 0.07 and type II 0.10 [Fleiss JL, Statistical Methods for Rates and Proportions (1981), pp 13–15].

Samples were divided into two PFS groups based on PFS median (33 weeks). A univariate Cox proportional hazardous model was used to evaluate correlation between clinical variables (i.e., age at study entry, gender, number of prior lines of therapy) and PFS groups. Correlation was assessed with the Wald test, and adjusted P values were calculated using FDR. These studies were conducted in accordance with recognized ethical guidelines such as the Belmont Report and the protocol was approved by the Institutional Review Board of Memorial Sloan Kettering Cancer Center. All patients provided written informed consent. A representativeness of Study Participants is included as Supplementary Table S1.

Cell culture

Palbociclib, abemaciclib, and trametinib were purchased from SelleckChem and used at the concentrations indicated in the text and figure legends.

The LS8817 well-differentiated and dedifferentiated liposarcoma (WD/DDLS) cell line was provided by Samuel Singer's lab and has been extensively characterized (6, 37). These cells are maintained in DMEM with high glucose (4,500 mg/L) supplemented with 10% FBS and 2 mmol/L glutamine. LS8817tetON-FMDM2 cells were generated by transduction of FLAG-tagged MDM2 cloned into the LT3 lentiviral vector backbone and selected with puromycin. Three lines were isolated and the results were similar between them.

Expression of targets in these cells were reduced by transducing quiescent cells (palbociclib + doxycycline) with pLKO.1 lentiviruses (Open Biosystems) expressing the appropriate shRNAs (sequences given in Supplementary Table ST2). pLKO.1-TRC cloning vector was a gift from David Root (Addgene plasmid #10878, RRID: Addgene_10878). Lentivirus vectors were prepared in 293T cells (ATCC, Catalog No. CRL-3216: RRID:CVCL_0063) as described previously (5).

A549 cells (ATCC, Catalog No. CCL-185, RRID:CVCL_0023) were cultured in DMEM with high glucose (4,500 mg/L) supplemented with 10% FBS and 2 mmol/L glutamine. Recombinant human ANGPTL4 was purchased from RD Biosystems (Full-length: 4487-AN-050; C-terminal: 3485-AN-050) and reconstituted in sterile PBS. A549 cells treated with palbociclib for 2 days were then incubated with palbociclib + 100, 1,000, or 2,000 ng/mL rANGPTL4 for a total of 6 days, with one media change.

HUH-7 cells (ATCC, Catalog No. Hep3B2.1–7, RRID:CVCL_0326) were a gift from Amaia Lujambio (Mt Sinai Medical Center) and cultured in DMEM with high glucose (4,500 mg/L) supplemented with 10% FBS and 2 mmol/L glutamine.

SK-HEP-1 cells obtained from ATCC (ATCC, Catalog No. HTB-52, RRID:CVCL_0525) and Hep G2 cells (ATCC, Catalog No. HB-8065, RRID:CVCL_0027) from Ping Chi at MSKCC were cultured in MEM supplemented with 10% FBS and 2 mmol/L glutamine.

MCF7 cells were originally obtained from ATCC (ATCC, Catalog No. HTB-22, RRID:CVCL_0031) and adapted to culture in DMEM with 10% FBS and 2 mmol/L glutamine. ZR-75–1 (ATCC, Catalog No. CRL-1500, RRID:CVCL_0588) and CAMA-1 (ATCC, Catalog No. HTB-21, RRID:CVCL_1290) were obtained from ATCC and cultured as pe their recommendations in RPMI1640 supplemented with 10% FBS and EMEM supplemented with 10% FBS, respectively.

Immunoblots

Proteins were extracted from cell lines in a buffer containing 50 mmol/L Tris-HCl, pH 7.4, 250 mmol/L NaCl, 5 mmol/L EDTA, 0.5% NP40, 2 mmol/L PMSF, and 1 mg/mL each aproteinin, leupeptin, and soybean trypsin inhibitor.

Extraction and immunoblotting conditions for patient biopsies were exactly as described previously (6), using an SDS-based lysis buffer.

Tubulin was detected with the Santa Cruz Biotechnology C-11 antibody at 1:2,000, FLAG-MDM2 with the Sigma M2 FLAG antibody at 1:1,000, and MDM2 using mAb SMP14.

Senescence assays

Senescence-associated  β-galactosidase (SA-β-gal) was assayed using the Senescence β-Galactosidase Staining Kit (Cell Signaling Technologies, 9860) according to the manufacturer's instructions.

Senescence-associated heterochromatic foci (SAHF), ATRX foci, BrdU incorporation were assayed by immunofluorescence as described previously (6).

Clonogenic growth arrest assays were performed as described previously (6), and colony formation was either qualitatively assessed (colonies ≥2 mm) or quantitatively determined by calculating the percent area of the well that contained colonies using ImageJ (ImageJ, RRID:SCR_003070).

Statistical analysis was carried out by one-tailed unpaired t tests using the Prism software, and P values are reported. Data are presented as a mean of technical replicates with error bars derived from the SEM.

Metabolic sequencing

Extracts were prepared as described (38) and profiled at Human Metabolome Technologies.

RNA sequencing

RNA was isolated as described previously (5). Quality was checked on a BioAnalyzer to ensure a minimum RNA Integrity Value of 7. Libraries were then generated using 500 ng of input RNA per sample, according to the manufacturer's instructions for TruSeq mRNA Library Prep Kit V2 (Illumina) with eight cycles of PCR. Libraries were pooled and run on an Illumina HiSeq 2500 high output to obtain 40 million paired-end 125 nucleotide-length reads. The reads were aligned to the human reference sequence hg19 using the STAR software version_2.4.0c package (STAR, RRID:SCR_004463). Raw counts were inputted into R studio version 3.2.0 (cran.r-project.org) and differential gene expression determined using the Bioconductor package DESeq2 (Bioconductor, RRID:SCR_006442; DESeq, RRID:SCR_000154) comparing expression in growth arrested quiescent LS8817TetONFMDM2 cells (DoxCDK4i) to each of the time points. Significant changes in gene expression were set at a fold change cutoff of ±1.5 and a FDR of 0.5% (Padj < 0.005). Heatmaps were generated using gplots software. Hierarchical cluster analysis was performed with the hclust function in R. Software from the Broad Institute was used to perform gene set enrichment analysis on the significantly altered transcripts identified by DESeq2 compared against the C5 Gene Ontology (GO) gene sets. Pathways with a FDR of Padj < 0.05 were considered significant. A conserved transcript signature was compiled from the overlap of gene expression or cytokine production in nine different RNA-seq, microarray, and secretome studies (11–13, 39–44). The data can be accessed from the GEO repository accession no. GSE241031.

RT-qPCR

RNA was extracted from cells using the QIAGEN RNeasy Kit according to the manufacturer's instructions. Complementary DNA (cDNA) was synthesized and qPCR performed on the QuantStudio 6 system as described previously (5). Data were analyzed using the ΔΔCt method, and changes in gene expression were plotted relative to untreated cell controls. Primer sequences can be found in Supplementary Table S2.

Single-cell analysis of cell lines

Palbociclib-treated and control single-cell suspensions were prepared following 10× Genomics guidelines for cultured cells. Suspensions were adjusted to concentration of 1,000 cells/μL and loaded to one capture well to obtain 10,000 cells. 10× genomics libraries were prepared using v3 chemistry and sequenced to an average of 23,000 reads per cell quantifying expression for a median of 1,509, and 2,733 genes per cell for palbociclib-treated and control, respectively. Sequence reads were aligned to the GRCh37/hg19 human genome reference using STAR (RRID:SCR_004463) as implemented by the Cellranger v3.0.2 pipeline. Subsequently, we analyzed the resulting BAM file with velocyto to obtain gene read counts (RRID:SCR_018167). Merged treatment and control spliced read counts were used for clustering with Seurat (RRID: SCR_007322), implementing the SCTransform pipeline, with 20 dimensions (45). Differential expression analyses between palbociclib treated versus control, and within treated cells, comparing ANGPTL4 high versus low were carried out in Monocle3 (46, 47). Both comparisons excluded the high mitochondria clusters. Significance thresholds were set at an FDR corrected P-value ≤0.05. The data can be accessed from the GEO repository accession no. GSE241225.

Immunofluorescence and IHC for patient biopsies

ANGPTL4 and CDKN2A expression was measured by RNAscope on serial sections from FFPE biopsies with probes ACD455358 and ACD310188, respectively, according to the manufacturer's instructions (ACD Inc.). Positive and negative control probes were provided by the manufacturer. 1:100 Tyramide Alexa Fluor 488 (Life Technologies, B40932) replaced the DAB step. Sections were counterstained with 10 mg/mL DAPI (Sigma-Aldrich, D9542) and mounted with MOwiol 18–88 mounting media (Sigma-Aldrich, 81365). Imaging was carried out using a confocal microscope at 63× magnification.

To quantify mRNA expression, slides were scanned on a Panoramic P250 Flash scanner (3DHistech) using 20×/0.8NA objective lens. Tumor regions were identified in H&E stained sections by pathology and exported as .tif files from those scans using Caseviewer (3DHistech). Each image was divided into a grid and each tile was analyzed using ImageJ/FIJI (NIH). Median filter, thresholding, and watershedding were used to segment the nuclei in the DAPI channel and obtain cell counts. Signal intensity was measured using a threshold, and maxima were found to determine foci counts per region. All tiles for a time point were then plotted and the median expression level defined. Changes in the distribution were assessed by both Mann–Whitney and Kolmogorov–Smirnov nonparametric tests.

IHC to detect Ki67, CD8, CD4, and FOXP3 positive cells was carried out in our clinical laboratory at Memorial Hospital. To quantify expression, slides were scanned and analyzed as above except that hematoxylin and DAB signals were separated using color deconvolution in ImageJ/FIJI (NIH) prior to counting. CDH18 was detected as described previously (4).

Statistical analysis

For analyzing the association of senescence with progression-free survival, a time-dependent cox model was run (48–50) on R version 4.0.5 using the survival package.

In addition, a Cox proportional hazards model with time-variable effects was specified and deployed by first defining the survival function, |${\boldsymbol{S}}( {\boldsymbol{t}} ) = \exp ( { - \gamma ( {\boldsymbol{t}} )} )$|as a measure of the survival probability after time t, where |$\gamma ( {\boldsymbol{t}} )$|is the cumulative hazard function. The cumulative hazard function defines the risk accumulated over the interval (0, t), and can be further specified as the time-integral over the instantaneous hazard rate |$\gamma ( {\boldsymbol{t}} ) = \mathop \int \nolimits_0^{\boldsymbol{t}} \lambda ( {\boldsymbol{s}} ){{\bf d}}{\boldsymbol{s}}$|⁠. To make the model tractable and allow for variation between the senescence+/inflammation+ and other groups, we introduce the coefficient β, which represents the effect of group on the instantaneous hazard rate. Further, we allow for β to be time varying, and model it as a Weiner process specified by |$\beta ( {{\boldsymbol{t}} + {\boldsymbol{u}}} ) - \beta ( {\boldsymbol{t}} )\sim{\boldsymbol{N}}( {0,{\boldsymbol{u}}} )$|⁠, which allows the model to learn slowly varying changes in the effect size. The Cox proportional hazards model can thus be specified as:
formula
where |${\lambda }_0( {\boldsymbol{t}} )$|is the baseline instantaneous hazard rate, and |${{\boldsymbol{x}}}_{\boldsymbol{i}}$| is the group index. To define the full model, we discretize the continuous variable t into blocks of 3 days, and treat each patient as having been exposed to the hazard rate |$\lambda ( {\boldsymbol{t}} )$|up until either tumor growth or censorship. The model is fully specified by modeling the observations of tumor growth as a Poisson process with underlying rate parameter |$\lambda ( {\boldsymbol{t}} )$|⁠:
formula
formula
formula
formula
where α and β were chosen to be 0.01, and where the subscripts i and j refer to the individual, and the discretized time interval, respectively. We sampled the model using PYMC3’s NUTS sampler with a burn-in trace of 1,000, followed by 1,000 samples from the posterior distribution. Our sampler showed strong convergence, with tight estimates of parameter |$\beta ( {{{\boldsymbol{t}}}_{{\boldsymbol{j}} + 1}} )$| up through the 1,500 day mark. Our posterior estimates of exp(β) indicate an overall effect of senescence+/inflammation+ being associated with a 58% increase in hazard rate, with significance of P = 0.0373 (measured using a ROPE correlate, defined as any decrease in hazard associated with being senescence+/inflammation+).

Data and materials availability

Sequencing data are available without restriction at GEO under the numbers GSE241031 (Therapy-induced senescence contributes to the efficacy of abemaciclib in patients with dedifferentiated liposarcoma) and GSE241225 (Gene expression at single cell level of untreated and 10-day palbociclib treated LS8817 cells). All other data are available in the main text or the Supplementary Materials and Methods and can be requested directly from Andrew Koff.

Single-agent abemaciclib phase II clinical trial in patients with dedifferentiated liposarcoma

We performed a single-center, nonrandomized, open-label phase II study of abemaciclib in patients with advanced DDLS (Experimental and Statistical Design, Materials and Methods). We have previously shown that the median progression-free survival of palbociclib in DDLS was 18 weeks (2, 3), but palbociclib was dosed intermittently, potentially allowing pre-senescent cells in the tumor the chance to return to the cell cycle during periods of drug withdrawal. Abemaciclib, however, is a structurally distinct CDK4/6i, which is dosed continuously (reviewed in ref. 10). In this phase II study, eligible patients, adults (>18 years) with an ECOG performance status of 0 or 1, any (or no) prior therapy, and progressive disease by RECIST v1.1 in the 6 months prior to study entry, received abemaciclib 200 mg orally twice a day in 4-week cycles. Two image-guided biopsies were performed for correlative analyses: one within 3 weeks prior to treatment start and another after 4 to 6 weeks of therapy. Tumor growth was assessed by an independent reference radiologist every 6 weeks for 36 weeks, and every 12 weeks thereafter. The primary endpoint was PFS at 12 weeks. The study would be positive if the 12-week PFS was ≥60%. Toxicities were graded according to the NCI's Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. The study was approved by the Institutional Review Board at Memorial Sloan Kettering Cancer Center and all patients provided informed written consent.

Between August 2016 and October 2018, 30 patients were enrolled. Study data were locked on June 1, 2021. One patient withdrew consent prior to the first follow-up scan. Patient demographics (Supplementary Fig. S1) and genomic profiles were consistent with previous trials with palbociclib (2, 3). Common genomic alterations included amplification of CDK4 and MDM2 on chromosome 12, along with variable loss and gain of other regions and some scattered patient-specific mutations. There was no significant association of PFS with demographics, copy-number variation, or prior therapy (Supplementary Table S3).

Nine patients underwent dose reduction to 150 mg, and 2 to 100 mg (Supplementary Table ST4), which was consistent with expectations based on the use of abemaciclib in breast cancer. On-treatment biopsies for patients 1, 8, and 30 were taken on the reduced dose of 150 mg and all others were on the original 200 mg dose.

Twenty-three of 30 patients were progression-free at 12 weeks (76.7%, two-sided 95% CI, 57.7%–90.1%), and thus the study met its primary endpoint (Fig. 1A). The median PFS as of the data lock was 33 weeks (two-sided 95% CI, 28–72 weeks). Three patients had a partial response and 6 had stable disease for 2 years or longer (Fig. 1B). As expected with abemaciclib, the most common adverse effects were gastrointestinal and hematologic, and were generally manageable (Supplementary Fig. S2).

With the exception of patient 10, all patients, regardless of their eventual response to the drug, experienced Ki67 reductions after a month of abemaciclib (Fig. 1C). However, we cannot rule out the possibility that the result with patient 10 is a sampling artefact because of the intratumoral heterogeneity that is possible in large liposarcomas (ref. 51 and unpublished data RGM and SS).

Thus, abemaciclib has meaningful activity in advanced and progressive DDLS and is now being evaluated in a large, randomized, phase III trial (NCT 04967521).

Transcriptome profiling during geroconversion in LS8817 cells

We wanted to develop an approach that would identify candidate gene products necessary for CDK4/6i TIS in DDLS cells, which might be useful as a “senescent-cell” marker in patients. To accomplish this, we took advantage of our understanding how the effect of palbociclib-induced changes in MDM2 affected the choice between senescence and quiescence in DDLS cell lines (6). Specifically, MDM2 accumulation is suppressed in DDLS cells that undergo senescence after palbociclib treatment, but enforcing MDM2 expression from a heterologous CMV promoter in such cells prevents them from undergoing senescence while still allowing them to exit the cell cycle (6). On the other hand, those DDLS cell lines wherein palbociclib did not induce the loss of MDM2 would exit the cell cycle but not senesce. Reducing MDM2 expression in such quiescent cells by infecting them with lentiviruses expressing shRNAs targeting MDM2, directly induced their conversion to senescence (6). Coupling these observations with inductive synchrony methods we reasoned that it we might be able to identify, in an unbiased manner, changes that occur during CDK4/6i TIS, specifically during geroconversion—the transitional pathway that occurs as a cell undergoes reversible (quiescence) to irreversible (senescence) exit.

We transduced LS8817 cells with a vector that allowed doxycycline-regulated expression of MDM2, and isolated individual single-cell clones, which were expanded into cell lines (LS8817TetOnFMDM2). Three clonal lines gave similar results. To be useful for inductive synchrony, we needed to first establish that the cells could be “locked” into a quiescent state. As shown in Supplementary Fig. S3, the cells were stably maintained in a noncycling reversible state in media containing doxycycline and palbociclib for up to 6 days, a time at which palbociclib alone would have induced stable arrest as measured by long-term clonogenicity assays, accumulation of the number of ATRX foci per cell, and the percentage of SA-β-gal positive cells.

After these cells were synchronized in quiescence, we removed doxycycline and continued to culture them in palbociclib alone for as long as 28 days (Fig. 2A). MDM2 levels decreased by approximately 50% within 3 days after removal of doxycycline (Fig. 2B). At that time, the cells began to acquire phenotypic changes associated with cellular senescence. Concurrent with the decrease in MDM2, an increase in the number of ATRX foci in each cell was noted (Fig. 2C). This was followed 2 days later by the accumulation of SA-β-gal (Fig. 2D) and SAHF (Fig. 2E). It was another week (day 14) before stable cell-cycle exit occurred, as measured by colony formation after replating in drug-free media (Fig. 2F). Metabolic changes are commonly reported in senescent cells (52, 53), and using and unbiased LC/MS approach, we found that the most profound metabolic changes occurred coincident with the time of stable growth arrest, well after SA-β-gal had peaked (Supplementary Fig. S4). Thus, synchronization was able to clearly separate those phenotypes most commonly used to define senescent cells—accumulation of SA-β-gal or SAHF—from its most foundational phenotypes: stable exit from the cell cycle. Similar results were obtained when senescence was induced with abemaciclib (data not shown).

As would be expected of cells undergoing senescence, the transcriptional landscape also changed over the course of geroconversion (Fig. 3A and B). Approximately 800 transcripts changed over the 28-day period (Supplementary Table S5), with the patterns of gene expression shifting from cell-cycle exit to immune system processes and extracellular communication (Fig. 3C)

Comparing these 800 transcripts to approximately 150 transcripts identified in primary cells induced to senesce by non-CDK4/6i stressors (Supplementary Table S6), we defined a 37-transcript signature overlap, and this is where we focused to identify potential regulators of geroconversion. The changes in this subset of transcripts during geroconversion are shown in Fig. 3D. Sixteen of them increased coincidently with stable growth arrest and remained elevated over the next 2 weeks (Fig. 3E). We hypothesized that the proteins encoded by one or more of these might be necessary for stable arrest, and thus a useful marker of an inflammation-provoking senescent cell.

ANGPTL4 is necessary for stable arrest and induction of the SASP program in LS8817 cells treated with palbociclib

We reasoned that a regulator of geroconversion during CDK4/6i TIS, which might eventually serve as a marker to define a stably arrested, inflammation-provoking senescent cell in a human, would be one that is also increased in other liposarcoma cell lines, which undergo CDK4/6i induced senescence, but not those that simply exited the cell cycle following treatment. From our previously published work when we had characterized cell line responses (quiescence or senescence) to palbociclib (5, 6, 10), we had stockpiled RNA, and thus we looked at the expression of eight of these transcripts in a panel of five genetically diverse DDLS cell lines treated with palbociclib, LS8817 cells induced to quiescence by serum starvation, and palbociclib treated quiescent U2OS osteosarcoma cells which are ATRX deficient. ANGPTL4 and MMP3 were commonly induced in cells that underwent senescence, but not in those that remained quiescent (Fig. 4). However, only ANGPTL4 was not induced when LS8817 cells were made quiescent by serum starvation. ANGPTL4 also increased in abemaciclib-treated LS8817TetONFMDM2 cells with similar kinetics to that observed with palbociclib (unpublished data). Thus, ANGPTL4 expression was increased in genetically diverse cell lines that undergo CDK4/6i induced senescence but not quiescence and might have potential to identify senescent cells in human biopsies.

The best marker would also be one that is necessary for the senescent cell type to form. Thus, we looked at whether ANGPTL4 was necessary for durable growth arrest and the triggering of the SASP. To accomplish this, we transduced quiescent LS8817TetONFMDM2 cells cultured in doxycycline and palbociclib with two different lentiviral shRNAs targeting ANGPTL4, or with shRNAs targeting other transcripts whose expression increased at day 14. Then, after selection, released them into palbociclib for 21 days to induce senescence (Fig. 5A). We chose to transduce the quiescent population as we did not want to deal with artefacts that might affect drug-induced cell-cycle exit, proliferation, or viability. Complete data for cells in which ANGPTL4 and IGFBP7 were knocked down is shown in Fig. 5B. The results of the two shRNA targeting vectors knocking down other transcripts are shown in Supplementary Fig. S5.

While reducing ANGPTL4 expression did not affect the accumulation of ATRX foci (Fig. 5C) or SA-β-gal (Fig. 5D), these cells did re-enter S-phase (Fig. 5E) and form colonies (Fig. 5F) when palbociclib was removed after 21 days. In contrast, reducing IGFBP7 expression did not affect the accumulation of ATRX foci or SA-β-gal, and the cells released from palbociclib 21 days later did not incorporate BrdU or form colonies in an appreciable fashion (Fig. 5CF).

The SASP is a cell-type and stressor-specific transcription and secretory program involving various chemokines, interleukins, matrix remodeling enzymes, and growth factors that could communicate to affect the surrounding tumor microenvironment (11–13). In p53-dependent models of senescence, the complete SASP program is dependent on the presence of unresolved DNA damage (54). However, CDK4/6i TIS is neither p53-dependent (6), nor associated with DNA damage (Supplementary Fig. S6); thus, we looked at how 17 different SASP-encoding transcripts that increased in LS8817 cells treated with CDK4/6i were affected by knocking down ANGPTL4 or IGFBP7. Several transcripts were reduced in the cells transduced with shANGPTL4, but not in those transduced with either shIGFBP7 or shSCRambled controls (Fig. 5G). These included CCL20, IL1B, and MMP3. The expression of IL1A and IL6 were more strongly affected by the loss of ANGPTL4 as well. In contrast, IL1A, IL1B, and IL6 levels were grossly higher in cells expressing shIGFBP7 compared with those expressing the nontargeting shRNA. The significance of this, if any, is unclear. On the other hand, all five of the mRNAs encoding growth factors (CSF1, TNFRSF11B, IGFBP5, ANG, and CSF2) were grossly upregulated in the shANGPTL4 cells compared with their expression in the cells expressing shIGFBP7 or shSCRambled. Thus, without ANGPTL4-associated stable arrest, some elements of the SASP program were dysregulated, but not all, similar to what occurs in p53-dependent DNA damage senescence models as well. Collectively, these data suggest that ANGPTL4 is a late-regulated necessary gene product required for the evolution of a SASP-producing senescent cell from a CDK4/6i arrested quiescent cell.

ANGPTL4 defines a SASP-expressing palbociclib-induced population of LS8817 cells

To determine whether ANGPTL4 was expressed in a small population of cells that might act in a paracrine fashion to induce the SASP in other cells, we looked at the association of ANGPTL4 with the expression of other SASP transcripts in palbociclib-treated LS8817 cells by single-cell sequencing. Unmanipulated parental LS8817 cells were treated with 1 mmol/L palbociclib for 10 days, a time sufficient to induce stable arrest and SASP-transcript expression (6) and subjected to single-cell sequencing. High-quality data were obtained from 7,562 untreated cells and 5,872 palbociclib-treated cells, and as expected, the treated cells were largely in the G1 phase (Fig. 6A and B). Although all the palbociclib-treated cells underwent a stable, long-term clonogenic arrest in vitro, ANGPTL4 was only expressed at detectable levels in about 25% of these cells (Fig. 6C). This was not a timing effect as similar results were observed in cells treated for 7, 14, 21, or 28 days (data not shown). We suspect that the relative expression of ANGPTL4 mRNA compared with others, coupled to the shallow depth of sequencing in a single-cell experiment, accounts for this. Consistent with this, CDKN2A, while expressed at its highest level in ANGPTL4+ cells, was expressed in both untreated cells and, more weakly, in the ANGPTL4 palbociclib-treated cells (Fig. 6D). Thus, ANGPTL4 expression clearly divides the palbociclib-treated cells into two groups, one we named PALBO-A (ANGPTL4 positive) and the other PALBO-B (ANGPTL4 negative).

Which of these expressed the SASP transcripts? Palbociclib-treated cells made up six Seurat clusters (1, 7, 9–11, 13), and the untreated cells seven (0, 2–4, 6, 8, and 14; Fig. 6E). ANGPTL4 expression was largely restricted to clusters 1 and 7, with a few positive cells also seen in cluster 10—the same clusters in which the SASP transcripts were largely expressed (Fig. 6F). Thus, PALBO-A represents stably arrested and SASP-producing senescent cells, and thus would be a suitable candidate to identify such cells in patients.

Having the single-cell sequencing data also gave us the opportunity to consider that an ideal clinical marker is one that can be scored positively or negatively, similarly to hormone receptors (Holland RL 2016 Advances in Precision Medicine). We detected 32 of the 37 transcripts in the single-cell data (again, likely due to the shallow sequencing). Five were clearly enriched in the treated population relative to the untreated population: CXCL11, IGFBP5, ANGPTL4, CXCL5, and FGF7 (Supplementary Fig. S7). ANGPTL4 and FGF7 were expressed in the largest number of treated cells (22% and 18%, respectively) and in very few (less than 1%) of the untreated cells. Collectively, these suggest that ANGPTL4, a necessary regulator of senescence induced by CDK4/6i, is likely to be a reliable candidate for detecting senescent cells in patient materials in vivo. First, ANGPTL4 expression was low in untreated and quiescent cells, and highly expressed in cells that underwent senescence in response to CDK4/6i. Second, ANGPTL4 was necessary for stable cell-cycle exit and was associated with subsequent SASP gene expression in response to palbociclib.

ANGPTL4 can collaborate with palbociclib-induced cell-cycle exit to drive other types of cancer cell lines into senescence

Studies in another experimental background lend further support to the conclusion that ANGPTL4 is involved in CDK4/6i therapy induced geroconversion. The nonsarcoma lung cancer derived cell line, A549, when treated with palbociclib, will exit the cell cycle but will only senesce if trametinib is added, either simultaneously or after pretreatment with palbociclib treatment (25). In these cells, ANGPTL4 expression increased in the presence of palbociclib and trametinib, but not when treated with palbociclib alone (Supplementary Fig. S8), whereas an increase in the number of ATRX foci per cell and the percentage of cells positive for SA-β-gal were equivalent with either treatment condition (Supplementary Fig. S8).

Thus, we asked if recombinant ANGPTL4 would be sufficient to replace trametinib and induce senescence in palbociclib-treated A549 cells. Adding recombinant ANGPTL4 directly to the media of palbociclib-treated A549 cells was able to increase the percentage of cells that underwent irreversible arrest (Supplementary Fig. S8), albeit not as effectively as trametinib.

Because ANGPTL4 processing is quite complex (55), this could indicate that the recombinant protein might not be processed fully to accomplish its senescent-promoting role. Thus, we also generated two isogenic clones of A549 cells, one expressing GFP and the other ANGPTL4 (Supplementary Fig. S9). ANGPTL4 expression was increased approximately 20-fold in those cells, yet BrdU incorporation, the number of ATRX foci per cell, and the accumulation of SA-β-gal were comparable with those expressing GFP and parental cells. Nevertheless, after 14 days in palbociclib, those ANGPTL4 expressing cells fail to form colonies when replated in drug-free medium, and the expression of a number of the idiotypic SASP-related transcripts were increased compared with the parental and control GFP expressing cells (Supplementary Fig. S9). These suggested that MEK inhibitor induced induction of ANGPTL4 might be sufficient to account for its ability to drive A549 cells through palbociclib-induced geroconversion.

However, the heterogeneity of cellular senescence portends that this may not be the same in all cell types treated with palbociclib. For example, we looked at the expression of ANGPTL4 and other senescent cell phenotypes in three ER+ breast cancer (MCF7, CAMA-1, ZR-75–1) and three hepatic cancer (SK-HEP-1, HUH-7, Hep G2) cell lines, all of which exited the cell cycle within 24 to 48 hours after treatment with 1 μmol/L palbociclib. All six to varying degrees underwent senescence following 14 days of palbociclib treatment as measured by long term clonogenic arrest (Supplementary Fig. S10). However, although the percentage of SA-β-gal staining cells increased in five of the lines and the number of nuclear ATRX foci in three of them, the expression of ANGPTL4 did not increase in any of the three ER+ breast cancer cell lines [and decreased significantly in MCF7 (P < 0.05)], and albeit trended positively in the liver cancer cell lines, it was only statistically significant in Hep G2 cells. Thus, ANGPTL4 would not be a useful marker of CDK4/6i TIS in patients with ER+ breast cancer or liver cancer. Nevertheless, the approach described above to identify such a marker is likely to be effective in such diseases as well.

The relationship between ANGPTL4 and infiltrating leukocytes in tumors

With the lines of cellular evidence above indicating a necessary association between the expression of ANGPTL4 with SASP expression in palbociclib-treated senescent dedifferentiated liposarcoma cells, and the potential limitations of this marker vis-a-vis the cancer type dependence, we were ready to use the rare correlative biopsies from the abemaciclib trial to determine if ANGPTL4 would be a marker to evaluate senescence in our patients in the future.

Acutely formed senescent cells in mice, either because of therapy or oncogene expression, induce an idiotypic inflammatory immune response which could involve macrophages, NK cells, induced NK cells, neutrophils, FoxP3+ Tregs or CD4+ or CD8+ cytotoxic cells (14, 18–21, 23–26, 40, 56). Thus, to determine if increased ANGPTL4 expression was associated with an immune response in patients treated with abemaciclib for 1 month, we prepared serial sections from 20 paired pretreatment and a 1-month on treatment biopsy samples and probed these for ANGPTL4 and CDKN2A by RNAscope, and CD4+, CD8+, and FoxP3+ Tregs (collectively TILs) by IHC and scored the changes from pretreatment to on-treatment. We also used a CD68 antibody to confirm that macrophages were quite abundant in the liposarcoma samples before treatment (57), and thus were unable to draw any conclusions about observed changes unless we measured their polarization state which we could not give the lack of material. Images from a patient positive for ANGPTL4 are shown in Fig. 7A. The data for all patients are compiled in a heat map (Fig. 7B), and the raw data with distributions in cancer bearing regions of the biopsy are in the supplemental material (Supplementary Figs. S11–S15).

CDKN2A expression increased in 15 patients, 14 of whom also had increased ANGPTL4 expression. Of those 5 patients in which CDKN2A was not increased, 2 had elevated expression of ANGPTL4. Thus, drug-induced accumulation of ANGPTL4 mRNA and CDKN2A mRNA were well correlated (Fig. 7C), consistent with the notion that ANGPTL4 expression is a later marker of senescent cells.

The number of TILs increased in 16 of the patients after 1 month on abemaciclib. Fifteen of those had an increase in CD4+ cells, and six showed an increase in FOXP3+ cells. One patient with elevated FOXP3+ did not have corresponding increase in CD4 positivity. Increases in CD8+ cells were rare. Thus, being ANGPTL4 and CDKN2A positive was well correlated with an increase in TILs (Fig. 7D), specifically CD4+ cells.

There was not enough material to determine if the immune response, vis-a-vis tolerance or exhaustion, was different in those patients with TILs who also had accumulated ANGPTL4/CDKN2A positive cells and those with TILs who had not accumulated ANGPTL4/CDKN2A positive cells. However, the fold increase in CD4+ cell number was significantly higher in the senescent group (Fig. 7E). Consequently, the infiltration of CD4+ cells define an effect of abemaciclib and suggest that coincident formation of senescent cells by the drug can augment its magnitude.

The effect of senescence on progression-free survival

With palbociclib in our prior studies, we reported some extension of median PFS when MDM2 expression was reduced posttreatment in extracts from biopsies (6), this was not obvious when using ANGPTL4 expression to define senescent cells in these patients treated with abemaciclib. Given this, we wanted to determine the extent to which CDH18 expression in pretreatment biopsies and MDM2 turnover in posttreatment biopsies, was related to ANGPTL4 in patients who received abemaciclib.

To evaluate drug-induced changes in MDM2, we made extracts from a different set of 10 paired pre- and on-treatment biopsies obtained from those patients in which a different set was used for the ANGPTL4/CDKN2A/TILs analysis above. MDM2 was reduced in 5 patients. Four of these patients were also CDH18+. However, neither the loss of MDM2, nor the expression of CDH18 prior to treatment, associated with the duration of PFS in patients treated with abemaciclib (Supplementary Fig. S16). Nevertheless, increased ANGPTL4 expression was associated with MDM2 downregulation further strengthening the notion that geroconversion associated with MDM2 loss and increased expression of ANGPTL4 occurred in some of the patients treated with abemaciclib. We suspect that dosing schedules, intermittent palbociclib versus continual abemaciclib, may obscure median PFS values as a chronic quiescent cell may look very similar to a stably arrested senescent cell, at least as far as tumor volume is concerned.

In addition, neither the presence of absence of TILs, or Tregs, or ratios between these cells before or after treatment associated with outcome. However, patients in which abemaciclib induced both senescence and an increase in TILs seemed to have a better initial outcome, many with stability up through 6 months, but then had a 58% increase in risk of progression compared with those who did not have both responses to the drug on the early on treatment biopsy (P = 0.0373, Fig. 7F). Using a ROPE correlate, the increase instantaneous hazard rate was most significant (P < 0.05) between days 194 and 246, which overlaps the median PFS of this cohort at the time of data lock (Fig. 7G). This raises the possibility that chronic persistence of senescent cells in patients who respond to CDK4/6i could create a permissive environment from which resistance occurs. Senolytics might be a useful way to improve outcomes in patients where this occurs.

Senescence is one potential mechanism by which CDK4/6 inhibitors may exert their effect in extending progression-free survival in patients. Senescent cells can stabilize cell-cycle exit or promote inflammation (8–10, 35). CDK4/6 inhibitors can also promote inflammation independently of senescence by increasing tumor cell antigenicity by inducing MHC class I antigen presentation (15), or they can directly affect the maturation of various types of immune cells (9, 58). Thus, any rational approach to improve CDK4/6i therapy is limited without a means to detect whether senescent cells arise during therapy. Co-ordinating a laboratory-based investigation of the molecular pathway underlying CDK4/6i-induced senescence with a phase II clinical study of abemaciclib in DDLS allowed us to define a necessary regulator required for geroconversion, affording us the opportunity to discover that senescence does contribute to inflammation in patients.

Senescent cells provoke inflammation in a variety of contexts (59, 60). The initial immune response at 1 month in the CDK4/6i treated patients was an increase in the numbers of infiltrating TILs, particularly CD4+ cells, and although this was not completely dependent on senescence, it was augmented when senescence occurs. Thus, senescence contributes to the immediate inflammatory response following treatment with the drug. There are no mouse models of DDLS, but in mouse models of other diseases, the nature of the cell types involved in the immunologic response varies with disease context in other models (18–26). This is likely due to the heterogeneity of the SASP effectors produced in a cell type and context-specific manner. The initial recruitment of CD4+ cells was noted in the earliest studies of oncogene-induced-senescence in mouse liver cancer (20) and probably contributes to the later recruitment of the NK cells that lead to their elimination (61). However, unlike mouse models where senescent cells can be eliminated by immunologic responses, tumors in patients are not typically eliminated by CDK4/6i, rather tumor progression or metastatic outgrowth is delayed.

It appears that TIS with a concomitant inflammatory response after exposure to abemaciclib is a possible predictor of early response to treatment. However, the long-term effects of senescence may not be beneficial in some patients. It is not clear in humans or mice how the chronic accumulation of senescent cells can alter the immunologic response over time during therapy. Our data suggest that it is critical to understanding why some patients with a robust initial senescence response develop resistance after a meaningful time on drug. In studies linking senescent cells to aging, it is accepted that the immune response to an acutely formed senescent cell is likely different than to a chronically persistent one. It is unknown how this evolves over time. It is plausible that both senescent cells co-evolve with the immune response. Taking multiple samples over time in patients treated with CDK4/6i and combining this with studies in mouse models where one can temporally manipulate the changes in senescent cells and immune responses would be useful. Nevertheless, it is reasonable to consider combining senolytics and immune modulators with abemaciclib in certain patients with DDLS.

Alternatively, some patients had prolonged stable disease even though their tumor cells did not exhibit TIS with a concomitant inflammatory response after initial exposure to abemaciclib. Longitudinal biopsies in these patients would also help inform if cells enter a senescent state later on in therapy, and if so, why kinetics may differ within tumors or between patients or if these patients benefit from continued quiescent or other mechanisms of drug activity.

We would be remiss to not discuss the encouraging activity of abemaciclib in DDLS, which is currently being investigated in a randomized phase III study (NCT 04967521). Abemaciclib may also be more efficacious in DDLS than palbociclib (2, 3). Although cross-study comparisons must be made with caution, the drugs have different potential off-target effects and are administered through different dosing schedules (62, 63). For the purpose of this work, it is important to recognize that continuous dosing could maintain cells in a noncycling quiescent state even if they do not progress into senescence. Such “stable” growth arrested cells could contribute to PFS and the interpretations of how these outcomes correlate with various biomarkers of senescence. We also note that the transcriptomic, metabolic, and proteomic profiles of both drugs side-by-side in the LS8817TetONFMDM2 synchronization system are very similar up to the time when ANGPTL4 is expressed but diverged after that (unpublished data). It is therefore reasonable to suspect that the drugs act differently in the senescent cell. This, coupled with their different dosing strategies, may account for any noted clinical difference between the drugs in DDLS or other cancers (i.e., ref. 64).

Understanding senescence as a biology is also important for its clinical application, which is why we ran these two efforts concurrently. It is well known that the SASP contains both tumor suppressing and promoting effectors. What is less clear is whether senescent cells evolve from a tumor suppressing state to one that might be more tumor promoting. In fact, we know very little about how cells transition from a reversible to a stably arrested state. Our current knowledge is limited to geroconversion in primary fibroblasts undergoing DNA damage induced senescence (65), a p53-dependent process completely unlike CDK4/6i TIS. Nevertheless, both systems clearly point out that there is a regulatable pathway between quiescence and senescence and not an either/or decision coming from the duress of a cycling cell. Our data also suggest that senescence is a multistage process, defined herein as pre-senescent, pro-senescent, and senescent states. A pre-senescent cell is reversibly arrested and has manifested changes in ATRX distribution, accumulated SA-β-gal as a marker of lysosomal changes, and experienced numerous transcriptional changes. The lack of stable arrest or the complete metabolic reprogramming should not define as these truly senescent cells. Regrettably, most of the markers used today to define senescence (p16, p53, p21, SA-β-gal), often identify these cells, that is those that have only begun to embark on the senescence pathway, but are not yet truly senescent.

At some point, pre-senescent cells enter a pro-senescent state. This is accompanied by the gross metabolic changes and their stable arrest. A truly senescent state is then achieved when these pro-senescent cells express the SASP. ANGPTL4 is necessary for this final transition in DDLS and A549 cells treated with CDK4/6i. ANGPTL4 is also upregulated in DNA damage induced primary fibroblasts and during wound repair in mice (unpublished data), but other gene products are likely important in other cancer cell lines treated with CDK4/6i or in other contexts in which senescence has been studied. For example, others have reported that components of the SASP can play a necessary role in secondary and paracrine senescence (66–68). However, in those models, loss of the gene product prevents cells from undergoing stimuli-induced cell-cycle exit, rather than moving from quiescence to senescence as we have shown. Thus, ANGPTL4 determines the fate of the noncycling cell and probably does not act as a paracrine or secondary senescence factor. Given that ANGPTL4 is involved in redox regulation, glucose homeostasis, lipid metabolism, and energy homeostasis (55), we suspect that such cytokines and their relationship to cellular metabolites might be a good place in which to mine key regulators of cellular fate related to senescence in different cellular contexts.

Finally, this body of work provides an adaptable process to define key regulators of senescence in diseases in which ANGPTL4 may not be able to define a senescent cell, for example ER+ breast cancer. Using synchronized cells will allow any investigator to reduce the complexity of the “senescence” associated gene products and limit experimental focus to those that associate with long-term clonogenic arrest and induction of the SASP and thus provide markers that are regulatory and useful for such cells.

In sum, it is critical to be able to define truly senescent cells in patients to improve therapeutic responses to CDK4/6i and possibly inform combinations with immune-modulators or senolytics. However, this will be slowed by the number of patients that are typically enrolled in translational clinical trials and the expanding parameters needed to evaluate multiple scientific hypotheses at a time. Bayesian statistical methods (69–72), while not immediately useful for changing treatment, are well suited to the scientific correlative investigations linked to clinical samples obtained from clinical trials. Insight will continue to require the adaptive synthesis of clinical and basic investigations running in an integrative manner as we have demonstrated here.

M.A. Dickson reports grants from Eli Lilly during the conduct of the study as well as grants from Sumitomo and Aadi outside the submitted work; in addition, M.A. Dickson has a patent for companion diagnostic pending to MSKCC. M.E. Klein (Dooley) reports other support from Memorial Sloan Kettering during the conduct of the study; in addition, M.E. Klein (Dooley) has a patent for US9,889,135B2 issued and a patent for W O 2017/156263 pending. S.P. D'Angelo reports personal fees from EMD Serono, Amgen, Nektar, Immune Design, GlaxoSmithKline, Incyte, Merck, Adaptimmune, Immunocore, Pfizer, Servier, Rain Therapeutics, GI Innovations, and Aadi Biosciences during the conduct of the study as well as personal fees from EMD Serono, Amgen, Nektar, Immune Design, GlaxoSmithKline, Incyte, Merck, Adaptimmune, Immunocore, Pfizer, Servier, Rain Therapeutics, GI Innovations, and Aadi Biosciences outside the submitted work. In addition, S.P. D'Angelo reports research funding from EMD Serono, Amgen, Merck, Incyte, Nektar, Bristol-Myers Squibb, and Deciphera; travel, accommodations, and expenses from Adaptimmune, EMD Serono, and Nektar; and participation on a data safety monitoring board or advisory board of GlaxoSmithKline, Nektar, Adaptimmune, and Merck. M.M. Gounder reports personal fees from Ayala Therapeutics, Boehringer, Epizyme, Karyopharm, Rain, Regeneron, Springworks, and TYME outside the submitted work. C.M. Kelly reports other support from Eli Lilly during the conduct of the study as well as other support from Daichii Sankyo, Merck, Amgen, Curadev, Servier, Inhibrx, Immuneonc Therapeutics, Kartos Pharmaceuticals, Xencor, Regeneron, Kartos, Servier, and Deciphera outside the submitted work. P. Chi reports grants from Deciphera and Pfizer/Array and grants and personal fees from NewBay outside the submitted work. S. Movva reports grants from Ascentage Pharma, Pfizer/Trillium, Hutchinson Medipharma, and Tracon and nonfinancial support from Merck and Clovis outside the submitted work. A.M. Crago reports grants from NIH during the conduct of the study as well as personal fees from Springworks Therapeutics and Wolters Kluwer outside the submitted work; in addition, A.M. Crago has a patent for US 9,889,135 Companion Diagnostic for CDK4 inhibitors issued and US Patent application 20190310259-A1 pending. V. Serra reports grants from Instituto de Salud Carlos III during the conduct of the study. B.J. Mehrara reports grants from Regeneron, Pfizer, Atyr, Integra, and Mediflix outside the submitted work. M. Kovatcheva reports grants from Spanish Association against Cancer (AECC), Milky Way Research Foundation, and European Research Council (ERC); nonfinancial support from European Molecular Biology Organization (EMBO); and other support from Galapagos NV, CD3, and Mesoestetic outside the submitted work; in addition, M. Kovatcheva has a patent for US9889135B2 issued, a patent for WO2016065044A1 pending, and a patent for US20190310259A1 pending. W.D. Tap reports other support from Eli Lilly during the conduct of the study as well as personal fees from Eli Lilly, EMD Serono, Mundipharma, C4 Therapeutics, Daiichi Sankyo, Deciphera, Adcendo, Ayala, Kowa, Servier, Bayer, Epizyme, Cogent, Medpacto, Foghorn Therapeutics, Amgen, AmMax Bio, Boehringer Ingelheim, BioAtla, Inhibrx, and PharmaEssentia outside the submitted work. In addition, W.D. Tap has a patent for Companion Diagnostic for CDK4 inhibitors 14/854,329 pending to MSKCC/SKI and a patent for Enigma and CDH18 as Companion Diagnostics for CDK4 Inhibition (SKI2016-021-969) pending to MSKCC/SKI; has served on the scientific advisory board of Certis Oncology Solutions; is a shareholder and cofounder of Atropos Therapeutics; and owns stock in and has served on the scientific advisory board of Innova Therapeutics. A. Koff reports grants from NIH, NCI, Jennifer Linn Fund Cycle for Survival, The Maloris Foundation, and Sarcoma Foundation of America and other support from Eli Lilly during the conduct of the study as well as personal fees from Eli Lilly, Concarlo Therapeutics, and PTT Global Chemical; grants from Eli Lilly and Novartis; and other support from Atropos Therapeutics outside the submitted work. In addition, A. Koff has a patent for US9889135 issued. No disclosures were reported by the other authors.

C.E. Gleason: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. M.A. Dickson: Conceptualization, resources, data curation, formal analysis, project administration. M.E. Klein (Dooley): Conceptualization, data curation, formal analysis, investigation. C.R. Antonescu: Formal analysis, validation. R. Gularte-Mérida: Data curation, formal analysis, investigation, methodology. M. Benitez: Investigation. J.I. Delgado: Investigation. R.P. Kataru: Resources. M.W.Y. Tan: Resources. M. Bradic: Data curation, formal analysis. T.E. Adamson: Data curation. K. Seier: Formal analysis. A.L. Richards: Resources. M. Palafox: Resources, methodology. E. Chan: Visualization, methodology. S.P. D'Angelo: Resources. M.M. Gounder: Resources. M.L. Keohan: Resources. C.M. Kelly: Resources. P. Chi: Resources, supervision. S. Movva: Resources. J. Landa: Resources. A.M. Crago: Resources. M.T.A. Donoghue: Supervision. L.-X. Qin: Formal analysis, supervision. V. Serra: Resources, supervision. M. Turkekul: Investigation, visualization, methodology. A. Barlas: Investigation, visualization, methodology. D.M. Firester: Formal analysis. K. Manova-Todorova: Supervision, visualization, methodology. B.J. Mehrara: Supervision, methodology. M. Kovatcheva: Investigation, writing–review and editing. N.S. Tan: Resources. S. Singer: Resources, funding acquisition, writing–review and editing. W.D. Tap: Conceptualization, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing. A. Koff: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We appreciate innumerable critical discussions with a number of investigators interested in cellular senescence and the mechanisms by which CDK4/6 inhibitors act, in particular, with Manual Collado, Yuri Lazebnik, Vivianna Risca, Sarat Chandarlapaty, Charles Rubin, Gary Schwartz, Geoffrey Shapiro, Charles Sherr, and John Petrini. We greatly appreciate the efforts of Sandra Gomez and Severin I Gharbi (Eli Lilly/Alcobendas) comparing the metabolic effects of abemaciclib and palbociclib over time. We acknowledge the use of the Integrated Genomics Operation Core at the Marie-Josée and Henry R. Kravis Center for Molecular Oncology. Funding for this work was generously provided by the NIH/National Cancer Support Grant P30 CA008748 to Memorial Sloan Kettering Cancer Center; the National Cancer Institute Soft Tissue Sarcoma SPORE P50 CA217694 (A. Koff, W.D. Tap, P. Chi, C.R. Antonescu, and S. Singer); the Jennifer Linn Fund from Cycle for Survival; The Maloris Foundation; a Heidi Connery Memorial Research Grant from the Sarcoma Foundation of America to A. Koff; the Mitchell Goodman fund for Sarcoma Research to M.A. Dickson; and Eli Lilly and Company, who provided abemaciclib and financial support for the clinical trial to M.A. Dickson and W.D. Tap.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

1.
Crago
AM
,
Singer
S
.
Clinical and molecular approaches to well differentiated and dedifferentiated liposarcoma
.
Curr Opin Oncol
2011
;
23
:
373
8
.
2.
Dickson
MA
,
Schwartz
GK
,
Keohan
ML
,
D'Angelo
SP
,
Gounder
MM
,
Chi
P
., et al
.
Progression-free survival among patients with well-differentiated or dedifferentiated liposarcoma treated with CDK4 inhibitor palbociclib: a phase 2 clinical trial
.
JAMA Oncol
2016
;
2
:
937
40
.
3.
Dickson
MA
,
Tap
WD
,
Keohan
ML
,
D'Angelo
SP
,
Gounder
MM
,
Antonescu
CR
., et al
.
Phase II trial of the CDK4 inhibitor PD0332991 in patients with advanced CDK4-amplified well-differentiated or dedifferentiated liposarcoma
.
J Clin Oncol
2013
;
31
:
2024
8
.
4.
Klein
ME
,
Dickson
MA
,
Antonescu
C
,
Qin
L-X
,
Dooley
SJ
,
Barlas
A
., et al
.
PDLIM7 and CDH18 regulate the turnover of MDM2 during CDK4/6 inhibitor therapy-induced senescence
.
Oncogene
2018
;
37
:
5066
78
.
5.
Kovatcheva
M
,
Liao
W
,
Klein
ME
,
Robine
N
,
Geiger
H
,
Crago
AM
., et al
.
ATRX is a regulator of therapy induced senescence in human cells
.
Nat Commun
2017
;
8
:
386
.
6.
Kovatcheva
M
,
Liu
DD
,
Dickson
MA
,
Klein
ME
,
O'Connor
R
,
Wilder
FO
., et al
.
MDM2 turnover and expression of ATRX determine the choice between quiescence and senescence in response to CDK4 inhibition
.
Oncotarget
2015
;
6
:
8226
43
.
7.
Abdul Razak
AR
,
Bauer
S
,
Suarez
C
,
Lin
C-C
,
Quek
R
,
Hütter-Krönke
ML
., et al
.
Co-targeting of MDM2 and CDK4/6 with siremadlin and ribociclib for the treatment of patients with well-differentiated or dedifferentiated liposarcoma: results from a proof-of-concept, phase Ib study
.
Clin Cancer Res
2022
;
28
:
1087
97
.
8.
Fassl
A
,
Geng
Y
,
Sicinski
P
.
CDK4 and CDK6 kinases: from basic science to cancer therapy
.
Science
2022
;
375
:
eabc1495
.
9.
Goel
S
,
Bergholz
JS
,
Zhao
JJ
.
Targeting CDK4 and CDK6 in cancer
.
Nat Rev Cancer
2022
;
22
:
356
72
.
10.
Klein
ME
,
Kovatcheva
M
,
Davis
LE
,
Tap
WD
.
Koff
A
.
CDK4/6 inhibitors: the mechanism of action may not be as simple as once thought
.
Cancer Cell
2018
;
34
:
9
20
.
11.
Acosta
JC
,
Banito
A
,
Wuestefeld
T
,
Georgilis
A
,
Janich
P
,
Morton
JP
., et al
.
A complex secretory program orchestrated by the inflammasome controls paracrine senescence
.
Nat Cell Biol
2013
;
15
:
978
90
.
12.
Kuilman
T
,
Peeper
DS
.
Senescence-messaging secretome: SMS-ing cellular stress
.
Nat Rev Cancer
2009
;
9
:
81
94
.
13.
Pribluda
A
,
Elyada
E
,
Wiener
Z
,
Hamza
H
,
Goldstein
RE
,
Biton
M
., et al
.
A senescence-inflammatory switch from cancer-inhibitory to cancer-promoting mechanism
.
Cancer Cell
2013
;
24
:
242
56
.
14.
Park
MH
,
Choi
JE
,
Kim
J-R
,
Bae
YK
.
Immunohistochemical expressions of senescence-associated secretory phenotype and its association with immune microenvironments and clinicopathological factors in invasive breast cancer
.
Pathol Oncol Res
2021
;
27
:
1609795
.
15.
Charles
A
,
Bourne
CM
,
Korontsvit
T
,
Aretz
ZEH
,
Mun
SS
,
Dao
T
., et al
.
Low-dose CDK4/6 inhibitors induce presentation of pathway specific MHC ligands as potential targets for cancer immunotherapy
.
Oncoimmunology
2021
;
10
:
1916243
.
16.
Goel
S
,
DeCristo
MJ
,
Watt
AC
,
BrinJones
H
,
Sceneay
J
,
Li
BB
., et al
.
CDK4/6 inhibition triggers anti-tumour immunity
.
Nature
2017
;
548
:
471
5
.
17.
Sikora
E
,
Bielak-Zmijewska
A
,
Mosieniak
G
.
A common signature of cellular senescence; does it exist?
Ageing Res Rev
2021
;
71
:
101458
.
18.
Arora
S
,
Thompson
PJ
,
Wang
Y
,
Bhattacharyya
A
,
Apostolopoulou
H
,
Hatano
R
., et al
.
Invariant natural killer T cells coordinate removal of senescent cells
.
Med
2021
;
2
:
938
50
.
19.
Binet
F
,
Cagnone
G
,
Crespo-Garcia
S
,
Hata
M
,
Neault
M
,
Dejda
A
, et al
.
Neutrophil extracellular traps target senescent vasculature for tissue remodeling in retinopathy
.
Science
2020
;
369
:
eaay5356
.
20.
Kang
T-W
,
Yevsa
T
,
Woller
N
,
Hoenicke
L
,
Wuestefeld
T
,
Dauch
D
., et al
.
Senescence surveillance of pre-malignant hepatocytes limits liver cancer development
.
Nature
2011
;
479
:
547
51
.
21.
Krizhanovsky
V
,
Xue
W
,
Zender
L
,
Yon
M
,
Hernando
E
,
Lowe
SW
.
Implications of cellular senescence in tissue damage response, tumor suppression, and stem cell biology
.
Cold Spring Harb Symp Quant Biol
2008
;
73
:
513
22
.
22.
Marin
I
,
Boix
O
,
Garcia-Garijo
A
,
Sirois
I
,
Caballe
A
,
Zarzuela
E
., et al
.
Cellular senescence is immunogenic and promotes antitumor immunity
.
Cancer Discov
2023
;
13
:
410
31
.
23.
Pereira
BI
,
Devine
OP
,
Vukmanovic-Stejic
M
,
Chambers
ES
,
Subramanian
P
,
Patel
N
., et al
.
Senescent cells evade immune clearance via HLA-E-mediated NK and CD8(+) T cell inhibition
.
Nat Commun
2019
;
10
:
2387
.
24.
Prata
LGPL
,
Ovsyannikova
IG
,
Tchkonia
T
,
Kirkland
JL
.
Senescent cell clearance by the immune system: emerging therapeutic opportunities
.
Semin Immunol
2018
;
40
:
101275
.
25.
Ruscetti
M
,
Leibold
J
,
Bott
MJ
,
Fennell
M
,
Kulick
A
,
Salgado
NR
., et al
.
NK cell-mediated cytotoxicity contributes to tumor control by a cytostatic drug combination
.
Science
2018
;
362
:
1416
22
.
26.
Xue
W
,
Zender
L
,
Miething
C
,
Dickins
RA
,
Hernando
E
,
Krizhanovsky
V
., et al
.
Senescence and tumour clearance is triggered by p53 restoration in murine liver carcinomas
.
Nature
2007
;
445
:
656
60
.
27.
Orjalo
AV
,
Bhaumik
D
,
Gengler
BK
,
Scott
GK
,
Campisi
J
.
Cell surface-bound IL-1alpha is an upstream regulator of the senescence-associated IL-6/IL-8 cytokine network
.
Proc Natl Acad Sci USA
2009
;
106
:
17031
6
.
28.
Acosta
JC
,
O'Loghlen
A
,
Banito
A
,
Guijarro
MV
,
Augert
A
,
Raguz
S
., et al
.
Chemokine signaling via the CXCR2 receptor reinforces senescence
.
Cell
2008
;
133
:
1006
18
.
29.
Kumari
R
,
Jat
P
.
Mechanisms of cellular senescence: cell cycle arrest and senescence associated secretory phenotype
.
Front Cell Dev Biol
2021
;
9
:
645593
.
30.
Liu
XL
,
Ding
J
,
Meng
LH
.
Oncogene-induced senescence: a double edged sword in cancer
.
Acta Pharmacol Sin
2018
;
39
:
1553
8
.
31.
Wang
L
,
Lankhorst
L
,
Bernards
R
.
Exploiting senescence for the treatment of cancer
.
Nat Rev Cancer
2022
;
22
:
340
55
.
32.
Collado
M
,
Serrano
M
.
Senescence in tumours: evidence from mice and humans
.
Nat Rev Cancer
2010
;
10
:
51
7
.
33.
Gorgoulis
V
,
Adams
PD
,
Alimonti
A
,
Bennett
DC
,
Bischof
O
,
Bishop
C
., et al
.
Cellular senescence: defining a path forward
.
Cell
2019
;
179
:
813
27
.
34.
Prasanna
PG
,
Citrin
DE
,
Hildesheim
J
,
Ahmed
MM
,
Venkatachalam
S
,
Riscuta
G
., et al
.
Therapy-induced senescence: opportunities to improve anticancer therapy
.
J Natl Cancer Inst
2021
;
113
:
1285
98
.
35.
Sharpless
NE
,
Sherr
CJ
.
Forging a signature of in vivo senescence
.
Nat Rev Cancer
2015
;
15
:
397
408
.
36.
Wagner
V
,
Gil
J
.
Senescence as a therapeutically relevant response to CDK4/6 inhibitors
.
Oncogene
2020
;
39
:
5165
76
.
37.
Barretina
J
,
Taylor
BS
,
Banerji
S
,
Ramos
AH
,
Lagos-Quintana
M
,
DeCarolis
PL
., et al
.
Subtype-specific genomic alterations define new targets for soft-tissue sarcoma therapy
.
Nat Genet
2010
;
42
:
715
21
.
38.
Takebayashi
S‐I
,
Tanaka
H
,
Hino
S
,
Nakatsu
Y
,
Igata
T
,
Sakamoto
A
., et al
.
Retinoblastoma protein promotes oxidative phosphorylation through upregulation of glycolytic genes in oncogene-induced senescent cells
.
Aging Cell
2015
;
14
:
689
97
.
39.
Chen
H
,
Ruiz
PD
,
McKimpson
WM
,
Novikov
L
,
Kitsis
RN
,
Gamble
MJ
.
MacroH2A1 and ATM play opposing roles in paracrine senescence and the senescence-associated secretory phenotype
.
Mol Cell
2015
;
59
:
719
31
.
40.
Coppé
J-P
,
Patil
CK
,
Rodier
F
,
Krtolica
A
,
Beauséjour
CM
,
Parrinello
S
., et al
.
A human-like senescence-associated secretory phenotype is conserved in mouse cells dependent on physiological oxygen
.
PLoS One
2010
;
5
:
e9188
.
41.
Coppé
J-P
,
Patil
CK
,
Rodier
F
,
Sun
Y
,
Muñoz
DP
,
Goldstein
J
., et al
.
Senescence-associated secretory phenotypes reveal cell-nonautonomous functions of oncogenic RAS and the p53 tumor suppressor
.
PLoS Biol
2008
;
6
:
2853
68
.
42.
Kuilman
T
,
Michaloglou
C
,
Vredeveld
LCW
,
Douma
S
,
van Doorn
R
,
Desmet
CJ
., et al
.
Oncogene-induced senescence relayed by an interleukin-dependent inflammatory network
.
Cell
2008
;
133
:
1019
31
.
43.
Lackner
DH
,
Hayashi
MT
,
Cesare
AJ
,
Karlseder
J
.
A genomics approach identifies senescence-specific gene expression regulation
.
Aging Cell
2014
;
13
:
946
50
.
44.
Özcan
S
,
Alessio
N
,
Acar
MB
,
Mert
E
,
Omerli
F
,
Peluso
G
., et al
.
Unbiased analysis of senescence associated secretory phenotype (SASP) to identify common components following different genotoxic stresses
.
Aging (Albany NY)
2016
;
8
:
1316
29
.
45.
Hafemeister
C
,
Satija
R
.
Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression
.
Genome Biol
2019
;
20
:
296
.
46.
Cao
J
,
Spielmann
M
,
Qiu
X
,
Huang
X
,
Ibrahim
DM
,
Hill
AJ
., et al
.
The single-cell transcriptional landscape of mammalian organogenesis
.
Nature
2019
;
566
:
496
502
.
47.
Ritchie
ME
,
Phipson
B
,
Wu
D
,
Hu
Y
,
Law
CW
,
Shi
W
., et al
.
limma powers differential expression analyses for RNA-sequencing and microarray studies
.
Nucleic Acids Res
2015
;
43
:
e47
.
48.
Bellera
CA
,
MacGrogan
G
,
Debled
M
,
de Lara
CT
,
Brouste
V
,
Mathoulin-Pélissier
S
.
Variables with time-varying effects and the Cox model: some statistical concepts illustrated with a prognostic factor study in breast cancer
.
BMC Med Res Methodol
2010
;
10
:
20
.
49.
Chen
Y
,
Shao
Z
,
Chen
W
,
Xie
H
,
Wu
Z
,
Qin
G
., et al
.
A varying-coefficient cox model for the effect of CA19–9 kinetics on overall survival in patients with advanced pancreatic cancer
.
Oncotarget
2017
;
8
:
29925
34
.
50.
Yu
Y
,
Carey
M
,
Pollett
W
,
Green
J
,
Dicks
E
,
Parfrey
P
., et al
.
The long-term survival characteristics of a cohort of colorectal cancer patients and baseline variables associated with survival outcomes with or without time-varying effects
.
BMC Med
2019
;
17
:
150
.
51.
Kanojia
D
,
Nagata
Y
,
Garg
M
,
Lee
DH
,
Sato
A
,
Yoshida
K
., et al
.
Genomic landscape of liposarcoma
.
Oncotarget
2015
;
6
:
42429
44
.
52.
Chan
M
,
Yuan
H
,
Soifer
I
,
Maile
TM
,
Wang
RY
,
Ireland
A
., et al
.
Novel insights from a multiomics dissection of the hayflick limit
.
eLife
2022
;
11
:
e70283
.
53.
Wiley
CD
,
Campisi
J
.
The metabolic roots of senescence: mechanisms and opportunities for intervention
.
Nat Metab
2021
;
3
:
1290
301
.
54.
Gabai
Y
,
Assouline
B
,
Ben-Porath
I
.
Senescent stromal cells: roles in the tumor microenvironment
.
Trends Cancer
2023
;
9
:
28
41
.
55.
La Paglia
L
,
Listì
A
,
Caruso
S
,
Amodeo
V
,
Passiglia
F
,
Bazan
V
., et al
.
Potential role of ANGPTL4 in the cross talk between metabolism and cancer through PPAR signaling pathway
.
PPAR Res
2017
;
2017
:
8187235
.
56.
Marin
I
,
Serrano
M
,
Pietrocola
F
.
Cellular senescence enhances adaptive anticancer immunosurveillance
.
Oncoimmunology
2023
;
12
:
2154115
.
57.
Dancsok
AR
,
Gao
D
,
Lee
AF
,
Steigen
SE
,
Blay
J-Y
,
Thomas
DM
., et al
.
Tumor-associated macrophages and macrophage-related immune checkpoint expression in sarcomas
.
Oncoimmunology
2020
;
9
:
1747340
.
58.
Ciznadija
D
,
Liu
Y
,
Pyonteck
SM
,
Holland
EC
,
Koff
A
.
Cyclin D1 and cdk4 mediate development of neurologically destructive oligodendroglioma
.
Cancer Res
2011
;
71
:
6174
83
.
59.
Pawelec
G
,
Goldeck
D
,
Derhovanessian
E
.
Inflammation, ageing and chronic disease
.
Curr Opin Immunol
2014
;
29
:
23
8
.
60.
Tchkonia
T
,
Zhu
Y
,
van Deursen
J
,
Campisi
J
,
Kirkland
JL
.
Cellular senescence and the senescent secretory phenotype: therapeutic opportunities
.
J Clin Invest
2013
;
123
:
966
72
.
61.
Eggert
T
,
Wolter
K
,
Ji
J
,
Ma
C
,
Yevsa
T
,
Klotz
S
., et al
.
Distinct functions of senescence-associated immune responses in liver tumor surveillance and tumor progression
.
Cancer Cell
2016
;
30
:
533
47
.
62.
Hafner
M
,
Mills
CE
,
Subramanian
K
,
Chen
C
,
Chung
M
,
Boswell
SA
., et al
.
Multiomics profiling establishes the polypharmacology of FDA-approved CDK4/6 inhibitors and the potential for differential clinical activity
.
Cell Chem Biol
2019
;
26
:
1067
80
.
63.
Kim
S
,
Tiedt
R
,
Loo
A
,
Horn
T
,
Delach
S
,
Kovats
S
., et al
.
The potent and selective cyclin-dependent kinases 4 and 6 inhibitor ribociclib (LEE011) is a versatile combination partner in preclinical cancer models
.
Oncotarget
2018
;
9
:
35226
40
.
64.
Braal
CL
,
Jongbloed
EM
,
Wilting
SM
,
Mathijssen
RHJ
,
Koolen
SLW
,
Jager
A
.
Inhibiting CDK4/6 in breast cancer with palbociclib, ribociclib, and abemaciclib: similarities and differences
.
Drugs
2021
;
81
:
317
31
.
65.
Imai
Y
,
Takahashi
A
,
Hanyu
A
,
Hori
S
,
Sato
S
,
Naka
K
., et al
.
Crosstalk between the Rb pathway and AKT signaling forms a quiescence-senescence switch
.
Cell Rep
2014
;
7
:
194
207
.
66.
Kortlever
RM
,
Bernards
R
.,
Senescence, wound healing and cancer: the PAI-1 connection
.
Cell Cycle
2006
;
5
:
2697
703
.
67.
Tremain
R
,
Marko
M
,
Kinnimulki
V
,
Ueno
H
,
Bottinger
E
,
Glick
A
.
Defects in TGF-beta signaling overcome senescence of mouse keratinocytes expressing v-Ha-ras
.
Oncogene
2000
;
19
:
1698
709
.
68.
Wajapeyee
N
,
Serra
RW
,
Zhu
X
,
Mahalingam
M
,
Green
MR
.
Role for IGFBP7 in senescence induction by BRAF
.
Cell
2010
;
141
:
746
7
.
69.
Berry
DA
.
Interim analyses in clinical trials: classical vs. Bayesian approaches
.
Stat Med
1985
;
4
:
521
6
.
70.
Kidwell
KM
,
Roychoudhury
S
,
Wendelberger
B
,
Scott
J
,
Moroz
T
,
Yin
S
., et al
.
Application of Bayesian methods to accelerate rare disease drug development: scopes and hurdles
.
Orphanet J Rare Dis
2022
;
17
:
186
.
71.
Lee
JJ
,
Chu
CT
.
Bayesian clinical trials in action
.
Stat Med
2012
;
31
:
2955
72
.
72.
Tudur Smith
C
,
Williamson
PR
,
Beresford
MW
.,
Methodology of clinical trials for rare diseases
.
Best Pract Res Clin Rheumatol
2014
;
28
:
247
62
.
This open access article is distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.

Supplementary data