Immunomodulation by anticancer cell cycle inhibitors

Abstract

Cell cycle proteins that are often dysregulated in malignant cells, such as cyclin-dependent kinase 4 (CDK4) and CDK6, have attracted considerable interest as potential targets for cancer therapy. In this context, multiple inhibitors of CDK4 and CDK6 have been developed, including three small molecules (palbociclib, abemaciclib and ribociclib) that are currently approved for the treatment of patients with breast cancer and are being extensively tested in individuals with other solid and haematological malignancies. Accumulating preclinical and clinical evidence indicates that the anticancer activity of CDK4/CDK6 inhibitors results not only from their ability to block the cell cycle in malignant cells but also from a range of immunostimulatory effects. In this Review, we discuss the ability of anticancer cell cycle inhibitors to modulate various immune functions in support of effective antitumour immunity.

The ability of malignant cells to ignore inhibitory inputs from the microenvironment and proliferate in an uncontrolled manner has been recognized for a long time as one of the hallmarks of cancer and has been considered a major therapeutic target for more than three decades¹. In fact, most (if not all) classical chemotherapeutics, as well as radiation therapy, which were empirically discovered and translated to the clinic well before the advent of modern cell biology, have some selectivity for tumour cells because of their increased proliferation rate². In line with this, conventional anticancer regimens also target rapidly proliferating normal tissues, such as the intestinal epithelium and hair follicles, which often underlies treatment-related toxicities³.

With the advent of modern genetics and molecular bio logy in the 1980s, we began to obtain an increasingly clear picture of the genetic alterations that underlie uncontrolled proliferation in malignant cells. These defects include (but are not limited to) the upregulation of proteins that drive cell cycle progression, such as cyclin D1 (CCND1), cyclin-dependent kinase 4 (CDK4) and CDK6, as well as the loss of genes encoding cell cycle inhibitors, such as retinoblastoma 1 (RB1) and cyclin-dependent kinase inhibitor 2A (CDKN2A; which encodes p16^INK4A)⁴ (BOX 1). The discovery of these molecular alterations generated considerable interest in developing specific cell cycle inhibitors for cancer therapy^5,6. Thus, starting with the demonstration that a pharmacological inhibitor of CDK4 and CDK6 (palbociclib; originally known as PD 0332991) blocked human primary myeloma cells in the G1 phase of the cell cycle⁷ (BOX 1) and was active as a single agent in patients with mantle cell lymphoma⁸, a series of preclinical and clinical studies have culminated in the regulatory approval of three such agents for use in humans over the past 5 years. Specifically, the FDA has licensed three distinct inhibitors of CDK4 and CDK6 (referred to here as CDK4/CDK6 inhibitors) for the treatment of patients with hormone receptor-positive breast cancer, namely palbociclib (PD 0332991; Ibrance), ribociclib (LEE-011; Kisqali) and abemaciclib (LY2835219; Verzenio)⁶. These and other CDK4/CDK6 inhibitors are now being investigated for safety and efficacy in a large number of clinical studies enrolling patients with haematological and solid malignancies (ClinicalTrials.gov).

Box 1 |. general organization of the cell cycle in mammalian cells.

Most adult mammalian cells exist in a differentiated, non-proliferative state commonly referred to as quiescence or G0, which enables their physiological functions. upon mitogenic stimulation, however, differentiated cells can engage in a (generally finite) number of divisions to generate progeny. Classically, the lifespan of an actively proliferating cell has been subdivided into four major phases (see the figure): G1, a long phase characterized by intense anabolism (including robust rNa and protein synthesis) in support of cellular growth; S, a phase dedicated to DNA synthesis; G2, a relatively short phase in preparation for cell division; and M, the actual mitotic phase. In mammals, cell cycle progression is orchestrated by a molecular machinery centred on the phase-specific expression of regulatory proteins, commonly known as cyclins, that assemble with (and hence activate) specific cyclin-dependent kinases (CDKs) to promote entry into the next phase. thus, the G1–S phase transition involves the sequential activation of CDK4 or CDK6 by D-type cyclins (CCNDs)¹⁴¹, followed by the activation of CDK2 by E-type cyclins (CCNes)^142,143. ultimately, this results in the release of retinoblastoma 1 (rB1) from transcription factors of the e2F family, culminating with the transactivation of numerous genes involved in DNA synthesis¹⁴⁴. along similar lines, CDK2 complexed with A-type cyclins (CCNAs) supports progression through the S phase, CDK1 bound to CCNAs or B-type cyclins (CCNBs) promotes the S–G2 or G2–M phase transition, respectively, whereas CDK3 assembled with C-type cyclins promotes the G0–G1 transition (not shown)^145,146. the M phase terminates with the proteasomal degradation of CCNBs, resulting in CDK1 inactivation and transition to either G1 for continued proliferation or to G0 for quiescence¹⁴⁷. importantly, the catalytic activity of cyclin–CDK complexes is negatively regulated by several endogenous inhibitors, which block cell cycle progression until optimal conditions are achieved. these inhibitors, whose affinity for different CDKs varies considerably, include p15^iNK4B (encoded by CDKN2B)¹⁴⁸, p16^iNK4a (encoded by CDKN2A)¹⁴⁹, p18^iNK4C (encoded by CDKN2C)¹⁵⁰, p19^iNK4D (encoded by CDKN2D)¹⁵¹, p21^CiP1 (encoded by CDKN1A)¹⁵², p27^KiP1 (encoded by CDKN1B)¹⁵³ and p57^KiP2 (encoded by CDKN1C)¹⁵⁴. whereas members of the iNK4 family specifically inhibit CDK4 and CDK6, p21^CiP1, p27^KiP1 and p57^KiP2 operate with relatively broad specificity for multiple CDKs. together, these inhibitors are crucial not only for regulating the cell cycle in physiological settings but also for the permanent proliferative inactivation of cells experiencing irreparable DNA damage, in other words, cellular senescence⁶⁸.

The clinical activity of CDK4/CDK6 inhibitors has largely been attributed to their capacity to selectively block malignant cells in the G1 phase of the cell cycle by inhibiting the CDK4/CDK6-mediated inactivating phosphorylation of RB1, thus restoring the inhibitory association of RB1 with transcriptional factors of the E2F family⁹ (BOX 1). Accordingly, some tumours that are insensitive to CDK4/CDK6 inhibition have genetic defects that prevent RB1 from enforcing cell cycle arrest, such as loss-of-function RB1 mutations¹⁰ or 13q14 deletion¹¹. However, not all cases of resistance to CDK4/CDK6 inhibitors can be attributed to cell cycle defects in malignant cells¹², which indicates that additional mechanisms may contribute to the therapeutic activity of palbociclib, ribociclib and abemaciclib. Indeed, CDK4 and CDK6 are involved in the regulation of multiple processes beyond (and independent of) cell cycle progression¹³, including energy metabolism^14,15, lysosomal functions¹⁶ and metastatic dissemination¹⁷. Moreover, emerging preclinical evidence suggests that systemic CDK4/CDK6 inhibition has various immunomodulatory effects that favour antitumour immunity¹⁸.

In this Review, we discuss the immunomodulatory activity of CDK4/CDK6 inhibitors, focusing on their regulatory effects on the proliferation and activation of immune cells, their capacity to promote cytokine secretion by malignant cells, their ability to regulate the expression of immunosuppressive ligands and the potential mechanisms whereby tumour cells evade immunity downstream of cell cycle inhibition. Non-immunological processes that may contribute to the clinical activity of CDK4/CDK6 inhibitors are not discussed here.

Cyclins and CDKs in immune cells

Perhaps not surprisingly, cyclins and CDKs are necessary for the development and physiological functions of immune cells¹⁹. However, not all immune cell populations require the same cyclins and CDKs for development, expansion and activation, which results in variable sensitivity to CDK inhibition. Moreover, multiple CDKs influence the activity of immune cells through cell cycle-independent pathways.

Cell cycle-dependent mechanisms.

Cdk4^−/− mice develop type 1 diabetes and sterility (in males), but no haematopoietic cell dysfunction²⁰, whereas Cdk6^−/− mice or mice lacking cyclin D3 (Ccnd3^−/− mice) develop multiple defects in the mature B cell^21–23 and granulocyte²⁴ compartments, as well as hypoplasia of the thymus and spleen^25–27. In line with this, Cdkn2c^−/− mice, which lack the endogenous CDK4/CDK6 inhibitor p18^INK4C (also known as CDKN2C) (BOX 1), have impaired plasma cell differentiation associated with an inability of B cells to arrest proliferation during physiological antibody responses, but they have no developmental defects^28,29.

Thymic atrophy in Cdk6^−/− mice correlates with a skewing towards double-negative thymocytes, largely owing to defective Notch signalling and consequent upregulation of IL-2 receptor subunit-α (IL-2RA; also known as CD25)^27,30,31. Indeed, CCND3 is highly expressed by double-negative thymocytes at the transition between the DN4 (CD25⁻CD44⁻) stage and the immature single-positive stage, following assembly of the pre-T cell receptor (pre-TCR), and the Ccnd3^−/− genotype imposes a block at the DN3 (CD25⁺CD44⁻) stage²⁷. Consistently, mice lacking the CDK4/CDK6 inhibitor p16^INK4A (BOX 1) develop thymic hyperplasia resulting from the accumulation of CD4⁺CD8⁺ thymocytes³². Of note, cyclin D2 (CCND2) expression in double-negative thymocytes seems to peak at the DN1–DN3 stage but becomes barely detectable after pre-TCR assembly²⁷, and Ccnd2^−/− mice have a specific defect in the development of CD5⁺ B cells³³. Thus, thymocytes and B cells at different developmental stages and from different lineages may require different D-type cyclins for their differentiation and maturation. However, neither the absence of CDK6 nor pharmacological CDK4/CDK6 inhibition affects the proliferation of mature effector T cells^24,34. Taken together, these observations highlight a key role for CDK6 (but less so CDK4) in the generation of a mature T cell repertoire, but not for effector T cell functions. Further supporting this notion, CD4⁺CD25⁺FOXP3⁺ regulatory T (T_reg) cells express CDK6 at higher levels than effector T cells³⁵, in which CDK4 and CDK6 activity is constitutively repressed by the expression of the CDK4/CDK6 inhibitor p18^INK4C (REF.³⁶) (BOX 1).

Cyclins and CDKs control not only the development of immune cells but also their differentiation and activation³⁷. Thus, CCND2 is required for B cells to proliferate in response to engagement of the B cell receptor, but not in response to CD40 signalling or lipopolysaccharide administration³³. Moreover, the differentiation of naive T cells to functional effector T cells requires CDK2, which normally operates downstream of CDK4 and CDK6 in cell cycle regulation³⁸. Consistent with this, Cdk2 deletion or overexpression of the broad-spectrum CDK inhibitor p27^KIP1 (also known as CDKN1B) (BOX 1) causes T cell anergy and inhibits the formation of IL-2-producing central memory T cells^39,40, whereas loss of p27^KIP1 promotes the differentiation of CD4⁺ T cells into effector T cells that produce interferon-γ (IFNγ), also in the absence of co-stimulation^39,41,42.

Cell cycle-independent mechanisms.

The differentiation of effector T cells also seems to require CDK5 (REF.³⁸), an atypical CDK that is activated upon hetero-dimerization with p35^NCK5A (also known as CDK5R1)⁴³. Intriguingly, CDK2 or CDK5 dysfunction not only compromises the activity of effector T cells but also promotes peripheral tolerance by favouring the differentiation of T_reg cells^41,42,44,45. Of note, CDK5 is not a canonical cell cycle regulator⁴⁶, which suggests that its influence on effector T cell functions is unrelated to cell cycle control. Further supporting this notion, CDK5 negatively regulates the expression of forkhead box P3 (FOXP3), which is crucial for the function of T_reg cells⁴⁷. Moreover, CDK5 can promote IL-2 expression by inhibiting histone deacetylase (HDAC) activity and binding to the IL2 promoter⁴⁸. The ability of CDK2 to support effector T cell functions may also originate, at least in part, from cell cycle-independent effects. Indeed, CDK2 shares with CDK5 the capacity to inhibit FOXP3 expression, although it does so by negatively regulating FOXP3 protein stability⁴⁹. Moreover, Cdk2^−/− mice do not have defects in lymphopoiesis, which is highly dependent on normal cell cycle progression⁴².

Cell cycle-independent mechanisms through which CDKs influence immune cell activity have also been reported for CDKs other than CDK2 and CDK5. For example, CDK4 can phosphorylate mitogen-activated protein kinase 8 (MAPK8; also known as JNK) in both human and mouse dendritic cells, resulting in IL-6 and IL-12 secretion downstream of AP-1-dependent transcriptional programmes (but no dendritic cell proliferation)⁵⁰. Moreover, CDK4 and CDK6 are involved in the ability of neutrophils to respond to microbial pathogens by extruding neutrophil extracellular traps, an activity that does not depend on cell cycle progression (as neutrophils are terminally differentiated cells) but is negatively regulated by the broad-spectrum CDK inhibitor p21^CIP1 (also known as CDKN1A)⁵¹. Intriguingly, in mouse and human dendritic cells, AP-1 activity seems to be inhibited by CDK8. CDK8 does not regulate cell cycle progression but operates together with mediator complex subunit 12 (MED12) and MED13 as part of the mediator complex⁵², with IL-10 secretion occurring upon pharmacological CDK8 inhibition⁵³. At least in part, such an effect can be attributed to the ability of CDK8 to catalyse the activating phosphorylation of signal transducer and activator of transcription 1 (STAT1), which generally supports dendritic cell immunostimulatory functions through the production of IFNγ^54,55. However, whether other factors are involved in the ability of CDK8 to suppress IL-10 secretion remains to be clarified.

STAT1 is also crucial for the development, activation and cytotoxic function of natural killer (NK) cells^56,57. In mouse embryonic fibroblasts, CDK8 bound to C-type cyclins phosphorylates STAT1 to support IFNγ secretion and the activation of transcriptional programmes that support resistance to infection⁵⁸. In line with this, pharmacological inhibition of CDK8 or its paralogue CDK19 promotes T_reg cell differentiation, ex vivo as well as in a mouse model of experimental autoimmune encephalomyelitis, as a consequence of decreased STAT1 signalling⁵⁹. Conversely, Cdk8 deletion from the NKp46⁺ compartment increases (instead of compromising) NK cell-mediated cytotoxicity and antitumour activity in vivo, in mouse models of melanoma, lymphoma and leukaemia⁶⁰. In this setting, however, STAT1 phosphorylation in NK cells is unaffected by the absence of CDK8, potentially owing to the compensatory upregulation of CDK19 expression⁶⁰. Thus, the superior effector functions of Cdk8^−/− NK cells may result from decreased activation of immunosuppressive CDK8-dependent signalling pathways, including (but not necessarily limited to) the SMAD family member 3 (SMAD3)-dependent secretion of transforming growth factor-β1 (TGFβ1)⁶¹. Additional work is required to elucidate this possibility.

Overall, these findings exemplify the key role of CDKs in the development and function of the immune system, as they delineate cell type-dependent and developmental stage-dependent requirements for specific CDKs, as well as numerous cell cycle-independent functions for CDKs (FIG. 1). In this context, CDK4 and CDK6 stand out as promising targets for the development of anticancer agents that do not compromise immuno surveillance, and at the same time support the (re)activation of tumour-targeting immune responses, as discussed below.

Fig. 1 |. Cyclins and cyclin-dependent kinases in immune cells.

Genetic data from whole-body or conditional knockout studies show that numerous immune cell types require the cell-intrinsic functions of specific cyclins (CCNs) and cyclin-dependent kinases (CDKs) for their development, maturation or activation. Importantly, the ability of specific CCNs and CDKs to regulate immune functions can involve cell cycle-dependent mechanisms (red arrows) as well as cell cycle-independent mechanisms (blue arrows). MED, mediator complex subunit; NK, natural killer; T_reg cell, regulatory T cell.

Immunostimulation upon CDK inhibition

Although the main consequence of CDK4/CDK6 inhibition is the establishment of an RB1-dependent cell cycle arrest, emerging preclinical data indicate that palbociclib, ribociclib and abemaciclib also promote various immunomodulatory effects as they interact with malignant cells as well as with immune cell populations of the tumour microenvironment (FIG. 2).

Fig. 2 |. Immunostimulation upon cyclin-dependent kinase inhibition.

Inhibitors of cyclin-dependent kinase 4 (CDK4) and CDK6 — herein referred to as CDK4/CDK6 inhibitors — can have various immunomodulatory effects as they target both malignant and non-malignant components of the tumour microenvironment. CDK4/CDK6 inhibition has been associated with immunostimulatory effects that largely reflect increased antigen presentation by tumour cells (1); the reactivation of endogenous retroviruses (ERVs) and the consequent secretion of type III interferon (2); production of multiple pro-inflammatory cytokines and chemokines, such as CC-chemokine ligand 5 (CCL5), as part of the senescence-associated secretory phenotype (SASP) (3); direct activation of effector T cells upon derepression of nuclear factor of activated T cells 1 (NFATC1; best known as NFAT) (4); or inhibition of regulatory T (T_reg) cell proliferation through repression of DNA methyltransferase 1 (DNMT1) expression (which is expressed downstream of the release of retinoblastoma 1 (RB1) from E2F family transcription factors by active CDK4/CDK6), and consequent upregulation of the cell cycle inhibitor p21^CIP1 (5). CCND, D-type cyclin; dsRNA, double-stranded RNA; P, phosphate.

SASP-related effects on tumour cells.

In multiple preclinical settings, the cell cycle block imposed on tumour cells by CDK4/CDK6 inhibitors is associated with the acquisition of phenotypic and biochemical biomarkers of cellular senescence^62–66, an irreversible proliferative arrest that is characterized by cellular enlargement, increased β-galactosidase activity and the presence of senescence-associated heterochromatin foci as a consequence of accumulating DNA damage^67,68 (BOX 2). Importantly, senescent cells have robust immun omodulatory effects, through at least two partially overlapping mechanisms. On the one hand, senescent tumour cells expose multiple activating ligands for NK cells on their surface, including UL16 binding protein 1 (ULBP1) and ULBP2 (REF.⁶⁹), which promote the activation of NK cells downstream of killer cell lectin-like receptor K1 (KLRK1; also known as NKG2D) signalling^70,71. On the other hand, senescent cells produce various soluble mediators, including growth factors such as hepatocyte growth factor (HGF), as well as numerous cytokines and chemokines, which are collectively known as the senescence-associated secretory phenotype (SASP)^72,73.

Box 2 | Key features of cellular senescence.

Mammalian cells respond to perturbations of the intracellular or extracellular microenvironment by activating various cytoprotective mechanisms that aim at limiting molecular damage and, hence, enabling the recovery of physiological cellular functions¹⁰⁴. However, when such perturbations overwhelm the cellular capacity for adaptation, cells can undergo either of two, non-mutually exclusive terminal fates: they can activate one of the mechanisms that drive cellular suicide via regulated cell death¹⁵⁵; or they can enter a state of irreversible proliferative arrest commonly known as cellular senescence⁶⁸. this decision depends on various factors, including (but not limited to) the type, intensity and duration of the initiating stress, as well as the expression levels of multiple proteins that support cellular survival and senescence over cell death, such as various anti-apoptotic members of the BCL-2 protein family¹⁵⁶. In line with this, inhibitors of the BCL-2 system, such as venetoclax (a drug that is commonly used for the treatment of chronic lymphocytic leukaemia), can promote the death of senescent cells¹⁵⁷. in addition to being unable to proliferate as a consequence of the stable overexpression of cell cycle inhibitors including p16^iNK4a and p21^CiP1 (REFS^158,159) (BOX 1), senescent cells have a panel of distinguishing features that have been commonly used for experimental assessments in vitro and in vivo. These features include an enlarged, flattened morphology; increased lysosomal β-galactosidase activity; accumulation of DNA defects that result in telomere shortening and the emergence of senescence-associated heterochromatin foci; oxidative damage to lipids and proteins as a consequence of altered metabolism and redox homeostasis; and secretion of various mitogenic and immunomodulatory factors that are cumulatively referred to as the senescence-associated secretory phenotype (SASP)⁶⁸. importantly, most of these features are not shared by cells that are temporarily withdrawing from replication in the context of quiescence⁶⁸ (BOX 1). thus, in addition to being reversible, quiescence is generally imposed by p27^KiP1, is not associated with macromolecular damage and does not involve an abundant secretion of bioactive factors⁶⁸.

Importantly, the net immunomodulatory effects of the SASP vary considerably in different experimental settings⁷³. At least in part, this reflects the fact that the precise mediators secreted by senescent tumour cells depend on multiple variables, including the molecular driver of senescence and the genetic configuration of the tumour⁶⁸. Thus, whereas tumour cells with operational tumour protein 53 (TP53; best known as p53) secrete high levels of CC-chemokine ligand 2 (CCL2) as they undergo DNA damage-driven senescence, and this is instrumental for the recruitment of NK cells and CD4⁺ T cells and the activation of senescence surveillance, the same does not hold true for p53-deficient cells^74–76 or for p53-competent cells undergoing senescence in response to other stimuli⁶⁸. Intriguingly, similar considerations apply to non-malignant cells, at least in some scenarios. Thus, whereas the SASP of p53-expressing non-malignant hepatic stellate cells promotes the polarization of local macrophages towards an oncosuppressive M1-like, pro-inflammatory phenotype, p53-deficient hepatic stellate cells produce M2-polarizing, pro-tumour cytokines, which support oncogenesis and tumour progression in the liver^77,78.

Another factor that influences the net immunomodulatory effects of the SASP is the phenotypic and biochemical heterogeneity that characterizes most, if not all, human tumours⁷⁹. In this context, some tumour cells can avert senescence and take advantage of the mitogenic and immunosuppressive SASP components that are secreted by their senescence-sensitive counterparts for proliferation and immunoevasion^80,81. Importantly, multiple cytokines secreted by senescent cells, notably CXC-chemokine receptor 2 (CXCR2) ligands as well as cytokines produced upon activation of the cytosolic DNA sensor cyclic GMP–AMP synthase^82,83, reinforce cellular senescence by autocrine or paracrine mechanisms^84,85. Thus, the establishment of cellular senescence downstream of CDK4/CDK6 inhibition is associated with immunomodulatory effects that, at least in some settings, support the therapeutic effects of anticancer cell cycle inhibitors. Importantly, such effects may be offset, at least partially, by inactivation of the p53 system, which is a common finding in human tumours⁸⁶.

Further supporting this contention, cellular senescence driven by CDK4/CDK6 inhibitors has been associated with increased tumour infiltration by CD45⁺ cells downstream of NF-κB activation and consequent secretion of CCL5 in immunocompetent models of melanoma^87–89. In silico data from The Cancer genome Atlas corroborate an association between elevated CCL5 levels and increased abundance of transcripts indicative of T cell infiltration and activation in human melanoma samples⁸⁷. Similarly, palbociclib combined with a mitogen-activated protein kinase (MAPK) inhibitor, but neither agent used alone, mediated therapeutic effects in a model of KRAS-mutant lung cancer that were linked to the accumulation of senescent cells in the tumour microenvironment and the secretion of SASP components including chemotactic and activating factors for NK cells, such as CCL2, CCL5 and IL-15 (REF.⁹⁰). Taken together, these findings exemplify the ability of the SASP associated with CDK4/CDK6 inhibition to reconfigure the immunological tumour microenvironment in support of antitumour immunity.

SASP-unrelated effects on tumour cells.

Importantly, CDK4/CDK6 inhibitors have also been shown to favour the recruitment and activation of immune effector cells in the context of cellular senescence but independently of conventional components of the SASP³⁴. In particular, in colorectal carcinoma and in various preclinical models of breast cancer, including patient-derived xenografts, abemaciclib administration has been linked to the upregulation of numerous components of the antigen processing and antigen presentation machinery, including MHC class I molecules (namely, HLA-A, HLA-B and HLA-C in human systems, and H2–D1, H2–K1 and B2M in mouse systems)^34,91. In support of MHC class I upregulation making a mechanistic contribution to the therapeutic effects of abemaciclib, ovalbumin (OVA)-expressing mouse tumour cells have increased sensitivity to OVA-specific OT-I T cells upon exposure to abemaciclib in vitro, and the therapeutic effects of abemaciclib in vivo, in syngeneic immunocompetent mouse models, are abrogated by CD8⁺ T cell depletion³⁴. Moreover, T cells that infiltrate abemaciclib-treated breast tumours have lower levels of exhaustion markers such as PD1 (also known as PDCD1) and cytotoxic T lymphocyte-associated protein 4 (CTLA4)³⁴, and the microenvironment of mouse colorectal carcinomas treated with abemaciclib is characterized by increased levels of transcripts encoding T cell activation markers such as IFNγ and granzyme B (GZMB), and co-stimulatory receptors such as TNF receptor superfamily member 9 (TNFRSF9; also known as CD137)⁹¹. Finally, data from The Cancer Genome Atlas breast cancer cohort point to a negative correlation between CCND1 amplification (and, hence, hyperactive CDK4 and CDK6) and MHC class I abundance³⁴. Notably, in the same models, increased antigen presentation has also been linked to the re-expression of endogenous retroviral elements and consequent accumulation of double-stranded RNA species in the cytosol of tumour cells. Ultimately, this has been shown to drive type III interferon secretion and, hence, activate STAT1 and STAT2 signalling through both autocrine and paracrine circuits³⁴. Similar results have recently been obtained with a chemical inhibitor of the upstream cell cycle regulator CDK7 (REF.⁹²). Specifically, CDK7 inhibition in non-small-cell lung carcinoma cells was associated with genomic instability accompanied by micronucleation and the consequent secretion of immunostimulatory cytokines, including tumour necrosis factor (TNF) and CXC-chemokine ligand 10 (CXCL10)⁹². Ultimately, pharmacological CDK7 inhibition resulted in increased tumour infiltration by immune cells together with a therapeutic immune response that could be boosted by PD1 blockade⁹².

Although the precise contribution of these immunomodulatory effects to the clinical efficacy of CDK4/CDK6 inhibitors remains to be determined, the state of viral mimicry that is induced in tumour cells by interferon signalling has been linked to superior antitumour immunity in various preclinical and clinical settings^93–95. Of note, CDK6 has been shown to support the secretion of the pro-angiogenic factor vascular endothelial growth factor A (VEGFA) in a model of acute lymphoid leukaemia⁹⁶. As VEGFA mediates immunosuppressive effects by favouring T_reg cell maturation^97,98, these findings could be interpreted as describing another mechanism for immunostimulation mediated by CDK4/CDK6 inhibitors. However, VEGFA secretion is induced by CDK6 independently of its kinase activity⁹⁶; thus, palbociclib, abemaciclib and ribociclib (which all operate at the catalytic site of CDK6) are not expected to have such activity.

Direct effects on immune cells.

CDK4/CDK6 inhibitors also mediate immunomodulatory effects by targeting immune cells. In particular, systemic administration of palbociclib or abemaciclib was reported to suppress the proliferation of tumour-infiltrating, splenic and lymph-node T_reg cells in mice³⁴, at least in part as a consequence of DNA methyltransferase 1 (DNMT1) inhibition and consequent upregulation of the endogenous cyclin-dependent kinase inhibitor p21^CIP1 (REFS^34,99). The specificity of CDK4/CDK6 inhibitors for T_reg cells over effector T cells in this respect seems to result from the increased dependence of the former (but not the latter) on CDK4 and CDK6 for proliferation^100,101. Moreover, palbociclib, abemaciclib and trilaciclib (another CDK4/CDK6 inhibitor) have been shown to boost effector T cell functions by supporting the activity of the key transcription factor nuclear factor of activated T cells 1 (NFATC1; also known as NFAT), resulting in increased synthesis of T cell mitogens (such as IL-2) and effector molecules (such as GZMB)^91,102. Of note, these observations were obtained both in vivo, in immunocompetent mice and humans¹⁰², and in vitro⁹¹, and the studies confirm a direct effect of CDK4/CDK6 inhibitors on effector T cells. However, it remains to be clarified whether such an immunostimulatory effect results from on-target effects on CDK4 and CDK6.

Irrespective of this and other open questions, it is now clear that CDK4/CDK6 inhibitors mediate a range of immunostimulatory effects that may contribute to their clinical activity (FIG. 2). However, in some settings, tumour cells can evade the immunosurveillance induced by CDK4/CDK6 inhibitors, which calls for the development of therapeutic strategies that restore antitumour immunity, as described below.

Immunoevasion upon CDK inhibition

In addition to escaping the cell cycle blockade that is imposed by CDK4/CDK6 inhibitors by losing RB1 expression or by upregulating expression of CDK4 or CDK6 (REFS^11,103), malignant cells can also evade the antitumour immune responses that are supported by CDK4/CDK6 inhibitors. Thus, the mechanisms underlying the capacity of malignant cells to evade the immune response in the context of cell cycle inhibition identify potential therapeutic targets to address clinical resistance to CDK4/CDK6 inhibitors (FIG. 3).

Fig. 3 |. Immunoevasion upon cyclin-dependent kinase inhibition.

Malignant cells harness numerous strategies to escape immunosurveillance in the context of inhibition of cyclin-dependent kinase 4 (CDK4) and CDK6. For example, specific components of the senescence-associated secretory phenotype (SASP) driven by CDK4/CDK6 inhibitors, such as CC-chemokine ligand 2 (CCL2) and transforming growth factor-β (TGFβ), favour the proliferation of senescence-resistant neighbouring tumour cells and the establishment of local immunosuppression by paracrine mechanisms. Moreover, inhibition of speckle type BTB/POZ protein (SPOP) by CDK4/CDK6 inhibitors limits PDL1 degradation by the proteasome, hence promoting immunosuppression via tumour cell-intrinsic pathways. Strategies that prevent or compensate for such mechanisms may be instrumental for the management of human tumours that are or become resistant to CDK4/CDK6 inhibitors. MDM2, MDM2 proto-oncogene; PTEN, phosphatase and tensin homologue; TIL, tumour-infiltrating lymphocyte.

Senescence-related immunosuppression.

Senescence has a key role in the immunostimulatory effects of CDK4/CDK6 inhibitors but can also promote immunosuppression or favour the proliferation of senescence-resistant tumour cells through specific components of the SASP^80,104. Thus, senescent cells can be targeted, at least theoretically, to limit immunoevasion downstream of cell cycle inhibition. In support of this notion, the combination of palbociclib with a senolytic has superior therapeutic effects in a xenograft model of melanoma¹⁰⁵. However, targeting senescence at large may compromise part of the immunostimulatory effects of CDK4/CDK6 inhibitors, and therefore more refined therapeutic strategies are required. In this context, it is tempting to speculate that blocking key immunosuppressive components of the SASP, such as TGFβ1, may be beneficial^106–108. However, recent data suggest that blocking TGFβ1 results in tumour cell-intrinsic resistance to palbociclib upon CDK6 upregulation¹⁰⁹. Along similar lines, CCL2 has been suggested as a potential target to limit tumour infiltration by immunosuppressive monocytes and macrophages downstream of cellular senescence^110,111. However, CCL2 also promotes the recruitment of immune effector cells that may contribute to the clinical efficacy of CDK4/CDK6 inhibitors⁷⁵. Moreover, discontinuing the use of CCL2-neutralizing antibodies has been associated with a paradoxical increase in metastatic dissemination and decreased survival in multiple syngeneic mouse models of breast cancer¹¹², which calls for caution when considering CCL2 as a target to limit immunoevasion downstream of CDK4/CDK6 inhibition.

The cell cycle blockade imposed by CDK4/CDK6 inhibitors is not necessarily irreversible^64,66, and this may contribute to disease progression not only upon the recovery of tumour cell proliferation but also upon the resolution of senescence-related immunostimulation. Genetic and pharmacological strategies have been successfully used to reinforce senescence in tumour cells that are temporarily arrested in the cell cycle by CDK4/CDK6 inhibitors, including downregulation of the proto-oncogene HRAS⁶⁵, the inhibition of autophagy^63,113 and the blockade of lysosomal trapping (which limits the ability of CDK4/CDK6 inhibitors to induce senescence)⁶⁶. Conversely, the ability of mechanistic target of rapamycin (mTOR) inhibitors to strengthen senescence in tumour cells responding to CDK4/CDK6 inhibitors is inconsistent^114,115. This may reflect, at least in part, the ability of mTOR inhibitors to promote autophagy¹¹⁶ and, hence, limit senescence, an effect to which some tumour cells may be more susceptible¹¹⁷, and to mediate robust anti-inflammatory effects that may support immunoevasion^118,119.

Immunosuppression in the context of p53 or PTEN inactivation.

As discussed above, p53 inactivation in tumour cells (which occurs by mutational or non-mutational mechanisms in more than 50% of all tumours) has a major impact on the immunomodulatory effects of cellular senescence, including reduced expression of NK cell-activating ligands by senescent cells as well as altered cytokine secretion^74,75,77. Similar considerations apply to deletion of phosphatase and tensin homologue (PTEN), which is also common across different malignancies¹²⁰. Indeed, cellular senescence in PTEN-deleted pro state tumours is associated with massive infiltration by immuno suppressive CD11b⁺Gr-1⁺ myeloid cells^121,122. Thus, strategies to compensate for p53 inactivation or PTEN deletion are promising approaches to limit immuno evasion downstream of CDK4/CDK6 inhibition. Numerous molecules, including PRIMA-1^MET, which restores the functions of some p53 mutants, and nutlin 3A, which inhibits p53 degradation by MDM2 proto-oncogene (MDM2), have demonstrated therapeutic effects in preclinical tumour models⁸⁶, including melanoma patient-derived xenografts resistant to palbociclib¹²³, but these molecules have never been investigated for their ability to restore the immunostimulatory effects of CDK4/CDK6 inhibition. Importantly, MDM2 not only degrades p53 (and hence skews the SASP towards an immunosuppressive profile) but also counteracts cellular senescence induced by CDK4/CDK6 inhibitors as it favours the secretion of IL-6, which further promotes local immunosuppression^65,124,125. Consistent with this notion, the MDM2 antagonist RG7388 synergizes with palbociclib for the control of liposarcomas in vivo¹²⁶, although the immunological correlates of this approach were not investigated. In addition to identifying another mechanism by which MDM2 inhibitors support the therapeutic effects of CDK4/CDK6 inhibition, these observations indicate that tocilizumab, a clinically available IL-6 blocker that is currently used for the management of cytokine release syndrome, may also support the activity of palbociclib, abemaciclib and ribociclib in patients with MDM2 amplifications.

PDL1-dependent immunoevasion.

CDK4/CDK6 inhibitors have been shown to stabilize the immunosuppressive molecule PDL1 by blocking its proteasomal degradation downstream of speckle-type BTB/POZ protein (SPOP) in various PTEN-competent cells, including a human breast cancer cell line, mouse embryonic fibroblasts and a mouse melanoma cell line¹²⁷. As PDL1 and its major receptor on T cells (PD1) are closely involved in immunoevasion, and numerous PD1 or PDL1 inhibitors are clinically available¹²⁸, these findings suggest that targeting the PD1–PDL1 axis may be beneficial in patients who acquire resistance to CDK4/CDK6 inhibitors. Lending further support to the validity of this approach, PD1 and PDL1 inhibitors have been shown to improve the therapeutic effects of palbociclib and abemaciclib in syngeneic mouse models of melanoma^89,127 and colorectal carcinoma^91,102. These findings led to multiple clinical trials testing CDK4/CDK6 inhibitors in combination with PD1 or PDL1 blockers in patients with cancer (ClinicalTrials.gov). However, inhibition of CDK4 and CDK6 has also been associated with decreased transcription from the PDL1 locus, as a consequence of decreased NF-κB activity downstream of RB1-dependent phosphorylation¹²⁹. Consistent with this, a small RB1-derived phosphomimic peptide synergizes with radiation in the control of Pten-null mouse prostate cancer cells growing in syngeneic immunocompetent hosts, which resembles the effects of PDL1 blockade¹²⁹. Thus, PDL1 blockade may not necessarily improve immunostimulation by CDK4/CDK6 inhibitors in PTEN-null tumours, although this possibility has not yet been directly investigated.

Other mechanisms of immunoevasion.

Interestingly, the palbociclib-induced SASP has also been linked to platelet aggregation¹³⁰, which reportedly supports metastatic dissemination by promoting cancer cell motility¹³⁰ and by shielding metastatic cells from immune effectors, including CD8⁺ T cells and NK cells^131,132. However, whether platelet aggregation can be therapeutically targeted to limit immunoevasion in the context of CDK4/CDK6 inhibition remains to be elucidated. Similarly, the CDK family has been proposed to be indispensable for type I interferon release by monocytes¹³³, but this effect seems to require the inhibition of multiple CDKs other than (or in addition to) CDK4 and CDK6.

Irrespective of these and other unknowns, it is now clear that cancer cells can evade immunosurveillance downstream of CDK4/CDK6 inhibition by multiple mechanisms, which opens up novel immunotherapeutic avenues to address resistance in the clinical setting.

Concluding remarks

In addition to blocking tumour cells in the G1 phase of the cell cycle, sometimes permanently, CDK4/CDK6 inhibitors mediate a wide range of immunomodulatory effects that involve both malignant and non-malignant components of the tumour microenvironment. Such effects must be carefully considered when novel treatment regimens are designed, as they might be actionable and thus enable the implementation of superior immunotherapeutic protocols. For example, the ability of CDK4 to promote PDL1 degradation, at least in some oncological settings¹²⁷, provides a strong rationale to combine CDK4/CDK6 inhibitors with immune checkpoint blockers targeting the PD1–PDL1 axis. Along similar lines, the systemic inhibition of T_reg cells by CDK4/CDK6 inhibitors should be carefully considered when using these agents in patients with cancer who have an increased risk of autoimmune phenomena, including (but not limited to) transplant recipients.

With some partial exceptions⁸⁹, the majority of the data that are currently available on immunomodulation by CDK4/CDK6 inhibitors have been obtained with bulk methods (in other words, techniques that are intrinsically unable to provide single-cell resolution) in preclinical cancer models¹³⁴. Thus, it remains to be determined whether palbociclib, abemaciclib and ribociclib can mediate immunomodulatory effects in patients with cancer. Along similar lines, the degree of heterogeneity and the durability of these effects — which fraction of a specific cell population is affected and how this shifts over time — remain to be explored. These aspects are particularly important to clarify because human tumours are highly heterogeneous⁸¹ and can evolve markedly over time⁷⁹, which indicates that CDK4/CDK6 inhibitors may mediate immunomodulatory effects that ultimately affect only specific clusters of malignant cells or that become manifest only at specific stages of the treatment schedule. Properly designed clinical trials that will investigate longitudinal samples at single-cell resolution are urgently required to obtain novel insights into immunomodulation by CDK4/CDK6 inhibitors in patients.

Several questions remain unanswered with respect to some of the molecular pathways through which CDK4/CDK6 inhibitors mediate their immunomo dulatory effects. Indeed, whereas the DNMT1-dependent epigenetic consequences of CDK4/CDK6 inhibition — resulting in the upregulation of endogenous retroviral elements — can explain (at least in part) the ability of abemaciclib to drive type III interferon secretion^34,135, the precise molecular links between CDK4 or CDK6 and NFAT remain to be fully elucidated^91,102. Intriguingly, NFAT has been reported to downregulate the expression of CDK4 (REF.¹³⁶), which points to the existence of a feedback loop that regulates cell cycle progression in T cells (high levels of CDK4 and NFAT inhibition) versus the acquisition of effector functions (CDK4 inhibition and NFAT activation) in a mutually exclusive manner. Irrespective of this possibility, it will be important to clarify the extent to which the immunomodulatory effects of CDK4/CDK6 inhibitors originate from on-target effects. Indeed, whereas abemaciclib inhibits several serine/threonine kinases in addition to CDK4 and CDK6 (REF.¹³⁷), palbociclib at therapeutic doses has exquisite selectivity for CDK4 and CDK6 compared with other kinases^138,139, which suggests that palbociclib could be harnessed as a molecular tool to assess specificity. However, palbociclib has been reported to kill cultured hepatocellular carcinoma cells by inhibiting the protein phosphatase 5 catalytic subunit (PPP5C) in a CDK4-independent and CDK6-independent manner¹⁴⁰, which indicates that targets other than kinases may contribute (at least theoretically) to the antitumour activity of palbociclib. Moreover, whether the ability of abemaciclib and ribociclib to inhibit CDKs other than CDK4 and CDK6 or other CDK-unrelated kinases contributes to their clinical efficacy remains to be elucidated. Finally, it will be important to clarify which compartments of the tumour microenvironment are involved in the immunomodulatory effects of CDK4/CDK6 inhibitors in vivo. Indeed, until now, considerable attention has been dedicated to malignant and immune cells, but the potential contributions of stromal cells and endothelial cells remain unknown.

In conclusion, CDK4/CDK6 inhibitors stand out as multifaceted immunotherapeutic agents in disguise, which we believe is instrumental for their full-blown clinical activity. We surmise that fully characterizing the molecular mechanisms whereby palbociclib, abemaciclib and ribociclib influence immune activation will enable not only the development of novel combinatorial regimens with superior anticancer potential but also the definition of therapeutic strategies to circumvent immunoevasion downstream of cell cycle blockade. Rationally designed clinical trials and modern technologies for the longitudinal study of the tumour microenvironment at single-cell resolution will be instrumental in this sense.

Retinoblastoma 1.

(RB1). A multifunctional tumour suppressor protein that is dysfunctional in most human cancers. Biallelic germline loss of RB1 is associated with hereditary paediatric retinoblastoma.

Hormone receptor-positive breast cancer.

A common form of breast cancer characterized by the expression of oestrogen receptor 1 and/or progesterone receptor.

13q14 deletion.

A frequent genomic alteration, consisting of the deletion of a portion of the long arm of chromosome 13 that includes the retinoblastoma 1 (RB1) gene.

Type 1 diabetes.

An autoimmune form of glucose intolerance that originates from the immunological destruction of insulin-producing pancreatic β-cells and, hence, can be treated with insulin administration.

Notch signaling.

A highly conserved signal transduction pathway that regulates the differentiation of multiple cell types.

Histone deacetylase.

(HDAC). one of multiple enzymes that repress transcriptional programmes by removing functional acetyl groups from histones.

Mediator complex.

A tetrameric cyclin-dependent kinase 8 (CDK8)-containing or CDK19-containing complex that is involved in transcriptional regulation.

Senescence surveillance.

An immunological mechanism that ensures the natural killer cell-dependent and T cell-dependent elimination of potentially pathogenic senescent cells.

Hepatic stellate cells.

Liver-resident stromal cells that are largely found in proximity of the vasculature and contribute to hepatic fibrosis and carcinogenesis.

The Cancer Genome Atlas.

A large, publicly available patient data set containing mutational, transcriptional and pathophysiological data from multiple independent cohorts of patients with cancer.

Mitogen-activated protein kinase.

(MAPK). The generic name commonly used to indicate one or more members of a large serine/threonine protein kinase family involved in numerous cellular functions, including proliferation.

Patient-derived xenografts.

Patient-derived tumour samples implanted into immunologically compatible mice for investigational purposes.

Endogenous retroviral elements.

DNA sequences encoded (but normally not expressed) by the genome of multiple mammals (including humans and mice) that resemble modern retroviruses.

Micronucleation.

The accumulation of small nuclear structures containing one or a few chromosomes resulting from one or more rounds of defective mitosis.

Viral mimicry.

The state whereby mammalian cells that are not exposed to an exogenous viral infection activate one or more pathways involved in viral resistance, including antigen presentation and interferon signalling.

Senolytic.

An agent that specifically kills senescent cells.

Autophagy.

A conserved mechanism through which all eukaryotic cells can dispose of superfluous or potentially toxic cytosolic material via lysosomal degradation.

Cytokine release syndrome.

A life-threatening condition characterized by systemic inflammation that can arise as a complication of some diseases, or as an adverse effect of some immunotherapeutic agents.