Cancer takes many paths through G1/S

Abstract

In the commonly accepted paradigm for control of the mammalian cell cycle, sequential CDK and cyclin activities drive the orderly transition from G1 to S-phase. However, recent studies using different technological approaches and examining a broad range of cancer cell types are challenging this established paradigm. An alternative model is evolving in which cell cycles utilize different drivers and take different trajectories through the G1-S transition. Moreover, we are discovering that cancer cells in particular can adapt their drivers and trajectories, which has important implications for antiproliferative therapies. These studies have helped to refine an understanding of how CDK inhibition impinges on proliferation and have significance in terms of understanding fundamental features of cell biology and cancer.

The evolving landscape of G1/S control:

A consensus understanding of cancer cell cycles and regulation of the G1/S transition emerged in the mid-1990s [1, 2]. In this adopted linear model of how mammalian cells traverse the G1/S transition, CDK4 and CDK6 (CDK4/6), acting with their cognate cyclins (Cyclin D1, D2, or D3), initiate progression through G1 via partial phosphorylation of the retinoblastoma (RB) tumor suppressor protein. Subsequent activation of CDK2 by Cyclin E is required to drive hyperphosphorylation of RB and progression into S-phase[3, 4]. Despite evidence and likely suspicion that this model was too simple, it has since appeared in a multitude of reviews and textbooks and has been a key guiding principle for innumerable research articles. With the emergence of highly specific CDK4/6 inhibitors[5–7] and their clinical success in ER+ metastatic breast cancer[8, 9], there has been considerable interest in further understanding the G1/S regulatory circuits to enhance therapeutic responses and understand mechanisms of therapeutic resistance. At the same time, innovations in CDK selective inhibitors, CDK and other sensors for live imaging of single cells as they cycle, multiplexed imaging, comprehensive gene-targeting approaches, and phosphoproteomics are being deployed to interrogate nuances of G1/S control (BOX 1). The totality of this work reveals a surprisingly complex coordination of G1/S and considerable deviations from the consensus linear model. Here we will discuss how these findings support a more flexible view of the cell cycle, in which multiple different mechanisms lead to the proliferation of cancer cells. Understanding diverse mechanisms for advancing the cell cycle is critical, as they likely reflect an underlying diversity through which cells in different tissues divide, and they invoke the need for diverse therapies for inhibiting cancer cell proliferation.

BOX 1:

The G1/S regulatory network involves CDK4 and/or CDK6 kinases becoming active leading to the phosphorylation of RB. This event yields the subsequent activation of CDK2, which fully inactivates RB and drives progression into S-phase. Results from a number of technological advancements, by enabling a more detailed analyses of the cell cycle at high-resolution, are challenging this paradigm.

Molecular sensors:

Means to visualize the cell cycle in living cells has had and continues to have a major impact on understanding of G1/S control. The first such sensors (e.g. FUCCI) relied on degradation of differentially fluorescent molecules at the G1/S and G2/M transitions. This approach enables tracking of individual cell fates relative to a successful completion of mitosis, which can be assessed morphologically with the emergence of daughter cells. A second newer set of sensors harbor substrate recognition motifs for specific CDK-cyclin complexes. Typically, such sensors exhibit changes in localization upon phosphorylation and can be followed with live-cell imaging to rapidly define the kinetics of CDK activation or inhibition in a variety of contexts.

CRISPR technologies:

Prior to the advent of broad-based use of CRISPR technologies, other methods (e.g. shRNA or RNAi) suffered from the potential of incomplete depletion of a given gene. Whereas CRISPR is a genetic event, it enables assessment of regulatory bypass or complete cessation of the cell cycle without the ambiguity that can afflict other gene depletion events. Furthermore, highly effective atlases of genetic dependences in a host of cancer cells have been developed that allow for addressing expected events as well as mining for new mediators of cell cycle control.

Single cell sequencing:

The single cell sequencing of cancer cells, normal tissues, and tumor samples enables the ability to develop trajectories based on velocity analyses. Such work can inform how a given normal or tumor cell is progressing through the cell cycle and the relative abundance of a transcript. A caveat of this approach is the incomplete correlation between gene expression and protein activity, albeit it can provide clues as to dominant drivers and expand the analyses to relatively small tissue compartments (e.g. stem cells) that may be challenging to approach through other means.

Multiplexed imaging:

A variety of tools are available to stain tissues or cells with multiple markers at one time. The methods can deploy anywhere from 8–40 markers at one point. Since protein is being measured, it enables the ability to probe for functional features of the cell cycle such as phosphorylation events. The challenge of this approach is the requirement of high-quality antibodies and imaging technologies to de-multiplex images and serially interrogate the same cell.

Phophoproteomics:

Advances in mass spectrometry technology have enabled identification of thousands of CDK phosphorylation sites across the proteome. While some common themes have emerged, for example the high prevalence of multi-site phosphorylation and the overrepresentation of sites in intrinsically disordered protein structures, substrate proteins have remarkably diverse functions. Remaining challenges are to further define substrate specificity among cell cycle CDKs and to determine to what extent there is redundancy among the CDKs.

Intrinsic cell-cycle heterogeneity:

It has been known that certain cancer cells do not conform to the consensus linear model for G1/S control. For example, it was realized years ago that RB-deficient tumors can divide in the absence of CDK4/6 activity[10, 11]. In fact, high expression of the endogenous CDK4/6 inhibitor p16INK4A is found in most RB-deficient tumor cells and clinical samples[12, 13]. The distinction between RB-deficient tumors with low CDK4/6 activity (high p16) and RB-proficient tumors with high CDK4/6 (low p16) already suggested different mechanisms by which cancer cells can subvert inhibition of the G1/S transition. With the use of CDK4/6 inhibitors and genetic screens, it has become clear that there is even more intrinsic heterogeneity than previously understood in how cancer cells, and potentially normal cells, inactivate RB to progress through G1. Recent analyses of the data from the DepMap consortium, which includes over 700 tumor cell lines, illustrates surprisingly diverse vulnerabilities imparted by the deletion of a single CDK or cyclin[14, 15]. These dependencies do not neatly conform to the linear model of cooperating G1/S CDK activity to inactivate RB. In contrast, the DepMap data suggest that at the single gene level there are diverse dependencies in RB-proficient tumor cell lines rather than a relatively common coordinate dependence on CDK4/6, CDK2, and their cognate cyclins (Figure 1). At first glance, this disparity in dependencies appears to track with differing expression of G1/S regulatory CDK, cyclin, and CDK-inhibitor genes in the various cancer cell lines[15, 16].

Figure 1. Differential intrinsic G1/S CDK and cyclin dependencies:

Analyses of DepMap data indicate the requirement of select G1/S CDK and cyclin genes for proliferation. In certain cells there are requirements for two genes that function in a complex, notably CDK6/CCND1, CDK6/CCND2 or D3, and CDK4/CCND1. These vulnerabilities are loosely associated with the corresponding CDK or cyclin gene amplification or the tissue of origin (e.g. CDK6/CCND2 or D3 vulnerability is associated with a hematological malignancies). Tumor cells dependent on CDK2 and CCNE1 are associated with CCNE1 amplification, which occurs frequently in gynecological malignancies. CCND1 can emerge as a vulnerability without a striking dependence on another CDK. This feature could be due to compensatory activity of CDK4 and CDK6, which will be missed in single gene screens, or due to the presence of non-catalytic activities of CCND1. Finally, there is a subset of tumor cells that is enriched for HPV-positivity or RB loss and that does not harbor a single G1/S cyclin or CDK dependence.

The intrinsic differences in cell-cycle drivers are apparent in tumors cell lines, and similar differences have also been observed in clinical samples. It is clear that CCND1, CCNE1, CDK4, CDK6 are subject to relatively frequent amplification, while CDKN2A, RB1, and AMBRA1 represent tumor suppressors that likely impinge on requirements for cell-cycle progression[17–19]. Recently employed multi-marker imaging in clinical samples has described a variety of cell-cycle states. Using cyclic immunofluorescence, heterogeneity has been described as a perturbation from a linear cell cycle and termed incoherence[20] or differential trajectories [21]. These non-canonical cell cycles are observed in a variety of tumor types and are associated with differential effects on survival. Conversely, in other studies, tumors have emerged as being dominated by particular G1/S cyclins (e.g. Cyclin D1 or Cyclin E), which is a pattern that tracks with the different dependencies observed in cell lines[15] (Figure 2).

Figure 2. Diversity of cell cycles in tumor tissue:

(A) Representative images from triple negative breast cancer showing two tumors. The tumor on the left is dominated by the expression of Cyclin E, while that on the right is dominated by Cyclin D1. Both tumors are RB positive and were stained at the same time on the same tissue microarray. (B) Representative image of a metastatic ER+ breast cancer. Eight indicated markers were stained in parallel. This tumor is dominated by Cyclin D1 expression and was responsive to CDK4/6-inhibitor based treatment.

Whether these differences reflect intrinsic heterogeneity in normal cell types or the cell of origin for tumors remains under-studied. It has been known for some time that mouse embryonic stem cells have different cell-cycle control mechanisms than differentiated cells; for example, CDK2 plays a more prominent role than CDK4/6, and the levels of cyclins fluctuate less throughout the cell cycle [22]. Multipotent radial glia cells transition from high expression of B-type cyclins to Cyclin D1 as they differentiate[23]. On the other hand, hematological malignancies exhibit dependencies on CDK6 and CCND2/D3, which is consistent with the role of these genes preferentially in hematopoietic stem cells[8, 24–26]. The extent to which cell-cycle trajectories vary in differentiated cell types has been poorly studied, in large part because the vast majority of cells in differentiated tissue have typically exited the cell cycle.

Adaptation in cell-cycle coordination:

While the presence of intrinsic heterogeneity among cancer cell cycles is becoming more apparent, the extent of cell-cycle adaptation and re-wiring is similarly changing established paradigms related to the relationships among CDK activities and the directionality of the G1/S phase of the cell cycle. The capacity for adaptation may have been anticipated by early genetic mouse models that examined CDK and cyclin knockout. A theme arising from those studies is the redundancy in CDK-cyclin function[25, 27–29], with perhaps the most striking example being that CDK1 is sufficient to complete cell division[30, 31]. It follows from the observations that CDKs can often compensate each other’s activities that cells should be able to adapt to perturbations that downregulate or upregulate specific Cdks and cyclins. However, since cancer cells can have wildly differing levels of CDK, cyclin, and inhibitors due to genetic aberrations, it remains challenging to define consistent expectations.

Indeed, cell cycle re-wiring has been observed upon treatment with specific CDK4/6 inhibitors. In particular, bypassing the requirement for CDK4/6 activity through upregulation of CDK2-Cyclin E or loss of RB expression has been observed as a mechanism of acquired resistance to CDK4/6 inhibitors in the clinic and laboratory[32, 33]. More recently, increasing expression of CDK6 has been observed as another acquired resistance mechanism upon long term use of approved CDK4/6 drugs in cell culture[34, 35]. Notably, the INK family of proteins, normally thought of as CDK inhibitors and growth suppressors, may promote proliferation in this context. In one study, INK4C or INK4B binding to CDK6 did not completely abolish kinase activity, and the residual activity was resistant to CDK4/6 inhibitors and thus contributes to therapeutic resistance[34]. This effect of the INK4 proteins was rescued using CDK6 degraders, suggesting strategies beyond catalytic inhibition may be important in targeting the cell cycle.

Related to the inter-dependency of CDK-cyclin activity in driving G1/S progression, several studies performed with CDK4/6 inhibitors have demonstrated the importance of coordinated downstream suppression of CDK2 activity for cell-cycle inhibition[36–39]. This “coupling” of CDK4/6 to CDK2 requires RB[15, 40], and it is presumably dependent on the dephosphorylation and activation of RB with CDK4/6 inhibition or other mechanisms to limit CDK2 activity [41] (Figure 3A). Other evidence for rewiring has come from recent work that has exploited high-resolution imaging approaches that track the fate of single cells while simultaneously monitoring CDK activity[36, 42–44]. CDK2 selective inhibitors have only recently been developed due to the similarity between the CDK1 and CDK2 catalytic domains [42, 45]. In the context of CDK2 inhibition, the key driver for G1/S control and RB phosphorylation becomes CDK4/6 (Figure 3B) [42]. These studies are reminiscent of older observations in mouse models of CDK2 depletion, in which CDK1 activity becomes important in driving G1/S[30]. Most importantly, the results from these studies argue that both CDK4/6 and CDK2 are capable of hyperphosphorylating and inactivating RB around the G1/S transition in human cells, which would be a prerequisite for an adaptive cell cycle.

Figure 3. Adaptive cell cycle transitions:

(A) CDK4/6 inhibition can elicit the suppression of CDK2 activity. This event is important for the cell-cycle arrest in response to CDK4/6 inhibitors and multiple mechanisms that contribute to resistance to CDK4/6 inhibition uncouple this event to allow for the continuation of CDK2 activity to drive cell cycle progression. (B) CDK2 inhibition will yield transient delays in cell-cycle progression; however, this inhibition is circumvented by the activity of CDK4/6 that promotes RB phosphorylation allowing for the accumulation of Cyclin A. Coordinate inhibition of CDK2 and CDK4/6 in this context potently arrests cellular proliferation. (C) The restriction point was defined in G1 as a point for which mitogenic signaling was no longer required for the onset of DNA-replication. Recent studies indicate that suppression of mitogenic signaling or CDK4/6 activity in S-phase is sufficient to re-set the cell cycle to a G1 state prior to progression through mitosis which associates with reduced Cyclin A expression.

Along with a changing perception of how cells can cross the G1-S transition, the significance of the G1-S transition itself is being questioned. A long-standing tenet of G1/S control is the concept of restriction point, wherein cells no longer require mitogens (growth factors) to advance their cell cycle following a specific point in G1[46]. This concept underlies the acceptance that many oncogene-targeting compounds (e.g. BRAF or EGFR inhibitors) will induce a G1 arrest, presumably because those mitogen-activated signaling proteins are mainly required during transition through G1 and not other phases of the cell cycle. Recent work utilizing single-cell imaging illustrates that mitogenic signaling and CDK4/6 activity are required for enabling progression through G2/M (Figure 3C) [43]. Modeling of this process and the observation of markers such as hypophosphorylated RB suggest that rather than progressing into mitosis, these cells regress into a G0-like state. A provocative lynchpin in adaptive mechanisms and resetting of the cell cycle in G2 is perhaps the expression of Cyclin A, which is controlled by RB and related genes (e.g. p107/p130) and, as a “universally required” gene, can support progression through S-phase and G2/M (Figure 3). Overall, these results are similar to observations of stress responses in G2 that can trigger a pseudo-G1/G0 arrest [21, 47], although they should prompt rethinking of cell-cycle trajectories as unidirectional and irreversible.

Another potential facilitator of cell-cycle rewiring that has not received sufficient attention is the activity of “atypical” CDK-cyclin complexes. In the textbook view of the cell cycle, specific CDKs and cyclins are paired together at certain times. For example CDK4/6 partner exclusively with D-type cyclins, while CDK2 pairs with Cyclin E and Cyclin A. However, in contexts in which CDK4/6 and or CDK2 is depleted, CDK1 has been found in complexes with Cyclin D1 and Cyclin E, and these complexes can drive transition to S-phase[30, 48]. In response to perturbations that yield high levels of Cyclin D, CDK2-Cyclin D complexes have been observed, and it is thought that these complexes may enable resistance to CDK4/6 inhibition [19, 49]. Other complexes, which include canonical CDK inhibitory proteins but are nevertheless active, have also been identified as resistant to CDK4/6 inhibition[34, 50]. A CDK4-Cyclin D complex with phosphorylated KIP1 is incapable of binding the CDK4/6 inhibitor palbociclib, and some preclinical studies have suggested high-levels of tyrosine phosphorylated KIP1 is associated with therapeutic resistance[51, 52]. While many such studies are focused on new roles for known players, it is possible that other less-well established CDKs (e.g. CDK3) or Cyclins (e.g. Cyclin E2) have a more significant impact on G1/S control. More work is needed to understand the biochemical properties of noncanonical CDK-cyclin complexes and the contexts, both in normal and cancer cell cycles, in which they are significant.

Broad mechanisms to block proliferation:

Based on multiple pre-clinical and clinical studies, a myriad of mechanisms can bypass CDK4/6 inhibition, and presumably converse mechanisms will enable escape from CDK2 inhibitors. Whether there are epigenetic or other strategies that could be employed to block such adaptation to “lock in” a dependency remains to be determined, although a number of combination strategies can dramatically enhance the response to CDK4/6 inhibition[32, 53]. Therefore, to develop a durable G1 cell cycle arrest, it may be necessary to inhibit both CDK4/6 and CDK2. Whether such approaches are going to be clinically efficacious or are too toxic is only now being tested. The first agent to harbor such activity, PF-06873600, has potent activity in cell and xenograft models, including those resistant to CDK4/6 inhibitors or CDK2 inhibition[37–39, 42]. However, clinical development has been discontinued (Pfizer.com), ostensibly due to the trade-off between efficacy and toxicity. Several selective CDK2 inhibitors are currently in clinical development, either for Cyclin E amplified tumors or for tumors that have progressed on CDK4/6 inhibitors (Clinicaltrial.gov). Time will tell if combinatorial CDK4/6 and CDK2 inhibitory strategies harbor clinical activity and are tolerable. However, by having separate drugs, it may be possible to alter dosing and optimize scheduling to bypass side-effect features while preventing the division of the tumor cells. The next key test will be if it is at all possible for cancer cells to bypass coordinate inhibition of CDK2 and CDK4/6.

Biomarkers for precise targeting of the cell cycle:

In spite of the relatively established mechanism of action for G1/S acting CDKs, features such as cell-cycle adaptation and heterogeneity represent a challenge for the clinical success of targeting the basic cell division machinery. Based on tumor heterogeneity and dependencies, it may be expected that interrogating cyclin amplification could represent a means to direct care. This supposition is generally supported in DepMap, and other large-scale analyses of cancer cell lines[54]. However, the recent outcomes of an NCI match trial directing treatment of CDK4/6 inhibitors based on CCND1, CCND2, or CCND3 amplification suggest that this strategy is not prudent in the absence of other information or without utilizing some form of combinatorial strategy[55]. One possible issue with assessing cyclin amplifications is the possibility that high-level cyclin expression supports activity of non-canonical complexes (e.g. CDK2-Cyclin D1)[19]. The newer CDK2 inhibitors are being directed towards conditions of CCNE1 amplification (e.g. Clinicaltrials.gov: NCT05252416 and NCT04553133), and whether this approach will be more successful than the experience with CDK4/6 inhibitors remains to be determined.

The population for which it is possible to assess biomarkers with large numbers of patients is in the context of ER+ metastatic breast cancer, for which CDK4/6 inhibitors are standard of care. A number of recently published studies have assessed gene expression and genetic features associated with duration of therapeutic response or progression to the resistant state. RB loss is a relatively infrequent mechanism occurring in ER+ breast cancer, although it is selected for during resistance to CDK4/6. [35, 56–58]. Rather, resistance has been associated with a variety of genes that lessen the ability of CDK4/6 inhibition to fully inhibit RB phosphorylation and/or deregulate CDK2 (e.g. CDK6 deregulation, deregulated mitogenic signaling, gain of Cyclin E)[35, 57–60]. In the Phase I study of abemaciclib, TP53 mutation was strongly associated with resistance to CDK4/6 inhibition[61]. TP53 function is generally not associated with immediate response to CDK4/6 inhibitors in preclinical models[62]; however, recent data has suggested TP53 could contribute selectively to cell-cycle re-entry following the cessation of CDK4/6 inhibitor treatment[63, 64]. Gene expression studies have suggested that deregulation of E2F-activity, which is indicative of significant G1/S deregulation, is associated with a short duration of CDK4/6 inhibitor response [60, 65–68]. Understanding common mechanisms for resistance to CDK4/6 inhibition will likely yield further complexity to how cancer cells can traverse G1/S.

Concluding Remarks:

Deregulated proliferation is a key hallmark of cancer and has been the subject of considerable basic and clinical research. We now better understand that cancer cell cycles are malleable and diverse (Outstanding Questions). This pliability impacts on likely all features of tumor biology. Notably, recent studies have provided a host of biological targets of RB activation on chromatin[69–71], and they have revealed broad-ranging effects of cell-cycle inhibition on not just the cancer cell but the tumor microenvironment[72, 73]. With this rapid evolution in thinking, we must temper expectations that experimental findings can be applied universally to understand the mechanisms underpinning cell division. Experimental context, including what cell or tissue type are studied, how perturbations are made to the cell cycle, and how observations are made, must be considered carefully. While we should no longer expect to find a single mechanism of cell-cycle control, if constants can be identified, they will aid in considering how to target the cell cycle effectively in cancer.

OUTSTANDING QUESTIONS:

What different ways is the cell cycle coordinated in normal tissues and cells?

Heterogeneity of tissue-specific cell cycles is a very much unstudied area of research with broad implications relevant to toxicities of CDK inhibitors and potential applications beyond cancer.

Does the human cancer cell cycle follow any rules?

Given the degree of heterogeneity or adaptation, it will become important to determine if there are constants. Presently, CDK4/6 and CDK2 as co-regulators of G1/S is perhaps one constant; furthermore, it appears that disruption of RB or RB-family member mediated transcriptional repression is required for G1/S progression.

Are CDKs going to continue to be attractive targets in oncology?

The success of CDK4/6 inhibitors in ER+ breast cancer likely ensure interest in CDK inhibitors for the treatment of cancer. The ability to define appropriate patient populations and combinatorial strategies will remain important goals.

Is it possible to “over-ride” a combined blockade of CDK4/6 and CDK2?

A number of preclinical studies support that this combinatorial strategy may provide potent tumor stasis. However, whether adaptations as seen in the mouse to employ CDK1 or utilization of other less well studied CDK-cyclin pairs will need to be determined.

Does CDK inhibition yield new vulnerabilities?

Approaches to target transiently arrested tumor cells or synthetic vulnerabilities related to CDK inhibition have only modestly advanced.

HIGHLIGHTS:

The consensus linear view of G1/S regulation has served the research community for many years.

Recent data indicate considerable heterogeneity in cancer cycles as controlled by distinct CDK and cyclin complexes.

Adaptation and unexpected dependencies reveal emerging themes that support a malleable model for cancer cell division.

The growing complexity of the cell cycle offers challenges and opportunities in rationally targeting proliferative control.