Cells on lockdown: long‐term consequences of CDK4/6 inhibition

Inhibition of cyclin‐dependent kinases Cdk4/6 is emerging as a useful anti‐proliferative chemotherapy, but it remains unclear how durable inhibition of cancer cell proliferation is achieved to promote a long‐lasting response in patients, or how toxicity is limited to cancer cells with minimal side effects. Two recent papers in The EMBO Journal investigating senescence induction following prolonged Cdk4/6 inhibitor treatment now reveal important insights into ways to increase anti‐tumour effects of Cdk4/6 inhibition and to reduce therapy‐induced side effects of senescence induction.

New studies reveal how G1‐blocking drugs like palbociclib arrest cancer cell proliferation at various stages and induce senescence with reduced toxic side effects.

The Saurin and Demaria groups set out with differing initial goals: Crozier et al (2022) focussed on identifying the mechanisms by which treatment with Cdk4/6 inhibitors (Cdk4/6i) promoted entry into senescence, while Wang et al (2022) aimed to characterise the downstream impact of senescence induction by Cdk4/6i. Both laboratories found that treating non‐cancer cells for a prolonged period of time (i.e. 7–8 days) with Cdk4/6i, followed by drug removal, caused cells to enter senescence, dependent on the tumour suppressor p53. Crozier et al discovered that this could occur in one of two ways: Cells either remain in a G0/G1‐like arrested state after drug withdrawal, or cells re‐enter S‐phase but arrest in G2 (Fig 1). Wang et al found that senescent cells induced by long‐term Cdk4/6i treatment secrete a unique set of proteins that lack pro‐tumorigenic properties and promote clearance of Cdk4/6i senescent cells by the immune system (Fig 1). Together, both studies give us exciting insights into how Cdk4/6i act to promote long‐term effects.

Figure 1. Long‐term Cdk4/6i treatment causes cells to enter into p53‐dependent senescence.

Cdk4/6i addition to proliferating cells causes them to arrest in a G0/G1 state. Prolonged Cdk4/6i treatment (7–8 days) leads to activation of p53 and its downstream targets and downregulation of proteins required for DNA replication origin licensing. After Cdk4/6i withdrawal, cells either (i) remain in a G0/G1‐arrested senescent state or (ii) re‐enter S‐phase with under‐licensed replication origins, generating replication stress and arrest in the subsequent G2. Both G0/G1‐ and G2‐arrested states are p53‐dependent. Cdk4/6i‐induced senescent cells secrete a set of p53‐associated proteins, the PASP. Unlike the SASP, the PASP lacks the ability to promote tumour cell proliferation, but signals to the immune system prompting the clearance of these senescent cells.

While p53 is not required for cells to enter or maintain a cell cycle arrest in the continued presence of Cdk4/6i, both laboratories found that it is essential for sustaining arrest after drug removal. In the presence of Cdk4/6i, p53 levels in the nucleus increase, as does expression of its transcriptional targets, including the CDK inhibitor, p21. Knocking out p53 allowed cells to re‐enter the cell cycle after Cdk4/6i removal. Using live single‐cell imaging, the Saurin laboratory showed that in the presence of p53, cells after Cdk4/6i washout either remain in G0/G1 or re‐enter S‐phase, complete DNA replication and arrest in the subsequent G2, where they withdraw from the cell cycle (Crozier et al, 2022). In either phase, arrest requires p53. They astutely observed that many of these cells displayed hallmarks of replication stress‐induced DNA damage. Proteomics analyses of Cdk4/6i‐arrested cells revealed that the proteins most downregulated after Cdk4/6i addition were components of the MCM2‐7 complex, required for DNA replication origin licensing. Many other DNA replication proteins were also downregulated. Licensing a sufficient number of genomic replication origins is essential to efficiently replicate the entire genome during the course of S‐phase (Limas & Cook, 2019). Indeed, Crozier et al found that after release from seven‐day Cdk4/6i treatment, DNA origin licensing was significantly reduced, thus identifying a source of replication stress. A recent study (Wander et al, 2020) identified TP53 alterations in approximately 60% of Cdk4/6i‐resistant breast cancer patients, but ruled out TP53 mutations as contributing to Cdk4/6i resistance since MCF7 cells, a breast cancer line with wild‐type p53 expression, arrested equally well during Cdk4/6i treatment in the presence or absence of p53. However, p53 alterations could prevent the sustained cell cycle withdrawal of cancer cells after Cdk4/6i treatment has ended, which could drive resistance.

An obvious question arising from both the Demaria and Saurin studies is what activates p53 during long‐term Cdk4/6i arrest and release? Replication stress, such as is generated due to the under‐licensing of replication origins, can activate p53 (Barr et al, 2017), and this likely contributes to the sustained G2 arrest in cells that re‐enter S‐phase after Cdk4/6i removal. However, how p53 is activated during a G0/G1 arrest, in the presence of Cdk4/6i, and how p53 is maintained after inhibitor washout, remains unknown. Increased levels of mitochondrial reactive oxygen species (ROS) in cells treated with Cdk4/6i may help to promote p53 activation, although this does not involve ROS‐mediated DNA damage (Wang et al, 2022). Determining the source of p53 activation would allow us to intelligently predict Cdk4/6i combination therapies to ensure that this particular p53 response is not diminished. It also remains unknown why some cells re‐enter S‐phase after Cdk4/6i washout, while others remain in a G0/G1‐arrested state. It was recently shown that the p53‐target, p21, is not required to maintain a G0/G1 arrest during Cdk4/6i treatment (Pennycook & Barr, 2021). However, based on the observations by the Saurin and Demaria teams, we would predict that p21, downstream of p53 activation, is required for the sustained cell cycle arrest phenotypes after Cdk4/6i washout—in which case, heterogeneity in p21 expression could perhaps contribute to whether cells re‐enter S‐phase or not.

Having elucidated these pathways in non‐transformed cells, the Saurin laboratory moved on to cancer cells and determined that MCF7 exhibited a similar downregulation of replisome components during arrest with Cdk4/6i. Examining a panel of p53 wild‐type and p53‐null cancer cell lines, Crozier et al (2022) showed that although G0/G1 arrest in the presence of Cdk4/6i was, on average, less complete in cancer cells than in non‐transformed cells, replication stress‐related DNA damage was still increased after Cdk4/6i washout, indicating that a complete G0/G1 arrest is not an absolute prerequisite. In fact, p53‐null cancer cells also accumulated DNA damage during Cdk4/6i treatment due to incomplete inhibition of cell proliferation generating replication stress. This suggests that irrespective of p53 status, Cdk4/6i can cause genotoxic stress in cancer cells. It would be interesting to further compare the mechanisms of replication stress induction between normal and cancer cells, and whether it still involved replisome downregulation in cancer cells, despite their lack of a robust G0/G1 arrest. As suggested by Crozier et al, combinations of Cdk4/6i and replication stress‐inducing agents could enhance the effects of Cdk4/6i treatment in cancer patients. Significantly, replication stress‐inducing agents should be administered after Cdk4/6i treatment, in contrast to the situation for paclitaxel and Cdk4/6i co‐treatment, where paclitaxel should be administered first and Cdk4/6i second, in order to prevent treatment‐induced DNA damage from being repaired by cancer cells (Salvador‐Barbero et al, 2020). Again, these studies demonstrate how understanding the detailed mechanisms of drug action on cells can guide their best use in the clinic.

Seeking to understand how the toxicity of Cdk4/6i is limited to cancer cells, Wang et al (2022) used in vivo experiments in mice to determine whether Cdk4/6i leads to senescence in non‐transformed cells. They treated cancer‐free mice with either Cdk4/6i or doxorubicin—a common DNA damage‐inducing chemotherapeutic—at doses normally capable of reducing tumour volume. Both drugs induced similar levels of senescence, as measured by an in vivo p16 luciferase reporter. The fact that downregulating p53 via PFT‐a, a p53 transcriptional inhibitor, prevented p16 upregulation by Cdk4/6i, suggests that senescence is bypassed in the absence of p53 in vivo.

Despite both agents inducing senescence in vivo, Cdk4/6i were less toxic to mice than doxorubicin. These studies were performed in cancer‐free mice, and it would be interesting to see how this compared to mice with tumours. To understand the mechanism for these differences in toxicity, Wang et al (2022) looked into the composition of the senescence‐associated secretory phenotype (SASP), a well‐characterised set of factors secreted by senescent cells. DNA damage responses induced by doxorubicin lead to the activation of a key signalling nexus, NF‐ϰB, which promotes a pro‐inflammatory SASP. This SASP has both pro‐ and anti‐tumourigenic effects by, for example, either promoting tumour cell proliferation or epithelial–mesenchymal transitions, or by initiating the clearance of senescent cells via the immune system, respectively (Birch & Gil, 2020). During Cdk4/6i treatment, DNA damage was not induced to the same level as by doxorubicin treatment, NF‐ϰB activation was not observed, and key components of the SASP were not secreted. Wang et al were also unable to detect typical NF‐ϰB‐dependent SASP components in plasma samples from breast cancer patients treated with Cdk4/6i, as compared to patients treated with paclitaxel. What the Demaria laboratory did identify, however, was a more limited set of proteins (including ISG15, IGFBP3, GDF15 and LIF) secreted by Cdk4/6i‐treated cells, which they termed the p53‐associated secretory phenotype (or PASP). The authors confirmed the absence of pro‐tumourigenic effects of Cdk4/6i‐induced senescence in vivo, by showing that growth of injected cancer cells was promoted by co‐injection of doxorubicin‐induced senescent cells, but not by co‐injection of Cdk4/6i‐induced senescent cells. Similar effects were observed in an orthotopic mouse model, where mice were pre‐treated with either doxorubicin or Cdk4/6i and then injected with murine breast cancer cells. Only doxorubicin pre‐treatment led to increased tumour growth and shorter survival periods. Intriguingly, the lack of SASP in Cdk4/6i‐induced senescent cells did not impair the clearance of murine senescent cells, which were in fact cleared faster than their doxorubicin counterparts.

A significant fraction (13 out of 32) of the secreted proteins Wang et al found upregulated after Cdk4/6i treatment in BJ human fibroblasts have also been reported to be upregulated after sustained p21 overexpression in mouse embryonic fibroblasts (Sturmlechner et al, 2021). This latter work identified a p21‐dependent secretome (also abbreviated PASP for p21‐activated secretory phenotype, but not to be confused with the p53‐associated secretory phenotype of Wang et al). The p21‐associated secretory phenotype was shown to promote clearance of stressed cells, in cases where p21 expression remained high for sustained periods. Taken together, it seems likely that the secretory response observed after long‐term Cdk4/6i treatment may be p21‐dependent, downstream of p53 activation.

The study by Crozier et al (2022) highlights a source of heterogeneity in the induction of senescence/long‐term arrest in cells after CDK4/6i washout—with cells arrested in either a G0/G1‐ or a G2‐like state. This has implications for the secretome phenotype identified by Wang et al (2022). Do both G0/G1‐ and G2‐arrested populations secrete the same factors? If not, does this affect how efficiently each population would be cleared by immune cells? Further investigation of these phenotypes may shed light on why some tumours respond better to Cdk4/6i treatment than others.

Finally, the identified secretome phenotypes described here have been characterised in fibroblasts. Future work should focus on whether cancer cells treated with Cdk4/6i display a similar secretory response, if this varies between tumour types, and how drugs used in combination with Cdk4/6i (e.g. MEK or PI3K inhibitors) may modify the repertoire of proteins secreted. This could start to give us some clues as to why some tumours respond better than others to Cdk4/6i treatment. Treatment of cancer‐free mice with Cdk4/6i appeared to only induce senescent‐like phenotypes in a fraction of cells (Wang et al, 2022). Identifying which normal cells and tissues are sensitive to senescence induction after Cdk4/6i treatment may help for understanding the effectiveness of Cdk4/6 inhibitors in patients.

The EMBO Journal (2022) 41: e110764.