CDK4/6 inhibitors: The mechanism of action may not be as simple as once thought

Abstract

CDK4/6 inhibitors are among a new generation of therapeutics. Building upon the striking success of the combination of CDK4/6 inhibitors and the hormone receptor antagonist letrozole in breast cancer, many other combinations have recently entered clinical trials in multiple diseases. To achieve maximal benefit with CDK4/6 inhibitors it will be critical to understand the cellular mechanisms by which they act. Here we highlight the mechanisms by which CDK4/6 inhibitors can exert their anti-tumor activities beyond simply enforcing cytostatic growth arrest and discuss how this knowledge may inform new combinations, improve outcomes and modify dosing schedules in the future.

THE IMPORTANCE OF CYCLIN D-CDK4/6 IN THE CELL CYCLE

With the war on cancer declared almost fifty years ago, a mechanistic understanding of the processes underlying the duplication and segregation of DNA was long considered a useful area for the identification of druggable targets for therapy. Nearly twenty years later the cyclin/cyclin-dependent kinase (CDK) holoenzyme was identified as a key driver of a number of cell cycle transitions and the first non-specific CDK inhibitors made it to the clinic over the next ten years. Such pan-CDK inhibitors had limited success, due in part to the dose limiting toxicities that hampered their use (Asghar et al., 2015). Today, far more specific inhibitors that target CDK4 and CDK6 (palbociclib, abemaciclib and ribociclib) exist. These have more limited toxicities, which allows for their broad use to treat a variety of neoplasms. Currently CDK4/6 inhibitors are being used both as single agents, and in a vast number of clinical trials evaluating their efficacy when combined with signaling pathway inhibitors. Thus, it is crucial to understand how they are exerting their anti-tumorigenic effects.

CDK4/6 activity bridges numerous extracellular signaling pathways to the cell cycle (Sherr and Roberts, 1999). Both non-immortal non-transformed cells (hereafter referred to as ‘normal’ cells) and many transformed tumor cells commit irreversibly to the mitotic cell cycle in the G₁ phase. Commitment depends on the phosphorylation and inactivation of the retinoblastoma tumor suppressor protein, Rb. The growth suppressive properties of Rb are largely, but not completely, associated with its binding to the transcription factor E2F and repressing transcription at target promoters (Classon and Harlow, 2002; Harbour and Dean, 2000; Stevaux and Dyson, 2002). Phosphorylation of Rb destabilizes its interaction with E2F and other transcriptional regulators. In normal cells, the phosphorylation of Rb is typically carried out by the sequential actions of the CDK4 or CDK6 kinases in complex with a positive regulatory D-type cyclin subunit, followed by cyclin E/CDK2 complexes (Harbour et al., 1999; Lundberg and Weinberg, 1998). Additionally, extracellular signals regulate the expression of cyclins and CDK inhibitors, like p16^Ink4a, p21^Cip1, and p27^Kip1, the first of which inhibits the CDK4/6 kinases whereas the latter two inhibit the CDK2 kinases (Sherr and Roberts, 1999).

In virtually all human cancer cells, this circuit is dysregulated by either overexpression of cyclin D1, loss of p16^Ink4a, the mutation of CDK4 to an Ink4-refractory state, or the loss of Rb itself (Classon and Harlow, 2002). This not only affects how the cancer cell responds to extracellular signals, it can also affect the requirement for sequential ordered phosphorylation by the CDKs during the inactivation of Rb. Thus, in some cells, CDK2 may be dispensable, or compensated for by other CDKs.

Although Rb is the primary cell cycle target of CDK4/6, Rb and other proteins that control the commitment decision have non-cell cycle related roles (Besson et al., 2008; Denicourt and Dowdy, 2004). CDK4/6 kinases can phosphorylate factors involved in cell differentiation affecting their transcriptional activity, apoptotic factors affecting their activity, and other factors that can directly affect mitochondrial activity (Hydbring et al., 2016; Lim and Kaldis, 2013). Where applicable we include potential non-Rb targets that could be participating in the immunological, senescence promoting, and metabolic outcomes associated with these drugs. However, our focus in this Perspective is on the outcome of inhibiting the CDK4/6 kinases in Rb-proficient tumors and we encourage readers with an interest in alternative substrates and interactions to seek out additional reviews.

CDK4/6 INHIBITORS: A TRIO OF COMPOUNDS WITH DISTINCT ADVANTAGES

Although palbociclib was the first CDK4/6 inhibitor to demonstrate clinical efficacy (Finn et al., 2016), there are two others that soon followed. Ribociclib is structurally very similar to palbociclib, and abemaciclib is significantly less similar to either one (Table 1). In vitro studies using cyclin D1/CDK4 and various cyclin D/CDK6 kinases determined that both abemaciclib and ribociclib are more potent against CDK4 than CDK6 (Gelbert et al., 2014; Tripathy et al., 2017) (Table 1). Palbociclib on the other hand has similar potency when comparing its activity on cyclin D1/CDK4 and cyclin D2/CDK6 (Fry et al., 2004). In such assays, abemaciclib also has modest activity, relative to its CDK4 inhibitory activity, against cyclin T1/CDK9, cyclin E2/CDK2, p25/CDK5 and p35/CDK5 (Gelbert et al., 2014) (Table 2). However, the remarkable specificity of all of these drugs to inhibit the proliferation of Rb-positive tumor cells but not Rb-negative tumor cells, suggests that differences in the in vitro profiles might not contribute that much to their in vivo activity.

Table 1.

Drug Characteristics of CDK4/6 inhibitors

	Structure	IC50in cell free assay	Cmax(nM)	tmax(hr)	t1/2(hr)	Kp,uu in mouse models	Toxicities in phase 3 trials	Dosing schedule
Palbociclib(PD0332991)		CDK4 (11 nM) CDK6 (15 nM)	200–260	4–8	28	0.01	Neutropenia	125 mg PO daily for 21 out of every 28 days (in combination with hormone therapy)
Ribociclib(LEE011)		CDK4 (10 nM) CDK6 (39 nM)	4,000–7,000	2–5	30–50	0.12	Neutropenia	600 mg PO daily for 21 out of every 28 days (in combination with hormone therapy)
Abemaciclib(LY2835219)		CDK4 (2 nM) CDK6 (9.9 nM)	500–600	4	NR (21 hr for a single dose)	0.03	GI distress, neutropenia (not dose-limiting)	200 mg PO daily continuously(as a monotherapy)

NR = not reported

Table 2.

IC₅₀ of CDK4/6 inhibitors in cell free assays.

	IC50(nM) by cell free assay
CDK family kinase complex	Palbociclib	Ribociclib	Abemaciclib
CDK4/CyclinD1	11	8	2
CDK4/CyclinD3	9	NR	NR
CDK6/CyclinD1	NR	NR	9.9
CDK6/CyclinD2	15	NR	NR
CDK6/CyclinD3	NR	39	NR
CDK1/CyclinB	>10 µM	>1.5 µM	1,627
CDK2/CyclinA	>10 µM	>1.5 µM	NR
CDK2/CyclinE2	>10 µM	>1.5 µM	504
CDK5/p25	>10 µM	>1.5 µM	355
CDK5/p35	NR	>1.5 µM	287
CDK7/CyclinH1	NR	>1.5 µM	3,910
CDK9/CyclinT1	NR	1510	57
References	Fry et al 2004	Tripathy etal 2017	Gelbert etal 2014

NR = not reported

All three CDK4/6 inhibitors are orally available, but each has differing pharmacokinetics and clinical toxicities (Table 1), necessitating different dosing strategies. Both palbociclib and ribociclib are dosed once daily whereas abemaciclib is dosed twice daily. Ribociclib is notable for achieving high maximum plasma concentrations (exceeding 2 µg/ml) with a long half-life (greater than 30 hours). This may translate to higher cerebrospinal fluid concentrations for ribocliclib compared to palbociclib and abemaciclib as noted in mouse models (DiPippo et al., 2016; Raub et al., 2015; Yin et al., 2017). Other rodent models indicate a strong efflux of palbociclib out of the central nervous system, and less so for abemaciclib (Raub et al., 2015). In the few human patients receiving abemaciclib, in which drug levels in the central nervous system were measured, accumulations in the range of 2.2–14.7 nmol/L were achieved (Patnaik et al., 2016).

There are marked differences in the toxicity profiles of the inhibitors for reasons that are not completely clear. Grade 3–4 neutropenia is observed in approximately 60% of patients taking palbociclib and ribociclib (Asghar et al., 2015; Hortobagyi et al., 2016). Abemaciclib appears to be better tolerated overall, with only 55% of patients experiencing significant adverse events (as compared to 70–80% with ribociclib and palbociclib) and only 21% with grade 3–4 neutropenia. However, 10% of the patients treated with abemaciclib develop grade 3 diarrhea, which is very rare with the other two inhibitors (Asghar et al., 2015; Hortobagyi et al., 2016). Unraveling the toxicities and overcoming them may lead to a better understanding of these drugs and their biologic availability.

Due to the significant myelotoxicity of palbociclib and ribociclib, both drugs require dose-interruption and are administered on a three-weeks-on / one-week-off schedule to allow marrow recovery. In contrast, abemaciclib is dosed continuously. When considering the cytostatic effect of CDK4/6 inhibitors, interrupted dosing can provide an opportunity for potent synergy with combination therapies dedicated to interfering with other cell cycle effects. For example, DNA damaging agents typically require active cell cycle progression. While co-administration of CDK4/6 inhibitors has been used in vivo to “protect” healthy cells from the toxic side effects of chemotherapy (He et al., 2017), interrupted dosing may provide an additional opportunity to use cytotoxic DNA damaging and other cell cycle targeting therapies as the cells become somewhat synchronized upon release, resulting in increased susceptibility to those compounds (Francis et al., 2017; Huang et al., 2012; Yang et al., 2015). Such alternative dosing strategies are being investigated in early phase multi-agent clinical trials in a variety of different tumor types including breast cancer, leukemia and soft tissue sarcomas (Table 3).

Table 3.

List of ongoing clinical trials with CDK4/6 inhibitors in combination with one or more other therapies

	Combination	Dosing schedule	Disease	Phase	Identifier
Palbociclib	Trastuzumab-DM1	Palbociclib days 5–18 (21 day cycle)	HER2+ breast cancer	Ib	NCT1976169
	(HER2 antibody)	Trastuzumab day 1
	Tucatinib (HER2i)	Palbociclib days 1–21 (28 day cycle)	HR+, HER2+ breast cancer	Ib/II	NCT03054363
	+ Letrozole (aromatasei)	Letrozole and Tucatinib days 1–28
	Anastrozole(aromatasei)	Palbociclib days 1–21 (28 day cycle)	HR+, HER2+ breast cancer	I/II	NCT03304080
	+ Trastuzumab	Anastrozole days 1–28
	+Pertuzumab (HER2i)	Trastuzumab and Pertuzumab once every 21 days
	Baxedoxifene (ER modulator)	Not stated	HR+ breast cancer	Ib/II	NCT02448771
	SAR439859 (ER degrader)	Palbociclib days 1–21 (28 day cycle)	ER+ breast cancer	I/II	NCT03284957
		SAR439859 days 1–28
	GDC-0810 (ER downregulator)	Palbociclib days 1–21 (28 day cycle)	ER+/HER2− breast cancer	I/II	NCT01823835
		GDC-0810 days 1–28
	Gedatolisib (PI3K/mTORi)	Palbociclib days 1–21 (28 day cycle)	ER+/HER2− breast cancer	I	NCT02626507
	+Fulvestrant (ER antagonist)	Gedatolisib days 1, 7, 14, 21; Fulvestrant day 1
	Gedatolisib (PI3K/mTORi)	Palbociclib days 1–21 (28 day cycle)	Solid tumors	I	NCT03065062
		Gedatolisib days 1, 7, 14, and 21
	Copanlisib (PI3Ki) + Letrozole	Palbociclib days 1–21 (28 day cycle)	HR+, HER2− breast cancer	Ib/II	NCT03128619
		Copanlisib days 1, 8, and 15; Letrozole days 1–28
	GDC-0077 (PI3Ki) + Letrozole	Palbociclib days 1–21 (28 day cycle)	PIK3CA mutant, HR+, HER2− breast cancer	I/II	NCT03006172
		GDC-0077 and Letrozole days 1–28
	AZD2014 (mTORC1/2i)	Not stated	ER+ breast cancer	I/II	NCT02599714
	+Fulvestrant
	Everolimus (mTORi)	Palbociclib days 1–21 (28 day cycle)	ER+, HER2− breast cancer	Ib/IIa	NCT02871791
	+ Exemestane (aromatasei)	Everolimus and Exemestane days 1–28
	PD-0325901 (MEKi)	Palbociclib and PD-0325901 days 1–21 (28 day cycle)	KRAS mutant non-smallcell lung cancer, solid tumors	I/II	NCT02022982
	Binimetinib (MEKi)	Palbociclib days 1–21 (28 day cycle)	KRAS mutant non-small cell lung cancer	I/II	NCT03170206
		Binimetinib days 1–28
	Neratinib (pan-ERBBi)	Palbociclib and Neratinib days 1–21 (28 day cycle)	EGFR, HER2/3/4amplified/mutated advanced cancers	I	NCT03065387
	Ibrutinib (BTKi)	Palbociclib days 1–21 (28 day cycle)	Mantle cell lymphoma	I	NCT02159775
		Ibrutinib days 1–28
	Erdafitinib (FGFRi)	Palbociclib days 1–21 (28 day cycle)	ER+/HER2−/FGFRamplified breast cancer	1	NCT03238196
	+ Fulvestrant	Erdafitinib days 1–28; Fulvestrant day 1
	Cetuximab (EGFRi)	Palbociclib days 1–21 (28 day cycle)	Squamous cell carcinoma of the head and neck	II	NCT02499120
		Cetuximab once weekly
	Sorafenib (RTKi) or Decitabine or Dexamethasone	Palbociclib days 1–21 (28 day cycle);Sorafenib days 1–28	Relapsed and refractory leukemias	I	NCT03132454
		Palbociclib days 1–7 (28 day cycle);Decitabine days 8–12
		Palbociclib days 1–21 (28 day cycle);Dexamethasone days 1–4 and days 15–18
	Bicalutamide (anti-androgen)	Palbociclib days 1–21 (28 day cycle)	AR+ breast cancer	I/II	NCT02605486
		Bicalutamide days 1–28
	Anastrozole (aromatasei)	Palbociclib days 1–21 (28 day cycle)	HER2− breast cancer	II	NCT02942355
		Anastrozole days 1–28
	Tamoxifen (anti-mitotic)	Palbociclib days 1–21 (28 day cycle)	HR+, HER2− breast cancer	II	NCT02668666
		Tamoxifen days 1–28
Cisplatin or Carboplatin	Palbociclib days 2–22 (28 day cycle)	Advanced solid tumors	I	NCT02897375
	Cisplatin or carboplatin day 1
Carboplatin	Palbociclib days 1–14 (21 day cycle)	Squamous cell carcinoma of the head and neck	II	NCT03194373
	Carboplatin day 1
5-FU (nucleotide analog)	Palbociclib days 1–7 (14 day cycle)	Advanced solid tumors	I	NCT01522989
+ Oxaliplatin (platinum-based)	5-fu/Oxaliplatin day 8
Bortezomib (proteasomei)	Palbociclib days 1–12 (21 day cycle)	Mantle cell lymphoma	I	NCT01111188
	Bortezomib days 8,11,15,18
Paclitaxel (anti-mitotic)	Palbociclib days 1–21 (28 day cycle)	Pancreatic ductal adenocarcinoma	I	NCT02501902
	Paclitaxel days 1, 8, 15
Paclitaxel	Not stated	Advanced breast cancer	I	NCT01320592
Avelumab (anti PD-L1)	Palbociclib days 2–22 (28 day cycle)	ER+/HER2− metastatic breast cancer	II	NCT03147287
+ Fulvestrant	Avelumab once every two weeks; Fulvestrant day 1
Pembrolizumab (PD-1i)	Palbociclib days 1–21 (28 day cycle)	ER+, HER2− breast cancer	II	NCT02778685
+ Letrozole	Pembrolizumab every 21 days; Letrozole days 1–28
Ribociclib	Trastuzumab (HER2 antibody)	Ribociclib days 5–18 (21 day cycle);Trastuzumab day 1	HER2+ breast cancer	I/II	NCT02657343
	LSZ102 (ER degrader)	Not stated	ER+ breast cancer	I	NCT02734615
	Everolimus (mTORi)	Ribociclib days 1–21 (28 day cycle)	Pancreatic adenocarcinoma	I/II	NCT02985125
		Everolimus days 1–28
	Everolimus + Letrozole	All drugs days 1–28 (28 day cycle)	Endometrial cancer	II	NCT03008408
	Everolimus	Ribociclib days 1–21 (28 day cycle)	Dedifferentiated liposarcoma andLeiomyosarcoma	II	NCT03114527
		Everolimus days 1–28
	Everolimus	Ribociclib days 1–21 (28 day cycle)	Neuroendocrine tumors	II	NCT03070301
		Everolimus days 1–28
	Everolimus	Ribociclib days 1–21 (28 day cycle)	HR+, HER2− breast cancer	I	NCT01857193
	+ Exemestane (aromatasei)	Everolimus + Exemestane days 1–28
	BLY719 (PI3Ki) + Letrozole	Ribociclib days 1–21 (28 day cycle)	ER+ breast cancer	I	NCT01872260
		BLY719 + Letrozole days 1–28
	BLY719 or BKM120 (pan-PI3Ki)	Ribociclib days 1–21 (28 day cycle)	ER+/HER2− breast cancer	I/II	NCT02088684
	+ Fulvestrant	BLY719 or BKM120 days 1–28; Fulvestrant day 1
	Trametinib (MEKi)	Not stated	Advanced solid tumors	I/II	NCT02703571
	MEK162 (MEKi)	Ribociclib days 1–21 (28 day cycle)	NRAS mutant melanoma	Ib/II	NCT01781572
		MEK162 days 1–28
	LGX818 (RAFi) + MEK162	Ribociclib days 1–21 (28 day cycle)	BRAF dependent advanced solid tumors	I/II	NCT01543698
		LGX818 + MEK162 days 1–28
	EGF816 (EGFRi)	Not stated	EGFR mutant non-smallcell lung cancer	I	NCT03333343
	Ceritinib (ALKi)	Not stated	ALK positive non-small cell lung cancer	I	NCT02292550
	Enzalutamide (anti-androgen)	Ribociclib days 1–21 (28 day cycle);Enzalutamide days 1–28	Prostate Cancer	I/II	NCT02555189
	Bicalutamide (anti-androgen)	Ribociclib days 1–21 (28 day cycle)	AR+ triple negative breast cancer	I/II	NCT03090165
		Bicalutamide days 1–28
	Carboplatin	Ribociclib days 1–4, 8–11, 15–18 (28 day cycle)	Ovarian cancer	I	NCT03056833
	+ Paclitaxel (anti-mitotic)	Paclitaxel + carboplatin days 1, 8, 15
	Paclitaxel	Not stated	Advanced breast cancer	I	NCT02599363
	Doxorubicin	Ribociclib days 1–7 (21 day cycle)	Advanced soft tissue sarcoma	I	NCT03009201
		Doxorubicin day 10
	Tamoxifen (anti-mitotic)	Ribociclib days 1–21 (28 day cycle)	ER+, HER2− breastcancer	I	NCT02586675
		Tamoxifen days 1–28
	Gemcitabine(nucleotide analog)	Ribociclib days 8–14 (21 day cycle)	Advanced solid tumors	I	NCT03237390
		Gemcitabine days 1, 8
	Docetaxel (anti-mitotic)	Ribociclib days 2–14 (21 day cycle)	Prostate cancer	I/II	NCT02494921
	+ Prednisone	Docetaxel and Prednisone days 1–21
	PDR001 (anti-PD1 antibody)	Ribociclib days 1–21 (28 day cycle)	HR+, HER2− breast andovarian cancer	I	NCT03294694
	± Fulvestrant	PDR001 days 1–28
Abemaciclib	LY3023414 (PI3K/mTORi)	Not stated	Pancreatic ductal adenocarcinoma	II	NCT02981342
	LY3214996 (ERK1/2i)	Not stated	Advanced solid tumors	I	NCT02857270
	Ramucirumab (anti-VEGFR2)	Abemaciclib days 1–28 (28 day cycle)	Advanced solid tumors and lymphoma	I	NCT02745769
		Ramucirumab days 1, 15
	Xentuzumab (IGF1/2i)	Abemaciclib daily, Xentuzumab once a week	Advanced solid tumors, HR+ breast cancer	I	NCT03099174
	LY3039478 (Notchi)	Both drugs daily	Advanced solid tumors	Ib	NCT02787495
	Exemestane (aromatasei) or Exemestane + Everolimus or LY3032414 + Fulvestrant or Letrozole (aromatasei) or Anastrozole (aromatasei) or Tamoxifen (anti-mitotic) or Trastuzumab (HER2 antibody)	All drugs daily	Metastatic breast cancer	Ib	NCT02057133
	Anastrozole or Letrozole	All dugs daily	HR+, HER2− breast cancer	III	NCT02246621
	Tamoxifen	Both drugs daily	HR+, HER2− breast cancer	II	NCT02747004
	Premetrexed (anti-folate) or Gemcitabine or Ramucirumab or LY3023414 or Pembrolizumab (PD-1i)	Abemaciclib daily (21 day cycle)	Non-small cell lung cancer	I	NCT02079636
		Premetrexed day 1
		Gemcitabine days 1, 8
		Ramucirumab days 1, 8
		Pembrolizumab day 1
	LY3300054 (anti-PD-L1)	Abemaciclib daily (28 day cycle); LY3300054 days 1, 15	Advanced solid tumors	I	NCT02791334

In summary, all three inhibitors target the proliferative function of the cyclin D associated kinases in Rb positive tumor cells to induce cell cycle exit and they are largely inactive in Rb negative cells. While this suggests that Rb is the sole important substrate, it is conceivable that other substrates of these kinases contribute to phenotypes after the cells have exited the cell cycle. The key to improving combination therapies may lie in recognizing the consequences of growth arrest induced by CDK4/6 inhibitors and understanding how signaling pathways contribute to maintaining cells in a quiescent state.

CDK4/6 INHIBITORS REINFORCE CYTOSTASIS INDUCED BY SIGNALING PATHWAY INHIBITORS

Many early phase clinical trials are combining CDK4/6 inhibitors with signaling pathwaytargeted inhibitors (Table 3). In cell lines and xenografts, intrinsic and acquired resistance to signaling pathway inhibitors against estrogen, RAF, EGFR, PI3K, and others is sometimes associated with mutations in p16^Ink4a, upregulated expression of cyclin D1 or other D-type cyclins, or upregulation of CDK4 or CDK6 (Jiang et al., 2016; Long et al., 2014; Yadav et al., 2014). In some of these models, resistance can be overcome by including CDK4/6 inhibitors in the treatment (Finn et al., 2009; Goel et al., 2016; Kwong et al., 2012; Vora et al., 2014; Zhou et al., 2017). Thus, increased therapeutic efficiency can occur by enforcing a more durable cell cycle exit (Figure 1A) as has been extensively reviewed (for example(Sherr et al., 2016)).

Figure 1. The consequences of CDK4/6 inhibition.

The four cellular mechanisms, labeled A-D, that contribute to the efficacy of CDK4/6 inhibitors are shown. Details are in the accompanying text.

However, a number of recent investigations provide an alternative mechanistic explanation for the clinical activity of CDK4/6 inhibitors in combination with other drugs. Outcomes can depend on the nature of the cytostatic effect in tumor cells, vis a vis whether it undergoes a reversible quiescence or a more stable senescence. CDK4/6 inhibition can also alter cellular metabolism, depleting antioxidants, increasing reactive oxygen species (ROS) and triggering apoptosis. CDK4/6 inhibition can also affect both the maturation of sentinel cells of the immune system and the expansion of regulatory T-cells. These mechanisms are summarized in Figure 1, and how they might impact combinatorial cancer therapies in the future are discussed individually below.

SENESCENCE AFTER CDK4/6 INHIBITOR INDUCED GROWTH ARREST

Recently, a number of groups have become interested in the decisions that cells make when they exit from the G₁ phase of the cell cycle into quiescence. Depending on the cell type and the transforming event, some Rb-positive cells undergo quiescence and others undergo senescence when treated with CDK4/6 inhibitors (Baughn et al., 2006; Choi et al., 2012; Kovatcheva et al., 2015; Michaud et al., 2010; Puyol et al., 2010). Unlike quiescent cells, senescent cells will not return to the cell cycle following removal of the inducing signal and are generally refractory to other proliferation-inducing signals (Rodier and Campisi, 2011). This outcome may be an important consideration when deciding whether to treat with abemaciclib or palbociclib/ribociclib, given their differing dosing schedules.

While exploring the effects of palbociclib and other CDK4/6 inhibitors, a decision point was identified at which quiescent cells decide whether or not to progress to senescence (Figure 1B). The transition from quiescence into senescence has been termed geroconversion and can also be described as senescence after growth arrest (SAGA). Which outcome is achieved depends on a cell-type intrinsic program that is activated following the withdrawal of the cell from the cell cycle. Specifically, down-regulation of MDM2, redistribution of the chromatin remodeling enzyme ATRX, and repression of HRAS transcription are necessary for the transition of CDK4 inhibitor-induced quiescence into senescence in a number of mesenchymal and epithelial cell lines derived from different tumor types, including breast, non-small cell lung cancer, soft tissue sarcoma and glioma (Kovatcheva et al., 2017; Kovatcheva et al., 2015).

In addition to creating a permissive senescence environment by fulfilling the requisite of cell cycle exit, CDK4/6 inhibition may elicit senescence via alternative CDK4/6 substrates. CDK4/6 inhibition leads to the loss of multi-site phosphorylation and the destabilization of the transcription factor Forkhead Box M1 (FOXM1) (Anders et al., 2011). FOXM1 suppression following CDK4/6 inhibition leads to the accumulation of reactive oxygen species and senescence in transformed melanomas, but not in normal melanocytes. Nevertheless, knockdown of FOXM1 does not completely recapitulate the senescence effect of CDK4/6 inhibition, suggesting CDK4/6 inhibitors have functions in addition to suppression of FOXM1. Recent work by others has shown that the induction of senescence after CDK4/6 inhibition is Rb-dependent in melanoma (Yoshida et al., 2016), thus raising the question of whether FOXM1 suppression requires a cell that has first been growth arrested, or if there are further context-specific effects of this pathway that need to be understood before it is targeted therapeutically.

It has also been suggested that the mammalian target of rapamycin (mTOR) can toggle the decision between quiescence and senescence (Korotchkina et al., 2010; Leontieva et al., 2012). This may be cell-type specific with the contextual clues not yet understood. For example, growth inhibition by over-expression of the CDK inhibitor p21 in the context of active mTOR signaling has been described to lead to a “futile” period of cell growth, ultimately leading to senescence (Korotchkina et al., 2010). Inhibition of mTOR indirectly following treatment with nutlin-3 or directly by treatment with rapamycin drives the cells into a quiescent state. In contrast to this however, mTOR inhibition is crucial to reach oncogene induced senescence in mouse models of melanoma, and mTOR inhibition can cooperate with CDK4/6 inhibition to drive senescence in melanoma cell lines (Damsky et al., 2015; Yoshida et al., 2016). As several mitogentic pathways – all frequently hyper-activated in cancer – impinge on mTOR activity, it may be valuable to further investigate how combining inhibitors of these pathways with CDK4/6 inhibitors affects mTOR activity and senescence and what other pathways can affect this outcome.

Collectively, these studies suggest that the cooperation of CDK4/6 inhibitors and signaling pathway inhibitors may be affecting sequential decisions in the tumor cell. Inhibitors that have minimal effect in cycling tumor cells may have more of an impact in non-cycling CDK4/6 inhibitor treated cells, perhaps pushing them into senescence. Indeed, quiescent CDK4/6 inhibitor arrested liposarcoma tumor cells are pushed into senescence upon knocking down HRAS, and enforced expression of near physiologic levels of HRAS mRNA is able to prevent cells from undergoing senescence following treatment with CDK4/6 inhibitors albeit they still exit the cell cycle (Kovatcheva 2017). Identifying the proteins that control the transition from quiescence into senescence may nominate new targets for combinatorial drug therapy and changing dosing schedules where cells are first induced to exit the cell cycle and are then treated with a second drug targeting such signaling events may even provide a therapeutic gain by minimizing the dose necessary and the subsequent toxicity associated with such inhibitors.

The consequences of driving a cell into senescence after growth arrest are particularly interesting. Senescent cells express a cell type and signal specific program of gene expression and cytokine secretion (the SASP), which can sculpt the immune response (Ohtani et al., 2012). On one hand, the SASP can induce the recruitment of immune cells that will mediate tumor clearance (Xue et al., 2007), or promote paracrine senescence (Acosta et al., 2013; Acosta et al., 2008). On the other hand, the SASP can create a pro-tumorigenic environment (Coppe et al., 2010; Krtolica et al., 2001; Ruhland et al., 2016). Consequently, the SASP is often referred to as a “double-edged sword” and a better understanding of this biology will be paramount for its clinical manipulation.

Furthermore, the SASP can induce cellular plasticity (Ritschka et al., 2017), and cancer cells that manage to escape senescence have a more aggressive, cancer stem cell-like identity (Milanovic et al., 2018). Indeed, persistent senescent cells may contribute to detrimental short and long-term side effects of treatment, as well as to metastases and relapse (Demaria et al., 2017). Collectively, these observations suggest that the most effective approach to harness the senescence promoting effects of CDK4/6 inhibitors may be to first induce senescence and then eliminate the persistent senescent cells. Evidence that BCL2 inhibitors can directly eliminate senescent cells (Chang et al., 2016; Yosef et al., 2016) suggests that combining this with other cancer therapies in a sequential manner might be useful. Future proteomic, transcriptomic, and metabolomic data from senescent cells holds promise to identify additional vulnerabilities.

THE IMPACT OF CDK4/6 INHIBITORS ON CELLULAR METABOLISM

It has long been recognized that cell division is coordinated with metabolic state. A number of non-Rb targets for CDK4/6 have been identified in the metabolic machinery. For example, phosphorylation of AMPKα2 by CDK4 is associated with increased glycolysis and decreased fatty acid oxidation in MEFS (Lopez-Mejia et al., 2017). In contrast, CDK4 phosphorylation of GCN5 can lead to acetylation of PGC-1α and is associated with decreased glucose metabolism in hepatic cells (Lee et al., 2014).

There is some evidence that manipulating metabolic pathways may be a useful addition of CDK4/6 inhibition (Figure 1C). Inhibition of CDK4/6 with any of the three drugs in pancreatic ductal adenocarcinoma cell lines alters glycolytic and oxidative metabolism leading to an increase in ROS in an Rb-dependent manner (Franco et al., 2016). Combining CDK4/6 inhibition with an mTOR inhibitor, a BCL2 inhibitor, or reducing the ROS scavengers drives these cells into apoptosis, whereas treatment with single agents alone is not sufficient. Interestingly, combined treatment with a MEK inhibitor uniquely drove these cells into senescence.

In T-cell acute lymphoblastic leukemia (T-ALL), inhibition of CDK6 or genetic repression of cyclin D3 induces apoptosis (Choi et al., 2012; Sawai et al., 2012). These cells express very low levels of cyclin D1, cyclin D2, and CDK4, and the proliferative decision is dependent upon cyclin D3 and CDK6 (Wang et al., 2017; Wolowiec et al., 1996). The cyclin D3/CDK6 kinase complex can phosphorylate 6-phosphofructokinase and pyruvate kinase M2 (Wang et al., 2017). This has the effect of pushing glycolytic intermediates into the pentose phosphate and serine pathways, and inhibition of CDK6 (through treatment with palbociclib, ribociclib, or knockdown of CDK6) depletes the antioxidants NADPH and glutathione, increasing ROS and apoptosis. It is likely that the changes in metabolism after CDK4/6 inhibition will be context-specific and may be dependent on the oncogenic drivers that set up unique metabolic pathways and vulnerabilities.

CDK4/6 IN THE TUMOR MICROENVIRONMENT

Our understanding of the relationship between cancer cells and the supporting cells that create a tumor microenvironment has significantly advanced during the last 10 years. The extraordinarily complex microenvironment is not only supportive for growth but can also drive the transition of slow growing, indolent tumors into a more aggressive state.

The importance of CDK4 activity within the microenvironment was first demonstrated by crossing an RCAS-PDGF/nestin-TvA mouse model of oligodendroglioma into a CDK4 deficient background (Ciznadija et al., 2011). CDK4 is required for the proliferation of the tumor cells; however, reconstituting incipient CDK4 deficient tumor cells with CDK4 expression vectors is not sufficient for tumors to progress to a more aggressive state when the rest of the animal is CDK4 deficient. This suggests there are tumor extrinsic roles for CDK4. The lack of progression in this model is associated with a defect in the maturation of tumor-associated microglia, which remain in a sentinel mode in the absence of CDK4. This is consistent with findings that the maturation of microglia from sentinel to reactive supports the progression of oligodendroglioma into its more aggressive form (Ghosh and Chaudhuri, 2010). Further reinforcing the notion that this is a cyclin D/CDK4 dependent phenomenon, similar observations were made in cyclin D1 deficient mice. Thus, CDK4 is required for both the proliferation of tumor cells and for the maturation of the tumor microenvironment, and both are necessary for the progression of disease in this model.

Recently, in a variety of breast cancer models, including patient-derived xenografts and an MMTV-HER2 mouse, it was demonstrated that abemaciclib or palbociclib induces growth arrest and up-regulation of antigen processing and presentation in the tumor cells (Goel et al., 2017). This enhances the immunogenicity of the tumor cells. Consistent with this, the number of CD3⁺ cells recruited into the tumor mass increases after treatment, allowing for the stimulation of cytotoxic T-cells (CTLs). Additionally, CDK4 is necessary for the development of CD4⁺FOXP3⁺ Treg cells that can suppress CTL responses. In both tumor bearing and non-tumor bearing animals, CDK4 deficiency is associated with a reduced number of infiltrating and circulating CD4⁺FOXP3⁺ Treg cells with minimal impact on other T-cell subsets (Chow et al., 2010; Goel et al., 2017). Thus, by enhancing the antigenicity of the tumor cell and suppressing the negative regulatory cells, one can achieve a substantial effect on tumor growth using CDK4/6 inhibitors (Figure 1D).

Interestingly, the action of CDK4/6 inhibitors on the microenvironment might not be only through its ability to block Rb phosphorylation and promote cell cycle exit. CDK4 was shown to have kinase activity towards SPOP, a cullin 3 E3 ubiquitin ligase adaptor protein that can interact with PD-L1 (Zhang et al., 2018). Treatment with palbociclib inhibits SPOP phosphorylation, promoting its degradation and blocking PD-L1 proteasome mediated degradation. CDK4/6 inhibition is also able to stimulate PD-1 expressing T-cells in vitro and enhance T cell infiltration (Deng et al., 2017). This enhancement is driven at least in part by CDK6-induced phosphorylation of NFAT4. Combining palbociclib with a PD-1 blockade enhanced tumor regression and improved overall survival in mouse xenograft models of colon adenocarcinoma (Deng et al., 2017; Zhang et al., 2018).

These examples illustrate the profound effect that CDK4/6 inhibition can have on tumor growth by acting on the cells of the tumor microenvironment, either by affecting proliferation and maturation, or by affecting antigen processing and other immunological features of the cells, both tumor and normal. Such observations empower new opportunities to combine CDK4/6 inhibitors with immune checkpoint blockades or other types of immunotherapy. Indeed, it also raises the mechanistic possibility that the effect of the signaling pathway inhibitor combinations may also strike at two (or more) different cellular targets to achieve response.

For example, although individually CDK4/6 inhibitors and PI3K inhibitors have little effect in a mouse model of triple negative breast cancer, the combination of CDK4/6 and PI3K inhibitors increases the expression of HLA antigens leading to an increase in tumor-infiltrating cytotoxic CD4⁺ and CD8⁺ T cells and natural killer cells, and a decrease in the CD4⁺FOXP3⁺ Treg suppressor cells (Teo et al., 2017). While these investigators did not experimentally address why the combination was better, they further showed that adding an immune checkpoint blockade induces complete and durable regression of established tumors.

Given the importance of CDK4 in the tumor microenvironment, it may be somewhat surprising that CDK4/6 inhibitors have no significant effect on Rb-negative tumors. It is possible that growth arrest in the tumor cell is necessary for the changes in histocompatibility to manifest. Alternatively, and by no means exclusively, the ability of CDK4/6 to sculpt the microenvironment may not be sufficient to reduce tumor burden. In the oligodendroglioma model described above (Ciznadija et al., 2011), ectopic PDGF expression is sufficient to drive the initiation of disease in CDK4 positive tumor cells, but if the microenvironment is CDK4 deficient these low-grade lesions would not progress to high grade lethal malignancies. It is possible that the experiments examining the effect of CDK4/6 inhibitors in Rb-negative tumors in both animals and in humans are simply not powered to identify more subtle changes in tumor grade. Regardless, multiple cellular targets of the CDK4/6 inhibitors, in the tumor cell and in the tumor microenvironment, can also cooperate to yield optimal clinical success.

CONCLUSION

It is clear that the mechanisms by which CDK4/6 inhibitors can retard cancer progression are far more diverse than originally thought. The in vivo functions of CDK4/6 inhibition are likely to extend beyond simply enforcing reversible cytostasis. Nevertheless, it is tempting to speculate that the alternative mechanisms discussed here may not be completely separate. Senescent cells are characterized by metabolic changes and elaboration of cytokines that modulate the immune response. Thus, the ability of CDK4/6 inhibitors to drive tumor cells into senescence may lead to changes in the immune response and cellular metabolism, yielding a unified mechanistic cellular response. Consequently, it will be exciting to examine each mechanism as we learn how drug cooperativity benefits each patient population.