The cell division cycle is driven by a collection of enzymes that coordinate DNA duplication and separation, ensuring that genomic information is faithfully and perpetually maintained. The activity of the effector proteins that perform and coordinate these biological processes oscillates by regulated expression and/or posttranslational modifications. Ubiquitylation is a cardinal cellular modification and is long known for driving cell cycle transitions. In this review, we emphasize emerging concepts of how ubiquitylation brings the necessary dynamicity and plasticity that underlie the processes of DNA replication and mitosis. New studies, often focusing on the regulation of chromosomal proteins like DNA polymerases or kinetochore kinases, are demonstrating that ubiquitylation is a versatile modification that can be used to fine-tune these cell cycle events, frequently through processes that do not involve proteasomal degradation. Understanding how the increasing variety of identified ubiquitin signals are transduced will allow us to develop a deeper mechanistic perception of how the multiple factors come together to faithfully propagate genomic information. Here, we discuss these and additional conceptual challenges that are currently under study toward understanding how ubiquitin governs cell cycle regulation.
Cell proliferation is a continuous cycle of DNA synthesis and subsequent chromosome separation. Posttranslational modifications of effector proteins ensure that these major events and their transitions are orchestrated so that genomic information is preserved. The covalent conjugation of the small protein ubiquitin through a process called ubiquitylation plays a critical role in the overall regulation of cell division. It is well established that ubiquitylation is a signal for protein degradation by the proteasome (Fig. 1, A and B), with special importance in assuring ordered and well-timed cell cycle transitions (Teixeira and Reed, 2013; Bassermann et al., 2014). However, ubiquitylation is not necessarily linked to protein degradation, and in recent years, an increasing number of nonproteolytic outcomes of protein ubiquitylation have been reported to play important cellular roles (Komander and Rape, 2012). Proteasome-independent regulation of an ubiquitylation target is achieved by changes in protein–protein interactions, subcellular localization, or enzyme activity (Fig. 1 B). As opposed to the irreversible fate of degradation, nonproteolytic outcomes of ubiquitylation allow for functional fine-tuning, dynamically and reversibly responding to intracellular cues instead of requiring de novo protein synthesis.
Ubiquitin conjugation to its targets requires the concerted action of an E1 ubiquitin-activating enzyme, E2 ubiquitin-conjugating enzyme, and E3 ubiquitin ligase. The latter binds specifically to the substrate and promotes the transfer of ubiquitin to one of its lysine residues (see text box for an overview of E3 ligases involved in cell cycle regulation). Because of multiple reactive sites on ubiquitin, more moieties may be added, establishing complex oligomers or chains (Fig. 1 A). This enables that multiple ubiquitin topologies generate individual signals, which are collectively referred to as the ubiquitin code (Komander and Rape, 2012). This code is read by downstream factors containing ubiquitin-binding domains, referred to as readers or decoders, which specifically recognize the chain topology and induce the appropriate signal (Husnjak and Dikic, 2012). For example, a polyubiquitin chain in which ubiquitin conjugates via its lysine-48 (K48) and/or K11 residues is read and as a result rapidly degraded by the 26S proteasome, an irreversible process that is often observed in cell cycle transitions (Grice and Nathan, 2016). Conversely, a monoubiquitin moiety or K63-linked chain can recruit factors that allow for a specific localized response, such as the recruitment of a DNA damage–tolerant polymerase to a site of replication stress (García-Rodríguez et al., 2016). In many cases, ubiquitylated proteins first need to be extracted from interacting partners or chromatin, a function typically attributed to the ATPase valosin-containing protein (VCP)/p97 (Cdc48 in yeast; Meyer et al., 2012; Franz et al., 2016). Importantly, specific proteases termed deubiquitylating enzymes (DUBs) can cleave off ubiquitin moieties and reverse the signal (Lim et al., 2016).
In this review, we summarize the main ubiquitin-mediated regulatory mechanisms that are believed to fine-tune DNA replication and segregation. We emphasize how E3 ubiquitin ligases orchestrate these processes in space and time, with a special focus highlighting nonproteolytic consequences of ubiquitylation. We aim to pinpoint current research challenges and suggest novel research approaches to decipher the complex ubiquitin-dependent network orchestrating cell cycle regulation.
Of the three described E3 ligase families, HECT, RING, and RING-between-RING E3 ligases (Spratt et al., 2014), the bulk of cell cycle regulation is performed by RING E3 ligases. In particular, the major family Cullin-RING E3 ligases (CRLs) and the anaphase-promoting complex/cyclosome (APC/C) take up most known cell cycle ubiquitylation events. CRLs use one of the six cullin proteins encoded by the human genome as a scaffolding subunit that brings together the ubiquitin-loaded E2 enzyme and the substrate. The E2 enzyme is recruited by the C-terminally bound RING subunit (RBX1 or RBX2). Substrates associate to CRLs via an N-terminal receptor module composed of a variable substrate-specific adaptor and a cullin-bound linker subunit, except in the case of CRL3. CRLs are activated by modification with NEDD8, termed neddylation, and they associate dynamically with regulators that modulate the neddylation state, block substrate access, or promote substrate receptor release and exchange (Lydeard et al., 2013). CRLs are thus regarded as modular, dynamic assemblies with substrate-specific adaptors that associate and dissociate in a regulated manner to ensure timely and specific substrate ubiquitylation (Craney and Rape, 2013). Specific adaptors are linked to individual cullins. CRL1 or SCF (SKP1–CUL1–F-box) E3 ligases contain an F-box protein; CRL3 contains a Broad complex, Tramtrack, and Bric-a-brac (BTB) domain–containing protein; and CRL4 has a DDB1- and CUL4-associated factor (DCAF) protein (Lydeard et al., 2013). Subdivided into CRL2 and CRL5, the Elongin B-C–CUL2/CUL5–SOCS box protein (ECS) E3 ligases recruit BC-box–containing adaptors, in particular VHL-box and SOCS-box proteins (Cai and Yang, 2016). Multiple cell cycle transitions critically depend on SCF E3 ligases, in particular for targeting degradation of cyclin-dependent kinase (CDK) inhibitors such as p27 and WEE1 at G1/S and G2/M, respectively. CRL4 complexes have been described for their functions in preventing DNA re-replication, whereas CRL3 is probably the most emergent CRL in cell cycle control, in particular by regulating mitosis. Several nonproteolytic functions of CRL3 and CRL4 complexes are now attributed (Table 1; Teixeira and Reed, 2013; Bassermann et al., 2014). DNA replication is one of the few cell cycle functions currently attributed to ECS E3 ligases (Table 1 and main text).
Although the APC/C is closely related to CRLs and contains the cullin-homology subunit APC2 (Yu et al., 1998), it is structurally divergent. The APC/C is composed of at least 14 different subunits, including the RING subunit APC11, plus one of two coactivators (CDC20 and CDH1) that also participate in substrate binding (Sivakumar and Gorbsky, 2015). The APC/C operates in mitosis and G1 and is mostly known for its ability to degrade mitotic cyclins and other mitotic factors so that chromosomes are separated and mitotic exit ensues (Zhou et al., 2016).
Dynamic control of DNA replication by ubiquitin
A cell duplicates its genomic information during S phase. Synthesis of the complementary DNA strands begins at localized replication origins, which are established during mitosis and G1 during replication licensing (Fragkos et al., 2015). After DNA duplex unwinding by the replicative helicase, the polymerases (Pol) Polε and Polδ elongate the “leading” and “lagging” DNA strands (Fig. 2 A). Once the duplication of a DNA stretch is complete, replication is terminated and components are removed from chromatin. Ubiquitylation impacts all stages of DNA replication (Moreno and Gambus, 2015; García-Rodríguez et al., 2016). Past research has focused on the global degradation of replication effectors when their function is no longer needed. For example, to prevent rereplication by prematurely assembling origins on newly replicated DNA, replication licensing factors are degraded in S phase, and cells degrade DNA replication factors such as the nuclease FEN1 after replication is complete (Guo et al., 2012; Moreno and Gambus, 2015). Altogether, the prevailing paradigm suggests that degradation of replication effectors is required to restrict their function to a narrow temporal window.
Regulation of lagging-strand synthesis
In recent years, localized proteolytic and several nonproteolytic ubiquitin-mediated regulatory processes have been discovered to regulate replication (Table 1 summarizes nonproteolytic cell cycle ubiquitylation events). An example of replication fine-tuning through selective and localized degradation arises during the process of lagging-strand synthesis (Fig. 2 B). The discontinuous synthesis of DNA requires a constant exchange of factors to prime, elongate, process, and ligate the so-called Okazaki fragments. Priming is performed by Polα, which synthesizes a RNA primer that is removed during the maturation step. Polδ functions during lagging-strand synthesis for consecutive extension of the primer and also for gap-filling during nick translation, a far less processive event (Zheng and Shen, 2011). It appears that in humans the composition of the four-subunit Polδ enzyme (Polδ4) is altered in order to promote this activity shift. Recent evidence argues that the cullin RING ligase (CRL) CRL4CDT2 mediates the destruction of the regulatory p12 subunit of Polδ4 during S phase (Cullin-RING and APC/C E3 ligases text box; Zhang et al., 2013), resulting in the formation of Polδ3, which has specialized properties such as increased proofreading activity. Polδ3 was also associated with gap-filling during DNA repair (Lee et al., 2012). Hence, one model is that the conversion from Polδ4 to Polδ3 generates a polymerase that is more suitable for gap-filling during Okazaki fragment processing (Fig. 2 B), explaining how the processivity of Polδ is locally adjusted (Lin et al., 2013; Lee et al., 2014), with local Polδ4 clearance important for the proper execution of DNA replication. Moreover, there is also a role for nonproteolytic ubiquitylation in lagging-strand synthesis through modulation of protein–protein interactions. MCM10 is a replication fork scaffolding protein involved in the recruitment of the replicative polymerases. Early evidence in yeast suggested that dimonoubiquitylation of MCM10 changes its interactions. Although the affinity of MCM10 for the primase Polα decreases, dimonoubiquitylation likely facilitates the recruitment of the elongating Polε/δ because of the concomitant increased MCM10 affinity to proliferating cell nuclear antigen (PCNA), the sliding clamp that brings these polymerases to DNA (Das-Bradoo et al., 2006; Thu and Bielinsky, 2014). Whether analogous mechanisms also regulate this switch in higher eukaryotes remains to be established.
Control of chromatin assembly during DNA replication
Recent work also uncovered a crucial nonproteolytic role for ubiquitin signaling in regulating the dynamic nucleosomal chromatin structure at advancing replication forks (Fig. 2, C and D). Nucleosome histones must be evicted from DNA and deposited in a semiconservative manner onto new DNA strands and the remaining gaps filled with newly synthesized histones. Thus, nucleosome assembly during S phase necessitates an adequate histone supply (Alabert and Groth, 2012), regulated through transcriptional induction and histone mRNA maturation by the processing factor stem-loop binding protein (SLBP; Fig. 2 D). Interestingly, histone mRNA processing is activated by human CRL4WDR23 through multimonoubiquitylation of SLBP (Brodersen et al., 2016). Indeed, cells lacking WDR23 or SLBP exhibit severe DNA replication defects caused by slow replication forks, suggesting that incorporation of newly synthesized histones is tightly coupled to fork progression. How ubiquitylation mechanistically impacts SLBP function remains to be investigated, but it is conceivable that ubiquitylation regulates its binding to interacting partners or directly affects enzymatic activity (Lampert et al., 2017). After S phase, SLBP is rapidly degraded by SCFcyclin F complexes (Dankert et al., 2016), and this proteolytic destruction is critical for genome maintenance upon genotoxic stress. Thus, nonproteolytic and proteolytic regulation of SLBP by ubiquitin cooperate in space and time to restrict histone synthesis to S phase and thereby maintain genome stability.
Both histone eviction and deposition require so-called histone chaperones. Available data suggest that nonproteolytic ubiquitin signaling mediated by cullin-4 and its putative yeast homologue, Rtt101 (Zaidi et al., 2008), coordinate histone-related processes by acting either on histone chaperones or on histones themselves. Rtt101 is required to target the histone chaperone facilitates chromatin transcription (FACT) complex to the replication fork through nonproteolytic polyubiquitylation of the FACT Spt16 subunit (Fig. 2 C; Han et al., 2010). The same E3 ligase promotes the deposition of newly synthesized histone H3–H4 dimers by ubiquitylating new, acetylated histone H3. The consequence is a switch in interactions between H3–H4 and the respective histone chaperones that allows their loading onto nucleosomes (Fig. 2 D; Han et al., 2013). A recent study clarified that Rtt101 is indeed tethered to replisomes to locally restrict its function to the vicinity of the replication fork (Buser et al., 2016). In humans, CRL4CDT2 is also recruited to active forks (Havens and Walter, 2009; Havens et al., 2012) and may thus perform an equivalent function.
Unloading of the replicative helicase
Rtt101 is not the only resident E3 ligase functioning at yeast replication forks. The replisome also binds the SCFDia2 E3 ligase (Morohashi et al., 2009), further underscoring the importance of local ubiquitylation of factors in the normal progression of replication forks. In the case of SCFDia2, the best described function is to promote the termination of DNA replication (Fig. 2 E). Hence, although Rtt101 is necessary during fork progression, SCFDia2 rather operates when chromosomal replication is finished. Because binding of SCFDia2 to the fork is important, it appears that SCFDia2 in some way senses when replisome function is complete, after which it ubiquitylates the Mcm7 subunit of the replicative helicase (Maculins et al., 2015). Mcm7 ubiquitylation promotes the extraction of the replicative helicase from DNA by Cdc48/p97 and hence the disassembly of the entire replisome, thereby terminating replication (Maric et al., 2014; Moreno et al., 2014). A similar mechanism exists in Xenopus laevis, and the E3 ligase was recently identified to be CRL2LRR1 (Moreno et al., 2014; Dewar et al., 2017). Of note, CRL2LRR1 seems to be specifically recruited to the chromatin at the time of termination instead of being tethered to the replisome like SCFDia2 (Dewar et al., 2017). It is currently unclear whether the K48-polyubiquitylated Mcm7 subunit is targeted to the proteasome or recycled.
Ubiquitin in DNA replication: Open questions
Altogether, ubiquitin can be used to signal specific and consequential modulation of the DNA replication machinery, especially for lagging-strand synthesis factor switching and nucleosome reassembly. Both proteasomal and nonproteolytic pathways can contribute to this behavior. Importantly, the fine-tuned response requires reversible effects, because a modified protein must be rapidly unmodified or replaced to initiate a new synthesis cycle. After local factor degradation, a sufficiently large protein pool must be available to allow dynamic regulation, as in the case of Polδ4 (Lee et al., 2014). Local degradation and replenishment of factors is experimentally challenging to identify, and the process of local cell cycle effector regulation may be more common than current evidence suggests. Likewise, nonproteolytic ubiquitylation is expected to rely on DUBs that remove the modification. However, only a few DUBs have been identified to date that regulate DNA replication, all but ensuring that a new chapter of discovery awaits. For example, is a DUB also tethered to the replication fork to reverse MCM10 dimonoubiquitylation? At which point is newly synthesized histone H3 deubiquitylated? Furthermore, we do not understand how regulatory ubiquitylation signals are translated into their response and which ubiquitin readers are involved. Finally, there are certainly more ubiquitylation substrates and perhaps more E3 ligases with functions in DNA replication awaiting discovery. For example, in Caenorhabditis elegans, the CRL2LRR-1 complex regulates DNA replication, and in Xenopus, the homologous E3 ligase is involved in replication termination (Ossareh-Nazari et al., 2016; Dewar et al., 2017). The direct substrates mediating DNA replication regulation are not known, and whether either function of CRL2LRR1 is conserved in humans remains to be tested. Finally, although the replicative helicase subunit MCM3 is ubiquitylated, the biological significance of this regulation is elusive despite considerable study (Mulvaney et al., 2016). Collectively, many questions remain to be answered, especially in identifying the players that erase and read critical ubiquitin signals during S phase.
Ubiquitin regulation of DNA segregation
Sister chromatids are segregated during mitosis in a process that involves chromosome condensation, nuclear envelope breakdown in animal cells, and centrosome separation to opposite poles. The activity of cyclin-dependent kinase 1, with its positive regulator cyclin B (CDK1/cyclin B), is the main trigger of these events (Gavet and Pines, 2010). In addition, the centromere of condensed chromosomes plays an important role in the assembly of kinetochores that mediate chromosome–spindle attachments and allow chromosome congression at the metaphase plate (Fig. 3 A). Finally, the spindle assembly checkpoint (SAC) monitors microtubule–kinetochore attachments to ensure faithful separation of sister chromatids.
Regulating APC/C E3 ligase activity
Well-timed protein degradation is a common event in the cell cycle, known to drive mitotic entry (G2/M) as well as the metaphase-to-anaphase transition (Teixeira and Reed, 2013; Bassermann et al., 2014). A frequent general question in these and other cell cycle processes is what defines the functional time window of an E3 ligase. In principle, either the activity of the E3 ligase may itself be regulated, or the substrate binding to the E3 ligase may depend on third-party factors such as kinases or scaffolding proteins. Mitosis provides a remarkable example of how an E3 ligase can be dynamically regulated, in this case to tightly coordinate the status of kinetochore–microtubule attachments with the onset of chromosome separation. It is long known that the metaphase-to-anaphase transition is driven by the E3 ligase anaphase-promoting complex/cyclosome (APC/C; see Cullin-RING and APC/C E3 ligases text box), activated by its subunit CDC20 (Teixeira and Reed, 2013; Bassermann et al., 2014). High APC/CCDC20 activity triggers anaphase and mitotic exit by mediating the degradation of cyclin B and securin, an inhibitor of the protease separase that cleaves the cohesin complex holding sister chromatids together (Hirano, 2015). Before anaphase, APC/CCDC20 is kept inhibited by the SAC until appropriate kinetochore–microtubule attachments are established for all chromosomes. A critical product of the SAC is the mitotic checkpoint complex (MCC), which inhibits APC/CCDC20 activity to prevent premature separation of sister chromatids (Lischetti and Nilsson, 2015).
Further studies provided deeper mechanistic insight into the dynamic regulation of the APC/CCDC20 E3 ligase (Fig. 3 B). Surprisingly, the APC/CCDC20 can itself promote the release of its inhibitor MCC through autoubiquitylation of CDC20, a process antagonized by the DUB USP44 (Reddy et al., 2007; Stegmeier et al., 2007). More recently, it was clarified that CDC20 ubiquitylation is brought about by a peculiar structural rearrangement, triggering CDC20 destruction and MCC disassembly (Mansfeld et al., 2011; Varetti et al., 2011; Foster and Morgan, 2012; Yamaguchi et al., 2016). Rather than occurring only at the point of anaphase onset, a model has been proposed in which constant MCC disassembly during metaphase generates a pool of uninhibited APC/C that can either rebind the MCC when unattached kinetochores are present or bind free CDC20 and thus be activated, triggering anaphase onset (Fig. 3 B). This dynamic view of APC/C release from inhibition is complemented by other specific mechanisms of MCC extraction (Westhorpe et al., 2011; Miniowitz-Shemtov et al., 2015; Kaisari et al., 2017). Interestingly, MCC disassembly is enhanced by the ubiquitin reader CUEDC2 (Fig. 3 B; Gao et al., 2011). Although experimental evidence demonstrated that the ubiquitin-binding domain (UBD) of CUEDC2 is important for its function, the ubiquitylated factor to which CUEDC2 binds remains to be determined. The UBD is not required for constitutive binding to CDC20, but we speculate that it might be the key in detecting CDC20 ubiquitylation to trigger MCC release from the APC/C. As a result, CDC20 would be available to the proteasome, with subsequent MCC disassembly.
Ordered degradation of the targets of a single E3 ligase
Another concept currently in focus is the pattern of ordered degradation of substrates of a single E3 ligase. Such pattern was observed for S phase targets of CRL4CDT2 and is established by distinct substrate binding affinities to the E3 ligase (Coleman et al., 2015). APC/CCDC20 likewise represents a prime example of coordinated sequential degradation of E3 ligase substrates, though it does not make use of identical mechanisms. Early observations debated that despite the fact that the MCC precludes the degradation of its late metaphase substrates, MCC-bound APC/CCDC20 can ubiquitylate other targets in prometaphase, namely cyclin A and the kinase NEK2A (Fig. 3 B; den Elzen and Pines, 2001; Geley et al., 2001; Hames et al., 2001). Thus, the very same E3 ligase mediates the destruction of several substrates at different time points. The mechanistic basis for selective substrate targeting includes increased affinity of the early substrates for APC/C binding, and APC/CCDC20 can generate branched ubiquitin chains that are better signals for proteasomal degradation (Meyer and Rape, 2014; Boekhout and Wolthuis, 2015; Di Fiore et al., 2015; Lu et al., 2015a). A summary of the current information on proteolytic ubiquitin signals generated by the APC/C and CRLs can be found in the respective text box.
Unlike the other known classes of E3 ligases, RING E3 ligases work by facilitating the direct transfer of ubiquitin from the E2 to the substrate lysine residue. A different E2 enzyme may be used to initiate and elongate a polyubiquitin chain (Deshaies and Joazeiro, 2009; Ye and Rape, 2009). Alternatively, as for a subset of CRLs, an independent E3 ligase may be recruited to catalyze the initiation step (Scott et al., 2016). The E2 enzyme used for chain elongation is the major determinant of ubiquitin chain topology (Deshaies and Joazeiro, 2009; Ye and Rape, 2009). In the case of CRLs, UBCH5 E2 enzymes allow for mono or multimonoubiquitylation, whereas CDC34 drives chain extension, forming canonical K48-linked polyubiquitin chains (Lydeard et al., 2013; Grice and Nathan, 2016).
Surprisingly, the metazoan APC/C appears to be the major cellular source of atypical K11-linked polyubiquitin chains, which is part of a signal for proteasomal degradation. The APC/C makes use of the E2 enzymes UBE2C and UBE2S to initiate and elongate these atypical chains, respectively (Jin et al., 2008; Garnett et al., 2009; Williamson et al., 2009; Matsumoto et al., 2010; Min et al., 2015; Brown et al., 2016). Despite considerable effort, a consensual structure for K11-linked chains is lacking (Bremm et al., 2010; Matsumoto et al., 2010; Castañeda et al., 2013). Nevertheless, recent studies clarified that homotypic K11 chains are not sufficient to signal proteasome-mediated degradation. Rather, heterotypic K11/K48–polyubiquitinated proteins are efficient proteolytic signals (Grice et al., 2015). Moreover, it was also observed that several ubiquitin chains can be extended from preformed ubiquitin oligomers, constituting branched K48/K11–polyubiquitin chains that appear to be better signals for proteasomal recognition (Meyer and Rape, 2014). These branched chains were suggested to facilitate the degradation of prometaphase APC/C substrates, a mitotic stage characterized by low APC/C activity (Meyer and Rape, 2014). The ability of the human APC/C to synthesize heterotypic ubiquitin chains does not appear to be conserved across all eukaryotes, as at least yeast APC/C substrates are modified with canonical K48-linked polyubiquitin (Rodrigo-Brenni and Morgan, 2007). Yeast might instead make use of complementary mechanisms that reassure the similarly ordered degradation pattern (Lu et al., 2014, 2015a).
Despite the importance of K11 chains as a degradation signal, the respective E2 UBE2S is not essential for cyclin B1 degradation (a canonical APC/C substrate; Garnett et al., 2009; Dimova et al., 2012), leading to the conclusion that multimonoubiquitylation can also constitute a signal for proteasomal degradation (Dimova et al., 2012). Indeed, single-molecule kinetic studies support the view that multimonoubiquitylation can efficiently induce substrate binding to the proteasome (Lu et al., 2015b). Hence, it appears that higher local concentration of ubiquitin moieties enhances binding to proteasomal ubiquitin readers, even though binding to the proteasome does not necessarily correlate with an increased rate of degradation (Lu et al., 2015b; Yau and Rape, 2016). Future research will likely reveal the determinants of the commitment of a substrate to degradation once it is bound to the proteasome.
Fine-tuning kinetochore protein localization
Other E3 ligases operate in mitosis, providing critical regulation often through nonproteolytic ubiquitylation. These signals during mitosis contribute to the remarkable resilience of the system so that cells readily adapt to changing conditions such as the status and quality of kinetochore–microtubule attachments. The importance of these ubiquitylation signals is twofold. Ubiquitylation triggers removal of factors from local chromosomal pools when their function is no longer required, and it can promote microtubule transport of effectors to their new sites of action. During metaphase, correct kinetochore–microtubule attachments must be stabilized, whereas erroneous attachments are destabilized in order to prevent chromosome instability. These processes are coordinated by two kinases, PLK1 and Aurora B (Zitouni et al., 2014; Krenn and Musacchio, 2015). Interestingly, the mitotic localization of Aurora B is regulated by nonproteolytic ubiquitylation at multiple points, including for its microtubule-mediated translocation (Fig. 3, C and E). Aurora B works at the kinetochore to destabilize incorrect microtubule attachments. The VCP–p97 complex ensures exclusive kinetochore localization by removing Aurora B from chromosomal arms, possibly after CRL3KLHL9-KLHL13-mediated polyubiquitylation (Ramadan et al., 2007; Sumara et al., 2007; Dobrynin et al., 2011). In anaphase, Aurora B translocates to the spindle midzone, a process initiated by CRL3KLHL21-dependent monoubiquitylation of Aurora B at attached kinetochores. Remarkably, this ubiquitin signal is decoded by the UBA-containing protein UBASH3B, which recruits ubiquitylated Aurora B to microtubules in the vicinity of the attached kinetochore (Maerki et al., 2009; Krupina et al., 2016). The microtubule-dependent translocation of Aurora B to the spindle midzone in anaphase is mediated by the kinesin MKLP2 (Gruneberg et al., 2004). Indeed, UBASH3B tethers MKLP2 and ubiquitylated Aurora B, thereby promoting microtubule-dependent Aurora B translocation (Fig. 3, C and E). Whether ubiquitylated Aurora B first needs to be extracted by VCP/p97 remains to be investigated.
Although PLK1 stabilizes correct kinetochore–microtubule attachments, its removal from kinetochores is required for faithful metaphase progression (Liu et al., 2012). Bipolar attachment creates tension across the kinetochore, and recent data suggest that this may activate CRL3KLHL22 to trigger rapid removal of PLK1 (Fig. 3 C; Beck et al., 2013). Ubiquitylation is counteracted by the DUB USP16, and thus a balance between CRL3KLHL22 and USP16 ensures the correct localization and function of PLK1 (Zhuo et al., 2015). This balance provides plasticity to this system, as ubiquitylation can be added or removed to fine-tune the localization of a subpopulation of PLK1. Because CRL3KLHL22 regulates PLK1 by nonproteolytic ubiquitylation, probably by monoubiquitylation, its displacement from kinetochores likely depends on a dedicated ubiquitin-binding protein such as VCP/p97. Because PLK1 is also translocated by the MKLP2 kinesin to the spindle midzone in anaphase (Neef et al., 2003), it also needs to be clarified whether ubiquitylated PLK1 is similarly recognized and translocated by UBASH3B or whether this process requires a different reader.
DUBs reveal additional roles of ubiquitin in microtubule transport
Kinase translocation in anaphase is not the only example of how protein ubiquitylation determines cargo for mitotic microtubule-based transport. Two studies reported that the DUBs CYLD and the BRISC complex are involved in the assembly and positioning of the mitotic spindle by regulating the function of the spindle assembly factor NuMA (Yang et al., 2014; Yan et al., 2015). NuMA promotes the tethering of microtubules to the spindle poles and also to the cell cortex and is transported to these sites along microtubules by cytoplasmic dynein (Radulescu and Cleveland, 2010). The BRISC complex appears to deliver ubiquitylated NuMA to spindle poles, indicating that NuMA ubiquitylation likely promotes its transport by dynein (Fig. 3 D; Yan et al., 2015). Cytoplasmic dynein was previously implicated in the transport of ubiquitylated protein aggregates, tethered by the ubiquitin-binding protein HDAC6, and perhaps another reader transports NuMA in a similar fashion (Kawaguchi et al., 2003; Ouyang et al., 2012).
Ubiquitin in mitosis: Open questions
Overall, in mitosis, ubiquitin operates to ensure genome integrity and well-timed DNA segregation by essentially two pathways. First, the peculiar regulation of APC/C by autoubiquitylation provides the necessary flexibility for the cell to quickly recognize changing conditions in the kinetochore–microtubule attachment state. Second, the plasticity of PLK1, Aurora B, and NuMA ubiquitylation ensures that the spindle is correctly assembled and that proper kinetochore–microtubule attachments are established. Today, cell cycle research faces the challenge of understanding how the observed dynamicity in ubiquitylation is achieved. The increased knowledge of APC/CCDC20 regulation might facilitate understanding of how other E3 ligases are regulated in space and time. For example, it seems that CRL3KLHL22 dynamically responds to microtubule–kinetochore tension to ubiquitylate PLK1, but the underlying mechanism remains elusive (Beck et al., 2013). To which extent other cell cycle E3 ligases are regulated in a comparable dynamic fashion will likely demand considerable research efforts. For instance, the APC/C E3 ligase was an early discovery in cell cycle research (Irniger et al., 1995; King et al., 1995; Sudakin et al., 1995), yet APC/C regulation is still an area of active investigation. Dynamic ubiquitylation can also be modulated at the level of the substrate by DUBs, but information regarding how their activity is modulated is mostly lacking. As another pressing and relatively obscure topic, further functional analysis will be required to identify specific readers involved in mitotic processes regulated by nonproteolytic ubiquitin signals. Finally, it will be of interest to determine whether ubiquitin-dependent microtubule motor binding is an ordinary feature in microtubule cargo transport.
Perspective: Reading ubiquitin signals
In this review, we summarized examples in which both proteolytic and nonproteolytic ubiquitin signals regulate cell cycle events. Ubiquitylation of key factors can be reversible, either by a DUB or through the rapid replenishment of a locally degraded factor, such as p12 or CDC20. Despite a growing catalog of nonproteolytic ubiquitin signals, surprisingly little is known about the mechanisms underlying cell cycle regulation that go beyond proteasome targeting. Although monoubiquitylation is widespread (Nakagawa and Nakayama, 2015), assessing nonproteolytic ubiquitin signals and elucidating how ubiquitin mechanistically alters the activity of a given target requires detailed understanding of the underlying process. Therefore, reading the information encoded in ubiquitin chains is now a major challenge in cell cycle research for nondegradative outcomes. The action of CUEDC2 and UBASH3B, in addition to VCP/p97, provides the first clues toward a more comprehensive understanding. We have summarized information regarding cell cycle proteins with UBDs and discovered that ∼25% of the putative human ubiquitin readers are also proteins associated with cell cycle regulation (Table 2). Nevertheless, in the majority of these cases, we do not yet understand the role of the UBD or that of the ubiquitylated binding proteins and subsequent response in the context of the cell cycle. For example, the yeast MCC component BUB3 can bind ubiquitin, but how it contributes to APC/C regulation remains elusive (Pashkova et al., 2010). Other examples are the endosomal sorting complexes required for transport (ESCRT)–related proteins TSG101 and ALIX, which regulate cytokinesis (Morita et al., 2007). Although their interaction with ubiquitin needs to be investigated (Bishop et al., 2002; Dowlatshahi et al., 2012), ALIX and other ESCRT proteins recruit ESCRT-III to promote cytokinetic abscission (Christ et al., 2016). Interestingly, ESCRT-III is directed to the reforming nuclear envelope by a VCP/p97-dependent mechanism to aid in nuclear envelope reformation after chromosome segregation (Olmos et al., 2015). Although speculative, it is thus possible that binding of ALIX to an ubiquitylated factor may similarly help to recruit ESCRT-III during late mitosis. Our efforts to compile cell cycle–associated readers (Table 2) are likely incomplete, and it is therefore clear that much remains to be discovered before the underlying processes of nonproteolytic ubiquitylation are well understood.
Technically, addressing nondegradative ubiquitylation can be a challenging task. In particular, when the bulk levels of a given target protein remain unchanged, it can be difficult to experimentally distinguish local degradation of a small but specific pool from ubiquitin-dependent changes promoting protein translocations and/or activity changes. Tagging specific proteins with a photoswitchable fluorescent protein (Zhou and Lin, 2013) and/or pulse-chase–type labeling with stable protein markers provide powerful tools to visualize ubiquitin-dependent translocations. The identification of specific ubiquitin readers may require siRNA or CRISPR-based screenings and/or mutagenesis of their UBDs. Because of the lack of tools for their detection, another technically challenging task is addressing the synthesis and functions of heterotypic (including branched) polyubiquitin chains in vivo. Ubiquitin linkage in polyubiquitin chains is often distinguished by linkage-specific polyubiquitin antibodies, but they cannot discern between homotypic and heterotypic chains. To overcome this limitation, bispecific bivalent antibodies that simultaneously and exclusively bind two distinct types of ubiquitin linkages within the same polyubiquitin chain have been developed (Rape, M., personal communication). Perhaps research will also lead to the identification of specific ubiquitin readers for these noncanonical linkages that in addition to their functional characterization could be exploited and employed to discriminate linkage types. We believe that new tools will be required to decipher the ubiquitin code. Despite the numerous challenges, it is clear that studying the roles of proteins that noncovalently bind ubiquitin will continue to shed light into how the complex network of ubiquitin-dependent signals cooperate to perpetually drive cells through ordered cycles of DNA synthesis and separation.
We apologize to colleagues whose work could not be included because of space limitations. We are grateful to Michael Rape for sharing results prior to publication, and we thank F. Lampert, R. Dechant, P. Kimmig, and A. Smith for helpful discussions and critical reading of the manuscript.
Work in the Peter laboratory is funded by the Eidgenössische Technische Hochschule Zürich (ETH-20 14-1 and ETH-46 16-1), the Swiss National Science Foundation (SNF 310030B_160321/1), and the European Research Council (ERC 268930 Rubinet).
The authors declare no competing financial interests.
cullin-RING E3 ligase
endosomal sorting complexes required for transport
facilitates chromatin transcription
mitotic checkpoint complex
proliferating cell nuclear antigen
spindle assembly checkpoint
stem-loop binding protein