Cell cycle regulation by complex nanomachines

The cell cycle is the essential biological process where one cell replicates its genome and segregates the resulting two copies into the daughter cells during mitosis. Several aspects of this process have fascinated humans since the nineteenth century. Today, the cell cycle is exhaustively investigated because of its profound connections with human diseases and cancer. At the heart of the molecular network controlling the cell cycle, we find the cyclin‐dependent kinases (CDKs) acting as an oscillator to impose an orderly and highly regulated progression through the different cell cycle phases. This oscillator integrates both internal and external signals via a multitude of signalling pathways involving posttranslational modifications including phosphorylation, protein ubiquitination and mechanisms of transcriptional regulation. These tasks are specifically performed by multi‐subunit complexes, which are intensively studied both biochemically and structurally with the aim to unveil mechanistic insights into their molecular function. The scope of this review is to summarise the structural biology of the cell cycle machinery, with specific focus on the core cell cycle machinery involving the CDK‐cyclin oscillator. We highlight the contribution of cryo‐electron microscopy, which has started to revolutionise our understanding of the molecular function and dynamics of the key players of the cell cycle.


Introduction
Orchestrating the cell cycle: switches, rheostats and oscillators The basic biology of multicellular organisms crucially depends on the cell cycle. When a cell enters the cell cycle, it grows, replicates its genome and segregates the replicated sister chromatids into the dividing daughter cells (Fig. 1A). The core mechanism of the cell cycle is a background force in the development of multicellular organisms, tissue homeostasis and tissue repair upon injury. Misregulation of the cell cycle can lead to genetic mutations, which are the basis for human developmental disorders and cancer [1][2][3]. During cell cycle progression, the cell assumes distinct functional states or phases and the transition from one phase to the next is rapid, switch-like and unidirectional [4,5]. Each functional state is characterised by a unique cellular event, which is executed only once per cell cycle. During S-phase, the DNA replication is performed; during mitosis, the genome condenses in the structure of the mitotic chromosomes, and the replicated sister chromatids within each chromosome are segregated by the mitotic spindle into the daughter cells (Fig. 1A). Both the S and M phases are preceded by gap phases where the cell either grows and prepares for DNA synthesis (i.e. G1-phase), or prepares for mitosis (i.e. G2). Influence diagrams of the cell cycle regulators involved are also represented. Relevant macromolecular complexes are schematically represented: SCF and APC/C. The APC/C-coactivator Cdc20 is indicated. Switch-like cell cycle transitions feature a positive feedback loop that is turned off by a negative feedback loop, thereby causing the oscillations in cyclin levels. (B) In a simplified model, the G1/S transition is promoted by the positive feedback loop involving activating E2F1-3, CDK4/6-cyclin D and CDK2cyclin E, which lead to the inactivation of the pRb by phosphorylation and to rising levels of cyclin E. This positive feedback loop is turned off by a negative feedback loop involving repressive E2F6-8 that are transcriptionally activated by the E2F1-3. Decreasing protein levels of E2Fs at the end of S-phase depend on their SCF CyclinF -dependent ubiquitination and subsequent proteasomal degradation. p27 Kip1 inhibits the CDK2-cyclin E complex. (C) The G2/M switch (not shown for simplicity) is activated at the end of a successful S-phase, and results in rising levels of CDK1-cyclin B activity, which stimulates the APC/C activity. Cyclin B oscillation depends on its APC/C-dependant ubiquitination and subsequent proteasomal degradation during mitotic progression. Protein phosphatase 2A (PP2A) counteracts CDK1-cyclin B activity. (D) Cartoon showing the reciprocal regulation between the APC/C and the MCC. The MCC consisting of BubR1, Cdc20, active Mad2 (C-Mad2) [201] and Bub3 is assembled by the SAC at kinetochores which are improperly attached to the mitotic spindle. Two molecules of Cdc20 take part in the APC/C MCC complex, one is the coactivator of the APC/C (Cdc20 APC/C ) and the other is the inhibiting Cdc20 part of the MCC (Cdc20 MCC ). The MCC is both inhibitor and substrate of the APC/C. Skp1-cullin-F-box containing complex (SCF) [10] and the anaphase-promoting complex/cyclosome (APC/C) [11,12] (Fig. 1B,C). Many more E3 ligases have been reported in the context of cell cycle regulation [13], although these findings seem to be confined to tumoral phenotypes, when, for example, the activity of the APC/ C is reduced [14]. E3 ligases catalyse the polyubiquitination of protein substrates, thereby targeting them to the proteasome for proteolysis [15]. Moreover, cyclin genes are subjected to transcriptional regulation by multiple families of regulators including the E2F, the MuvB and the retinoblastoma proteins (pRb) [16]. Ultimately, the activity of the CDKs is regulated by cellular checkpoints, which delay a cell cycle transition until the cellular events of the previous cell cycle phase are successfully completed, in case of DNA damage [17], and within mitosis, in case of either incomplete or wrong assembly of the spindle [18].
Mathematical modelling efforts performed during the last decades reveal that cell cycle transitions are governed by bistable switches [19]. A bistable switch is a system that can adopt only two defined steady states and no intermediate states. In a bistable switch, the transition between the underlying states is characterised by a high degree of irreversibility [4].
Bistable switches govern the cell cycle entry decision (Fig. 1A,B). A cell is committed to divide from a reversible cell cycle arrest state named quiescence or G0 [20] when it is exposed to extrinsic signalling molecules called mitogens. This allows for increased transcription and subsequent increased protein levels of cyclin D that in conjunction with CDK4 and 6 phosphorylate and inactivate the pRb repressor. This repressor binds and inhibits the E2F transcriptional activators, thereby repressing the transcription of cell cycle genes. Once pRb is phosphorylated by CDK4 and 6, the E2F transcriptional programmes for S-phase entry are activated. Once mitogens stimulate the accumulation of a threshold level of CDK activity, the switch to S-phase is triggered irreversibly. This is due to the fact that CDK activity, inhibitory for pRb, is supported by a positive feedback loop where the E2Fs activate the transcription of cyclin E and cyclin A, which further strengthen CDK activity [21,22] (Fig. 1B). Importantly, the positive feedback loop triggering the cell cycle transition is coupled with a negative feedback loop, which completes the oscillation of CDK activity and avoids relicencing of replication origins [23,24] (Fig. 1B).
Entry into mitosis involves the counteracting activity of CDK1-cyclin B and the protein phosphatase PP2A [25] (Fig. 1C), and depends on the successful completion of DNA replication and on the absence of DNA damage, which switches off the G2/M checkpoint [17,26].
The molecular components that operate during mitotic progression and orchestrate the chromosome segregation process include the APC/C and its inhibitor, namely the mitotic checkpoint complex (MCC). The MCC is the effector of the spindle assembly checkpoint (SAC), a signalling cascade starting when kinetochores are improperly attached to the spindle microtubules [18,27-29] (Fig 1A,C,D). Chromosome segregation or anaphase is triggered by the APC/Cmediated ubiquitination and proteasomal degradation of two anaphase inhibitors, namely cyclin B [30] and Securin [31]. Securin degradation leads to the activation of its target Separase, which cleaves the Kleisin subunit of the cohesin complex [32]. This event relieves intrasister chromatid cohesion and triggers sister chromatid separation [33]. Cyclin B and Securin degradation spring up once all the chromosomes achieve correct bipolar attachments to the mitotic spindle. This event switches off the SAC and stops MCC production [18]. The APC/C is irreversibly activated because the MCC, apart from being an inhibitor, is also an APC/ C substrate [34-36] (Fig. 1D). Since the strength of the SAC and its effects on delaying anaphase depend on the number of chromosomes with incorrect attachments to the spindle, the transition to anaphase is more rheostat-like, rather than switch-like [37,38].
In light of the above, biochemically speaking, the underlying components of any cell cycle switch function by establishing various protein-protein interactions, which impact on the subunit composition of the protein complex involved. Each key player of the cell cycle interacts with a multitude of inhibitory or activating modules, thereby forming a multitude of protein complexes with cell cycle-specific functions. An additional layer of regulation comes from the multitude of conformational states which characterises the key players of the cell cycle. For instance, at least four conformational states have been described as functionally relevant for the APC/C complex (reviewed in [39][40][41]). The scope of this review is to highlight the contribution of recent developments in structural determination by cryoelectron microscopy (cryo-EM) in elucidating the role of compositional and conformational heterogeneity in protein complexes regulating the cell cycle and discuss some perspectives in this exciting field. . Strikingly, the latest revolutionary developments in cryo-EM [47,48] are now hugely enriching the field with high-resolution structures of large protein nanomachines where scaffolds, adaptors, substrates and catalytic modules are captured simultaneously while "in action," and while they perform their biochemical function [49]. In cryo-EM, a vitrified specimen is directly imaged on an EM support grid in the vacuum of an electron microscope. The electron beam that passes through the sample produces 2D projection images of fields of particles on a detector in a transmission electron microscope set-up. The recorded electron micrographs are fed into a single-particle analysis pipeline where images of individual particles are aligned together after determining their relative orientation. This produces 2D images with an improved signal-tonoise ratio. When the sample assumes random orientations on the grid, 2D images representative of different views are back-projected unto a 3D volume, which corresponds to the 3D reconstruction of the molecule of interest. The recent developments in both the hardware side and the software side of this pipeline allow the determination of high-resolution 3D reconstructions of the molecule of interest directly in solution, without the need to obtain protein crystals, and without the size limits imposed by NMR structural determination [50]. One major breakthrough in single-particle cryo-EM has been the introduction of direct-electron detection cameras with high detective quantum efficiency (DQE). These cameras can run at high frame rate, thereby allowing the recording of cryo-EM images as movie stacks, which can be corrected for beam-induced motion. This correction alleviates the effects of radiation damage. The other major breakthrough came from the new image processing algorithms, which allow to classify structural heterogeneity in silico and greatly improve the accuracy of image alignment of all the particles used to reconstruct the density map [47].

Main sections
In the following paragraphs, we attempt to summarise the structural biology of the cell cycle core components and highlight, where possible, the contribution of single-particle cryo-EM in helping to decipher the molecular function of this system.

Cyclin-CDK complexes in cell cycle progression
CDK-cyclin complexes form a large family of heterodimeric serine/threonine protein kinases involved in controlling progression through the cell cycle [51], with some families also involved in gene regulation [52]. The CDK activity is counteracted by protein phosphatase complexes, which are reviewed in Ref. [53]. CDK proteins are defined by a catalytic core comprising the ATP-binding pocket, the PSTAIRE-like cyclinbinding domain and an activating T-loop motif [54] ( Fig. 2A). The active site of a monomeric CDK is blocked by a glycine-rich loop in the N-terminal lobe which buckles the activation loop hanging from the Cterminal lobe [55] ( Fig. 2A). Cyclin binding repositions the PSTAIRE motif-containing aC helix that results in a repositioning of the aL12 helix at the start of the activation loop [56] (Fig. 2B). Due to this, the activation loop loosens the interaction with the glycine-rich loop and it opens up. Phosphorylation at Threonine 160 by the master cyclin-activating kinase (CAK) induces melting of the aL12 helix. The latter causes a further rearrangement of the activating loop, which augments its interactions with the cyclin, thereby allowing the formation of a complete active site [57] (Fig. 2C). Although this mechanism of CDK activation is conserved between CDK2-cyclin A and CDK1cyclin B complexes [58], the CDK4-cyclin D complex features a distinct activation mechanism that involves a phosphorylated form of the CDK inhibitor (CKI) p27 Kip . In a landmark study from the Rubin lab, the crystal structure of a CDK4-cyclin D complex with phospho-p27 Kip shows that p27 Kip is required for shaping the CDK4 catalytic site and that nonreceptor tyrosine kinase (NRTK)-mediated phosphorylation of p27 Kip relieves the inhibition on the ATP-binding site. This activation is required for pRb phosphorylation. Importantly, this effect is specific for CDK4, and not for CDK2 where p27 Kip functions exclusively as an inhibitor [59].
Cyclin proteins are defined by a~100-residue-long sequence called cyclin box. This region folds in a conserved domain consisting of one central helix (a3) surrounded by the remaining a1, 2, 4 and 5 (Fig. 2D). Two copies of this structural domain are often present in tandem. The cyclin box is also present in other cell cycle regulators described later in the text such as the retinoblastoma transcriptional regulators [60] (Fig. 2E). More N-terminally to the cyclin box, both S Substrate recognition involves the CDK active site which recognises the consensus S/T-P-x-K/R sequence [61], a conserved patch on the cyclin subunit which is specific for the RxLxF motif [62], and an accessory subunit named Cks which recognises phosphorylated S/T-P motifs [63][64][65][66] (Fig. 2C) [66]. CKIs such as p27 Kip1 compete with CDK substrates by tightly binding to the substrate recognition sites on the CDK and cyclin subunits. Moreover, p27 Kip1 inhibits ATP binding and deform the catalytic site of CDK. [67] ( Fig. 2F). p27 Kip1 is considered a marker of quiescent cells [68]; conversely, p21 Cip1 is associated with quiescence induced after DNA damage occurring during the previous cell cycle [69].
Our current view of CDK activation and substrate recognition for the cell cycle-related CDKs is mainly based on a subset of CDK-cyclin complexes [58, 59,70,71], which are excellent targets for crystallisation in different phosphorylation states and in complex with several types of inhibitors. Extending these studies to other cell cycle regulating CDK complexes and to CDK-substrate complexes will be critical to augment our mechanistic view on substrate recognition and regulation of this class of enzymes during cell cycle progression. Recently, Greber et al. [72,73] demonstrated that cryo-EM methods can be employed to determine high-resolution structures of CDK-cyclin complex refractory to crystallisation. In this study, the structure of the human CAK, comprising CDK7, cyclin H and the assembly factor MAT1, was determined by single-particle cryo-EM. This structure shows that, contrary to the CDK2-cyclin A complex, the C-terminal cyclin fold of cyclin H and the Cterminal lobe of CDK7 are rotated away from each other, thereby vacating a space occupied by MAT1, which stabilises the overall complex assembly. The study has huge potential in expanding the possibilities of anticancer drug design targeting the CDK-cyclin complexes [73][74][75][76]. Moreover, this system can also be employed to study highly dynamic processes such as the CAK-dependent activation of the downstream CDK-cyclin complexes.
Furthermore, a cryo-EM structure of CDK4 in complex with Hsp90-Cdc37 has provided key insights into CDK4 maturation, another highly dynamic process [77].

SCF complexes in interphase regulation
Timely protein degradation of cell cycle regulators depends on the concerted action of multi-subunit E3 ubiquitin ligases belonging to the family of cullin-RING ligases (CRLs). These E3s are responsible for timely 'culling' cyclins, CKIs and a plethora of other proteins via the proteasome, thus allowing cell cycle progression. CRLs account for~20% of all proteasomal degradation [13, 78,79]. During the last 20 years, seminal studies on the CRLs have shed light on their core complex assembly, interactions with substrates and activation/inactivation mechanisms. Due to its key role in cell cycle, early structural studies were focused on a subfamily of CRLs, the Skp1-Cul1-F-Box-protein or SCF E3 ubiquitin ligase which formed the archetypical member of the CRLs [80,81] (Fig. 3A). Cul1 forms an elongated a-helical repeat-containing scaffold where the N terminus interacts with the adaptor module Sphase-kinase-associated protein-1 (Skp1). Skp1 interacts with the F-box domain of one F-box protein (FBP) that act as substrate receptors [82]. The C terminus of Cul1 forms a globular complex with the catalytic RING domain (RBX1) which is responsible for recruiting and activating a ubiquitin-loaded E2 conjugating enzyme (Fig. 3A).
The structural basis of the Cul1-RBX1 interaction involves a Cul1 cavity formed by a four-helical bundle, which extends into an a/b domain nesting an RBX1 bstrand. The latter is further stabilised by two Cul1 winged-helix motifs, namely WHA and WHB (Fig. 3A). In this arrangement, revealed by X-ray crystallography, the active module (the C-terminal Cul1-RBX1 subcomplex) is located more than 50 A apart from the substrate-binding module at the N terminus [82]. Many studies showed that in this orientation, the SCF complex and many other CRLs are in an inactive conformation. This conformation is further stabilised by the binding of another factor, CAND1 [83,84]. CAND1 'hugs' Cul1, thereby sterically inhibiting interactions with Skp1 and the F-box substrate receptors [84].
The activity of CRLs is stimulated by the covalent attachment of NEDD8, a ubiquitin-like modification [85,86]. NEDD8 ligation competes with binding of CAND1, thereby shifting the equilibrium towards an active SCF E3 complex [83,86]. A highly conserved lysine residue within the WHB domain is the target of this reversible NEDD8 modification [86]. NEDD8 induces large conformational changes allowing the RING domain to move closer to the substrate receptor and its bound target protein. A more definite structural model of this was only possible through the use of cryo-EM [83,[86][87][88] (Fig. 3B). In this study, a snapshot of the ubiquitin-transfer onto a phosphorylated substrate recruited through an FBP was obtained, and the key role of covalently attached NEDD8 into inducing large structural changes on Cul1 was finally revealed [87]. NEDD8 acts as a hub of interactions between different parts of the Cul1 protein but also enhances the active E2~Ub closed conformation by making contact with the so-called 'backside' of the E2 [87] (Fig. 3B). These interactions allow the RING domain to become mobile and prime the E2~Ub for transfer onto the juxtaposed substrate. Single-particle cryo-EM has also provided snapshots of how a related multi-subunit complex, the COP9-signalosome (CSN) binds to and removes the NEDD8 moiety from the SCF complex thereby inactivating it [89,90]. SCF complexes rely on nearly 70 FBPs for substrate recruitment, allowing them to polyubiquitinate very diverse protein substrates related to the cell cycle but also to every other aspect of eukaryotic biology [81,91,92]. Apart from the cullin protein being regulated by posttranslational modifications (i.e. NEDD8 attachment), an additional layer of control is exerted on substrate recruitment. FBPs recognise their substrates in multiple ways. Initially, a degron that bears a phosphorylated residue (phosphodegron) was considered the paradigm for SCF substrate recruitment. However, as more FBPs were paired with their respective substrates, it became evident that this rule has many exceptions, some of which are discussed below [92].
The founding member of the FBPs, cyclin F, is the largest cyclin protein, and its protein levels oscillate throughout the cell cycle. As the name suggests, it contains a cyclin domain and, importantly, also a small 40 residue motif which the Elledge lab named F-box [91,93]. This F-box is responsible for the interaction with the adaptor protein Skp1 as part of the SCF [94] (Fig. 3B). Cyclin F differs from the canonical cyclin-CDK paradigm as it does not activate a CDK and plays no role in phosphorylating a substrate [95]. Similar to other cyclins, the cyclin fold of cyclin F interacts with target proteins through an RxLxF-like motif that in this context acts as a degron for polyubiquitination by the SCF-cyclin F complex and subsequent proteasomal degradation [92,95]. The majority of the validated SCF-cyclin F substrates are involved in cell cycle regulation. Among them, we find the APC/Ccoactivator Cdh1, which is also an APC/C substrate [95,96]. This contributes to inactivating the APC/C during late G1-phase and allows for S-phase entry. Interestingly, in a double negative feedback loop control, APC/C is targeting cyclin F for degradation through the Cdh1 component in early G1-phase [96].
Following the discovery of cyclin F, other proteins that contain the F-box at their N terminus were identified. As in the case of cyclin F, these proteins form distinct SCF complexes through a direct interaction with Skp1 [81,92]. FBPs are categorised into three main classes, depending on the structural fold at their respective C terminus which mediates the protein-protein interaction with the E3 ligase substrate. FBXW are FBPs that contain a WD40 domain, FBXL contain a Leu-rich repeat (LRR) and FBXO proteins, including cyclin F, have miscellaneous or 'Other' domains [91] (Fig. 3B).
Skp2 (S-phase kinase-associated protein 2, also known as FBXL1) remains one of the best characterised FBPs, both functionally and structurally. It is well established that Skp2 is overexpressed in numerous human cancers [97]. Skp2 plays a key role in the G1/S transition as it targets CKIs of the Cip/Kip family (namely p21 Cip1 , p27 Kip1 , p57 Kip2 ) that interfere with the activity of cyclin A-, B-, D-, E-dependent kinase complexes [98,99]. In another negative feedback inhibition example, p27 Kip1 is phosphorylated by CDK2-cyclin A/E on Thr187 [100]. This allows recognition of the p27 Kip1 phosphodegron by the SCF Skp2 and leads to its ubiquitination with the help of the accessory subunit Cks1 [99,101]. The crystal structure of the Skp1-Skp2-Cks1 complex bound to the p27 Kip1 phosphodegron showed that Skp2 interacts with Skp1 through a canonical F-box interaction, whereas the LRR motif of Skp2 embraces mainly Cks1 that in turn binds the pThr187 of p27 Kip1 [92,101]. Another Skp2 substrate is retinoblastoma-like protein 2 (RBL2; also known as p130), a retinoblastoma-like 'pocket protein' which binds to and restricts activity of E2F transcription factors [102][103][104]. As quiescent cells re-enter the cell cycle, RBL2 is phosphorylated by cyclin D-CDK4/ 6 leading to a conformational change and release from the RBL2-E2F repressive complexes [102,105]. This phosphorylated form of RBL2 is then recognised by SCF-Skp2 in complex with Cks1 and is subsequently ubiquitinated and degraded [103,104]. Future structural studies will show whether Cks1 plays a similar role in RBL2 recognition by SCF-Skp2 as it does for the recognition of p27 Kip1 .
Another SCF E3 ligase with a distinct mechanism of substrate recruitment is SCF-FBXW7 which polyubiquitinates cyclin E, the activating partner of CDK2, at the G1/S boundary [106,107]. Structural analysis of the Skp1-FBXW7-cyclin E complex shows that the WD40 domain of FBXW7 can recognise two different phosphodegrons of cyclin E: a doubly phosphorylated pThr380/pSer384 peptide and a mono-phosphorylated peptide around pThr62 [108]. Crystal structures of the Skp1-FBXW7 complex bound to the two cyclin E phosphodegrons showed that each phosphopeptide occupies the same binding site on top of the main FBXW7 WD40 cavity [108]. The authors additionally showed that FBXW7 can dimerise, raising an interesting avidity-driven model of E3 ligase action where two SCF-FBXW7 complexes bind one cyclin E substrate through its two phosphodegrons leading to increased processivity [108]. Future studies are required to understand how the two E3s cooperate to effectively ubiquitinate a single substrate and why this is required, given that SCF E3 enzymes are highly processive even when stoichiometric to their substrate. Along this line, a recently published study reveals another insight into this SCF substrate pair and another E3 ligase, named ARIH1, which is required for efficient ubiquitination of cyclin E [109][110][111]. A series of cryo-EM structures show that ARIH1 binds to the Cul1-RBX1 interface creating an E3-E3 superassembly where ARIH1 transfers the ubiquitin to the substrate, while SCF has no catalytic role in this instance [111]. The authors suggest that ARIH1 may compensate for substrates that are bulkier or show less mobility, a situation where a conventional SCF-only catalysis is less effective.
Proteins from the FBXO family [91] remain largely uncharacterised both in terms of their substrates but also structurally. Recent structural studies have revealed unique-binding domains, such as the double b-barrel motifs of FBXO31 that interact with a phosphodegron of cyclin D1 as part of the DNA damage response [112,113]. This structure begins to highlight the structural diversity of FBPs and further hints to the need for further characterisation of other SCF E3s, given the tumour-suppressive or oncogenic functions of many FBPs [114].
Within the last two decades, the majority of the structural knowledge on the SCF E3 ligases came from X-ray crystallography [81]. However, single-particle cryo-EM has recently contributed with further answers to long-standing questions regarding the structural rearrangements and the activation/inactivation mechanisms of the CRL E3 complexes [87][88][89][90]111]. With cryo-EM methods now streamlined and further improving at a fast pace, it is possible that we will soon have further insights into ubiquitination mediated by SCFs. How do different E2s coordinate ubiquitination? How does the conformation of the substratebound SCF complex change to accommodate ubiquitin chain elongation? Does the size and shape of the substrate affect SCF polyubiquitination and is there any correlation with the FBPs? What is the role of SCF oligomerisation in enhancing substrate ubiquitination? What is the optimal positioning (if there is an optimum) and number of lysines at the ubiquitination sites? Why very often only one partner in a multiprotein complex gets ubiquitinated and degraded, while the others remain untouched? These are some of the questions we anticipate will be answered in the near future and might further help in the development of novel therapeutics, including proteolysis-targeting chimeras that bridge E3s with non-native, medically important substrates for degradation [115].

APC/C complexes in mitotic progression
The APC/C is an unusually large cullin-RING E3 ubiquitin ligase. Structural efforts performed during the last decades, which involved also the latest developments in cryo-EM unveiled an extremely complex and dynamic cell cycle regulator [39][40][41]116]. Similar to the SCF complex, the APC/C features a catalytic module comprising a cullin subunit (Apc2) which binds a RING domain protein (Apc11) via its C terminus [117] and a substrate receptor protein named coactivator involved in substrate recognition [39,118] (Fig. 3A-D). Cdc20 is the coactivator required for recognising the APC/C mitotic substrates cyclin B and Securin. Conversely, the coactivator Cdh1 is required for recognising substrates that are degraded later in the cell cycle, such as Cdc20 during mitotic exit and cyclin F in G1-phase [96]. The former is essential for the execution of mitotic exit, establishment of G1 and quiescence (G0) [119][120][121]. The molecular details on how the APC/C-coactivator complexes recognise target substrates have been reviewed extensively [39,46], and it involves specific SLiMs on target substrates, namely the D box [8], KEN box [122] and ABBA motifs [123,124]. These motifs are recognised by distinct patches on the coactivator WD40 domain [39]. The D box is recognised by a coreceptor involving the coactivator and the Apc10 core subunit [125]. Another SLiM named MR motif is coactivator-independent [126][127][128][129]. Similarly to SCF, the APC/C recognises phosphodegron motifs as reviewed in Ref. [39]. When comparing the APC/C with other E3 ligases, it is quite striking to notice that the adaptor module connecting the catalytic site with the substrate receptor is an unusually large and complicated protein scaffold forming a central cavity [117] (Fig. 3C). This adaptor module allows substrate recognition and catalysis to happen in a confined and highly regulated environment [39, [130][131][132]. The APC/C adaptor module is composed of a socalled tetratricopeptide repeat (TPR) [133] lobe and a platform module (Fig. 3C). The platform module includes the largest subunit of the APC/C called Apc1, which, together with Apc4, recruits the catalytic module [117,128] (Fig. 3C). The TPR lobe is composed of TPR-containing homodimers which stack on top of each other, thereby forming a TPR superhelix. The TPR lobe, in synergy with the platform, recruits the coactivator subunit via specific 'hook'-like dipeptide motifs containing an arginine and a hydrophobic residue [39,46]. One of these motifs, namely C box, is located on the N-terminal domain (NTD) of the coactivator and binds the APC/C in a cleft, right at the centre of the APC/C cavity between Apc1 and the Apc8 homodimer B, at the interface between platform and TPR lobe [117] (Fig. 3C). As shown by cryo-EM, this interaction induces a remarkable conformational change on the catalytic site. Here, the Apc2 C terminus and Apc11 are lifted upwards from the platform, thereby exposing the recruitment site for the E2 enzyme [132,134] (Fig. 3C,D). As in the SCF complex, the E2 contacts the C-terminal WHB domain of the cullin subunit Apc2 and the RING domain of Apc11 (Fig. 3A-D).
Recent cryo-EM studies have shown that the complex of APC/C with the mitotic coactivator Cdc20 and its inhibitor complex MCC features striking conformational flexibilities [131,135]. In one conformation named APC/C MCC closed, the MCC blocks substrate recognition by obstructing the APC/C substrate recognition sites with specific pseudosubstrate sequences present in the MCC protein named BubR1 (Figs 1D  and 3D,E). Furthermore, in APC/C MCC closed, the MCC obstructs the E2 enzyme-binding site thereby competing with the E2 and inhibiting ubiquitin chain initiation (Fig. 3D,E). In another conformer named APC/C MCC open (Fig. 3F), the MCC module is rotated in such a way that the catalytic inhibition is released and the binding of both the MCC and the E2 becomes possible within the same molecule of APC/C. In this complex, the MCC switches from an APC/C inhibitor to an APC/C substrate [131,135]. Pioneering work has demonstrated that the MCC is indeed ubiquitinated by the APC/C during mitosis in a cellular context and that this process depends on the small subunit called Apc15 [34-36] (Fig 3C). Removal of Apc15 locks APC/C MCC in closed conformation [131,135] and prolongs mitosis [36]. These data support a fascinating model where the APC/C MCC enzyme is not fully inhibited. The APC/C MCC is capable of binding the E2 enzyme and promoting the ubiquitination of its own inhibitor MCC. This allows the APC/C to act as a 'sensor' of the MCC levels produced at kinetochores, where the APC/C activity against its mitotic substrates is delayed by the MCC, which functions as a competitive inhibitor substrate, until there are no unattached kinetochores left. Once the attachments are complete, the APC/C will extensively ubiquitinate the MCC, which leads to MCC disassembly [136]. Then, the APC/C-Cdc20 will readily ubiquitinate cyclin B and Securin for triggering anaphase (Fig. 1A,D). Even though this model is tempting and would help explain how anaphase is synchronised with the assembly of the spindle, it lacks experimental validation in vivo. For example, it will be critical to assess whether mutations that promote the APC/C MCC open state could accelerate mitosis in an opposite fashion to Apc15 depletions. It is also critical to understand how the APC/C MCC closed to open transition is regulated in the cell by post-translational modifications such as phosphorylation or SUMOylation.
In conclusion, cryo-EM has been an excellent tool for exploring the APC/C complex assembly and conformational variability. The cryo-EM studies on the APC/C MCC represent an example of how EM allows to visualise multiple conformations within a single sample preparation, thereby providing snapshots of a 'nanomachine in action'. These structural data can now be used for generating new hypotheses to probe the APC/C function during mitotic progression in vivo.

Transcriptional complexes in cell cycle regulation
The majority of cell cycle-specific genes are transcriptionally regulated by multiple families of regulators including the E2F, B-MYB, FOXM1 transcription factors, the retinoblastoma 'pocket proteins' and the multivulva class B (MuvB) complex [16,23]. There are eight human E2F1-E2F8 genes which encode several protein products due to expression from alternative promoters (such as E2F3a and b) and alternative splicing (E2F7) [23,137]. Every E2F protein contains a conserved DNA-binding domain (DBD) that engages its target gene promoters at a similar consensus DNA sequence. E2Fs can be classified based on their transcriptional capacity, with some being activators (E2F1, E2F2, E2F3a) and some repressors (E2F3b, E2F4-E2F8) [23, [137][138][139]. In a rather simplified model of E2F function, oscillations in cell cycle-regulated gene expression derive from sequential action of both E2F activators and suppressors [137,140] (Fig. 1B).
The E2F transcriptional oscillator is intertwined with the protein oscillator dictated by the SCF and APC/C E3s [141]. The SCF-cyclin F E3 ligase is responsible for degrading the transcriptional activators E2F1-3a in late S-phase and after DNA replication, whereas the APC/C-Cdh1 targets the repressive E2F7-8 during G1, thereby allowing levels of E2F1-E2F3a to increase and initiate another round of DNA replication [141][142][143][144]. Furthermore, targeting of E2F7-8 during G2 by SCF-cyclin F is required for DNA repair after DNA replication [145].
Additionally, the activity of E2Fs is modulated by protein-protein interactions. E2F1-6 bind to dimerization  [137,139,141]. Importantly, E2F1-E2F5 proteins also contain a transactivation domain that either recruits transcriptional coactivators or one of the three RB proteins namely pRB, and the RB-like proteins RBL1/p107 or RBL2/ p130. RB proteins mask the E2F transactivation domain, thereby repressing its transcriptional activating function [137,146,147]. The RB family of proteins are key components of the cell cycle. However, their bistable switch controlling E2F-mediated transcription and cell cycle progression [102,148] is often inactivated in cancerous cells [149]. All three RB proteins share a similar architecture, with a central domain that is termed 'pocket' region because, within this fold, a pocket embeds the E2F transactivation domain. The pocket is flanked by an N-terminal (RB-N) [150] and a Cterminal domain (RB-C) [151] (Figs 2E and 4B). Both the pocket domain and the RB-N are structurally composed of two flexibly linked subdomains, A and B, which bear structural homology to the cyclin box fold [150,152] (Fig. 2E). The pocket domain interacts with a plethora of RB-binding proteins, including the E2F transactivation domain of the E2F-DP complex, the Cdh1 substrate receptor of the APC/C and, importantly, proteins that contain a short LxCxE motif such as cyclin D, histone deacetylases and chromatin remodelling complexes [146,147,151,[153][154][155][156][157] (Figs 2E and  4B). The RB-N regulates the protein-protein interaction capabilities of the pocket domain upon phosphorylation [105,150]. The RB-C domain mediates further interactions with the E2F1-DP1 heterodimer (Fig. 4B). Furthermore, RB proteins recruit the CDK4/6-cyclin D kinase at G1-phase via an RxLxF motif and a distinct C-terminal helix [157,158]. It is estimated that a third of the RB protein sequence lacks stable secondary structure elements. Several disordered segments of RB are targeted by CDK-cyclin-dependent phosphorylation at multiple sites [151]. When RB proteins are hyperphosphorylated, RB-N interacts with the pocket domain thereby masking the LxCxE-binding site and indirectly inactivating the E2F-binding side. This inhibits RB-E2F complex formation and promotes E2F activity on its target genes [105,151,158]. E2F and RB proteins present different affinities for each other: E2F1-E2F3 interact almost exclusively with pRB, whereas E2F4 binds to RBL1 and RBL2 (but also to pRB in some cell types) and E2F5 associates with RBL2 [23, 137,159,160]. The RB-E2F complex formation varies throughout the cell cycle. pRb-E2F complexes are most abundant when cells proceed from G1phase to S-phase, whereas RBL1-E2F complexes are most prevalent in S-phase. RBL2-E2F complexes are present mainly during quiescence and early G1-phase [23,161].
Strikingly, when cells exit the cell cycle and enter quiescence, a massive transcriptional suppression of around 1000 cell cycle-specific genes occurs. This is achieved by the repressive activity of the DREAM (DP, RB-like, E2F and MuvB) complex, which is formed through the association of the RBL2-E2F-DP module with another multiprotein complex, the MuvB (multivulva class B), containing LIN54, LIN9, LIN37, LIN52 and RBBP4 [16,162]. The DREAM complex is recruited to target gene promoters through E2F-DP, which binds to a cell cycle-dependent element (CDE), and LIN54, which binds to the downstream cell cycle genes homology region (CHR) [16, [162][163][164][165]. Structures of several cell cycle transcription factor DBDs in complex with target DNA sequences have been determined and provide insights into DNA consensus recognition. For example, structures are known for the E2F4-DP2 DBD [166], the E2F8 DBD [167] and the LIN54 DBD [168] (Fig. 4). The interaction between E2F-DP-RBL complex and MuvB is mediated through the smallest subunit LIN52 when phosphorylated by DYRK1A [169]. Biochemical and structural analyses of this interaction revealed that LIN52 binds to the LxCxE cleft of the pocket domain of RBL proteins using the combination of a linear peptide containing a suboptimal LxSxExL sequence and a phosphate at a nearby Ser28 of LIN52 [170] (Figs 2E and 4B). Importantly, the absence of the phosphate-interacting region on the pRb protein explains why MuvB complex specifically assembles with RB-like proteins RBL1-RBL2 and not pRb [170]. The functions of LIN9 and LIN37 are less understood, both functionally and structurally, with initial hints pointing towards LIN9 being a structural component essential for DREAM complex formation, and LIN37 being actively required for gene repression [164,171,172]. RBBP4 is a subunit shared among several chromatin-regulating complexes, and it has been proposed to mediate the interactions between DREAM and chromatin [173,174] (Fig. 4B).
Recent studies have established the key role of the DREAM complex as a master regulator of cell cycledependent gene expression and have shown that perturbations in DREAM complex formation are frequently observed in cancers [16,175]. Despite the importance of the DREAM complex, its precise molecular function remains unclear. Unlike chromatin remodelling complexes, the eight-subunit DREAM complex lacks any obvious catalytic subunits posing the question of how it exactly enforces transcriptional repression. Currently, there are two proposed models of action for the DREAM complex. In the first, the DREAM complex acts as a bridging factor to recruit protein complexes with enzymatic activity. In support of this model, the DREAM complex was shown to interact with Sin3B, a scaffolding protein interacting with HDAC1 (histone deacetylase 1) that has wellestablished roles in chromatin compaction and subsequent gene repression [120]. In the second model, the DREAM complex could participate in nucleosome assembly and positioning at gene bodies, a process that can occlude RNA polymerase binding and transcription initiation [176].
Another unanswered question is how the same MuvB complex can switch from a quiescence-specific repressor to an activator during cell proliferation. This process involves the dissociation of RB-like proteins from MuvB, which requires the CDK-cyclindependent phosphorylation of RBL proteins upon cell cycle entry [161] (Fig. 4A,C). Furthermore, during Sphase, MuvB binds to the B-MYB transcription factor and reassociates with the promoters of a subset of DREAM target genes including mitotic genes [165,177]. Binding of B-MYB to MuvB stimulates recruitment of FOXM1 [177,178]. In late S-phase, B-MYB is phosphorylated by CDK2-cyclin A kinase, which leads to B-MYB ubiquitination by the SCF-Skp2 E3 ligase and proteasomal degradation. Furthermore, FOXM1 is phosphorylated by Plk1. The latter events stimulate the MuvB-FOXM1-dependant transcriptional activation of genes required in the G2/Mphase [179][180][181][182]. To date, it is not understood how B-MYB mediates the recruitment of FOXM1 onto the MuvB complex, but also why B-MYB needs to be degraded for transcriptional activation [177,178]. The crystal structure of the domains mediating the MuvB-B-MYB interaction revealed a coiled-coil interaction between LIN9 and LIN52 and a B-MYB helix-turnhelix peptide that binds across this coiled-coil interface [172]. Importantly, this structure shows that the RBL2 and the B-MYB-binding sites on LIN52 do not overlap, pointing to a more complex mechanism of MuvB complex switching from a transcription repressor to an activator [170,172].
In summary, the mechanistic role of MuvB and the precise function of its complexes in regulating cell cycle-dependant transcription remain largely unresolved. Structural studies by X-ray crystallography have been limited to isolated modules of this complex system, as many of these proteins have significant regions of disorder. We envision that cryo-EM has great potential to help in deciphering the structures and the dynamics of these complexes, and in establishing structure-function links that will elucidate their precise mechanism of action. In fact, related chromatin-associated protein complexes have been successfully resolved by cryo-EM as reviewed in Ref. [183]. Among these complexes, we highlight the polycomb repressive complex 2 [184][185][186], the mixed lineage leukaemia complex [187][188][189] and the nucleosome remodelling and deacetylase complex [190]. Intriguingly, all these complexes share the RBBP4 core subunit with the MuvB complex, and therefore, it is tempting to think that some of the mechanisms of assembly and chromatin engagement may be conserved among all of these complexes.

Conclusion
The eukaryotic cell cycle is a complex process, which is central to the basic biology of multicellular organisms, and its misregulation lies at the root of several human diseases including cancer. At the heart of this process, the biological oscillator system of the CDKcyclin kinases is controlled at both the gene level and the protein level by transcription factors and E3 ubiquitin ligases, respectively. During the last decades, high-resolution structural studies on this system have largely relied on X-ray crystallography and NMR to reveal intricate details about the key proteins controlling the cell cycle proteome, actively contributing to our current understanding of different cell cycle events. With cryo-EM now being an integral part of the structural biology toolkit, it is possible to explore these proteins as part of larger assemblies and gain insights into their dynamic nature required for biological activity. Due to the latest advances in both hardware and software, the 'fuzzy' picture of these nanomachines has started to clear up, revealing new high-resolution structures and offering new avenues for structurebased drug design. Nevertheless, single-particle cryo-EM is still limited by the intrinsic flexibility within the protein assemblies. Therefore, X-ray crystallography, NMR and other complementary methods including cross-linking mass spectrometry studies converge together with singleparticle cryo-EM in an integrative structural biology effort [191][192][193]. In addition, tools that allow visualisation of these complexes in their native cellular environment under near-physiological conditions are also emerging very rapidly. Cryo-electron tomography (cryo-TEM) and correlative light and electron microscopy techniques are providing an exciting new era for structural biology [194][195][196][197]. Ultimately, time-resolved structural studies on cell cycle regulators will be essential to investigate intermediates of reaction, assembly and remodelling involving these systems [198][199][200].