The fate of metaphase kinetochores is weighed in the balance of SUMOylation during S phase

Introduction: The SUMO Pathway

Conjugation of Small Ubiquitin-like Modifier proteins (SUMOs) to cellular target proteins regulates many processes, including nucleo-cytoplasmic transport, mitosis, gene expression and DNA-metabolism.¹ SUMOs undergo post-translational proteolytic processing, and mature SUMOs become attached to lysine residues within their target proteins through a conserved enzymatic cascade requiring the sequential action of an activating or E1 enzyme (Aos1/Uba2), a conjugating or E2 enzyme (Ubc9) and usually one of several ligases or E3 enzymes.² SUMOylation is highly dynamic and is readily reversed by a single family of SUMO proteases.^3,4 The same group of proteases both generates mature SUMOs by proteolytic processing (peptide bond cleavage) and removes SUMO from the conjugated substrates (isopeptide bond cleavage). Budding yeast has two SUMO proteases, Ulp1 and 2,⁵ while mammals have six genes encoding SUMO proteases, SENP1, 2, 3, 5, 6 and 7.⁴

Fungi have a single SUMO (Smt3p in budding yeast). Vertebrates possess three commonly expressed SUMOs: SUMO-2 and 3 are ∼95% identical to each other, and each shows ∼45% identity with SUMO-1.² Due to their high sequence identity, SUMO-2 and 3 remain indistinguishable in most contexts, and can be collectively called SUMO-2/3. Most SUMOylation involves SUMO-2/3, since they are present at significantly higher concentrations than SUMO-1,⁶ and are more dynamic.⁷ Notably, SUMO-2/3 can also form conjugated SUMO chains.⁸ SUMO-1 does not typically form chains in vivo,⁸ and SUMO-1 conjugation is relatively more stable.⁹ Interestingly, SUMO-1 is dispensable for mouse development.¹⁰ All mammalian SENPs act on SUMO-2/3-conjugated species, while only SENP-1 and 2 act on SUMO-1-conjugated species.¹¹ Chains of Smt3p or SUMO-2/3 are disassembled by a sub-family of SUMO proteases, which includes Ulp2p in yeast and SENP6 and SENP7 in vertebrates.^12–16 These enzymes are collectively called SUMO chain editors.

SUMOylation can cause a variety of outcomes for any particular target protein, including changes in the target's sub-cellular localization, activity, protein-protein interactions and stability. Mechanistically, these changes often reflect the loss or acquisition of interaction surfaces upon conjugation of the SUMO group.¹ Many cellular proteins contain SUMO-interacting motifs (SIMs) that allow low-affinity interactions with free or conjugated SUMOs.^17–20 Some SIMs bind particular SUMO polypeptides selectively,^19,21 allowing paralog-specific interactions. SIM-SUMO interactions frequently promote assembly of higher-order protein complexes.^22,23

Recently, SIM-SUMO-based interactions were shown to be crucial for proteasomal degradation of some poly-SUMOylated targets through the action of SIM-containing ubiquitin ligases, dubbed as SUMO targeted ubiquitin ligases (STUbLs) (Fig. 1). The most studied STUbLs are Slx5/Slx8 in yeast and RNF4 in vertebrates.^24–30 The biological roles of this degradation pathway are not yet fully understood; notably, SUMOylated species are often formed as intermediates in the assembly of large protein complexes,^22,23,31 so it is possible STUbLs contribute to quality control mechanisms for disposal of mis-assembled or stalled intermediates.

Deletion of budding yeast Ulp2p disrupts mitotic progression and chromosome segregation.^5,32 We tested whether depletion of the vertebrate SUMO chain editor SENP6 from HeLa cells could provoke analogous defects. We found that SENP6 depletion disrupted chromosome congression and caused prolonged delays in mitosis.³³ These findings can be understood in the context of a novel role for SUMOylation in governing inner kinetochore assembly. Here, we present an overview of how SUMOylation controls kinetochore function within mitosis, as well as our findings regarding its distinct role in the regulation of inner kinetochore assembly.

SUMOylation during Mitosis: Outer Kinetochore and Inner Centromeric Proteins

Centromeres are chromatin domains on each chromosome that possess a specialized histone H3 variant, CEN-H3 (CENP-A in vertebrates and Cse4 in yeast).³⁴ Kinetochores are large proteinaceous structures, which are assembled onto centromeres and which capture microtubules (MTs) to form the kinetochore fibers (k-fibers) that bi-orient the sister chromatids and allow their anaphase segregation to daughter cells.³⁵ Kinetochores consist of around a hundred different proteins, arranged in multiple sub-complexes.³⁶ Kinetochores have three other important roles: (1) Kinetochores nucleate MTs during early prometaphase that enhance polar MT capture;³⁷ (2) Kinetochores generate a “wait anaphase” signal in the absence of proper MT attachment (bi-orientation).³⁸ The pathway through which this signal is generated is called the Spindle Assembly Checkpoint (SAC); (3) Kinetochores generate the mechanochemical force required for anaphase separation of sister chromatids.³⁹

When visualized by electron microscopy, vertebrate kinetochores show two electron dense regions (EDRs) separated by a translucent gap.⁴⁰ The EDR proximal to the chromosome is known as the inner kinetochore, and the distal EDR is known as the outer kinetochore. Many of the enzymatic functions of kinetochores probably reside within the outer kinetochore, whereas the inner kinetochore may play a more structural role. Consistent with this idea, inner kinetochore components are epigenetically maintained in association with centromeres throughout the cell cycle while the outer kinetochore is reassembled prior to each mitosis.^41,42 The fibrous corona is a region associated with the external face of the outer kinetochore that contains components of the SAC pathway and motor proteins. The inner centromeric region (ICR) is the domain of the centromeres that lies between the inner kinetochore plates.

In vertebrates, mitotic SUMOylation has been demonstrated for ICR, outer kinetochore and fibrous corona proteins (Table 1). Experiments in mitotic Xenopus egg extracts (XEEs) using GFP-SUMO-2/3 demonstrate considerable accumulation of SUMO-2/3 in the ICR region.⁴³ ICR components that are confirmed SUMOylation targets include Topoisomerase II,⁴⁴ poly(ADP-ribose) polymerase 1 (PARP1)⁴⁵ and the chromosome passenger complex component Borealin.⁴⁶ SUMOylation appears to be important for remodeling of attachments of Topoisomerase II to chromatin⁴³ and for negative regulation of PARP1 activity against chromosomal substrates.⁴⁵ The biological role of Borealin SUMOylation is not clear.⁴⁶

Outer kinetochore and fibrous corona proteins that are confirmed SUMOylation targets include BubR1 (a SAC component), Nuf2 (a member of the Hec1 MT binding complex) and CENP-E (a plus end-directed MT motor) (Table 1). In addition, CENP-E is a poly-SUMO-2/3 binding protein.⁴⁷ Suppression of mitotic SUMOylation in HeLa cells by overexpression of SUMO protease SENP2 disrupts CENP-E targeting to kinetochores.⁴⁷ Notably, mutation of SIMs within the C-terminal tail of CENP-E also blocks its recruitment to kinetochores, suggesting that its capacity to recognize poly-SUMOylated species is particularly important for its localization at this site. Nuf2 and BubR1 are implicated in targeting of CENP-E to kinetochores,^48,49 so it is attractive to speculate that their SUMOylation may directly promote CENP-E recruitment through its SIM domains.⁴⁷

Finally, a complex containing the nucleoporin RanBP2, SUMO-1-conjugated RanGAP1 and Ubc9 is recruited to outer kinetochores after initial attachments have formed between centrosome-nucleated MTs and kinetochores.⁵⁰ RanGAP1 is the GTPase activating protein for Ran, a Ras-family GTPase with critical roles in interphase nuclear trafficking and in mitotic spindle assembly.^51,52 Recruitment of this complex to kinetochores appears to be important for suppressing MT nucleation at kinetochores⁵³ and for forming correct bi-oriented k-fibers.^54,55 Interestingly, RanBP2 also possesses a domain with SUMO E3 activity, which has been implicated as a ligase for ICR substrates.⁴⁶

SUMOylation of all known targets within the ICR, outer kinetochore and fibrous corona occurs within mitosis. While some of these proteins can be modified in other contexts,⁵⁶ it is not clear whether SUMOylation at other points of the cell cycle can control their mitotic functions. While we have not performed an exhaustive search, it is notable that none of the ICR, outer kinetochore or fibrous corona proteins that we have tested to date appear to be substrates of SENP6, since they showed neither substantial increases in their SUMOylated populations nor were they destabilized after SENP6 depletion. It thus appears likely that SUMOylated targets within these structures are deconjugated by other proteases, and that SUMOylation may act in distinct ways during different phases of the cell cycle to control kinetochore function.

SUMOylation of CENP-H/I/K during S Phase

The inner kinetochore components that directly interact with CENP-A nucleosomes have been classified as the nucleosome-associated complex (NAC), and those that interact with CENP-A indirectly are known as CENP-A nucleosome-associated distal (CAD) complex;⁵⁷ together, they are known as constitutive centromere-associated network (CCAN)⁴² (Supp. table 1 and Fig. 2). While CCAN components are essential for chromosome congression, spindle assembly and sister kinetochore bi-orientation, the biochemical functions of most individual CCAN proteins remain undefined. However, it is clear that they can play distinct roles in kinetochore assembly, and that multiple CCAN sub-complexes may be subject to SUMOylation: CENP-C is essential for recruitment of the outer kinetochore proteins.⁵⁸ CENP-C interacts with the SUMO pathway genetically,⁵⁹ and it can become SUMO conjugated in vitro,⁶⁰ but the in vivo circumstances and molecular function of CENP-C SUMOylation remain obscure. The CENP-H/I/K complex is important for CENP-A loading in G1 phase,^61,62 as well as for inner kinetochore assembly through recruitment of the CENP-O/P/Q/U/R⁶³ and CENP-S/X⁶⁴ complexes (Fig. 2). CENP-H/I/K has also been implicated in the control of MT plus end dynamics and flux at the kinetochore.⁶⁵ As discussed below, we found that CENP-H/I/K components are important SUMO-2/3 conjugation substrates that are modified during S phase.

We examined HeLa cells depleted of SENP6 by siRNA mediated knockdown³³ to determine whether they would show mitotic defects, analogous to those observed in budding yeast lacking Ulp2p.^5,32 While many aspects of spindle assembly appeared grossly normal, cells lacking SENP6 consistently showed an increase in pole-to-pole distance of around 15%, as well as a broader distribution of chromosome within the metaphase plate and increased distance between sister kinetochores.³³ Moreover, although the bulk of chromosomes in cells lacking SENP6 congressed onto the metaphase plate, a few chromosomes typically remained unaligned and close to the poles. The unaligned kinetochores activated the SAC, prolonging mitosis for several hours, but the cells did not achieve complete chromosome congression despite this delay. The depleted cells frequently slipped out of mitosis with unaligned chromosomes, resulting in mis-segregation. We found only modest increase in the overall level of SUMOylated species in SENP6-depleted cells.¹³

To understand this phenotype, we systematically surveyed the abundance of representative members of all major kinetochore complexes. Most remarkably, while CENP-A and CENP-C were correctly localized, the CENP-H/I/K complex was absent from the kinetochore.³³ We measured the levels of CENP-H and CENP-I and found the abundance of both the proteins was dramatically reduced after SENP6 depletion,³³ suggesting that their absence was the result of destabilization. We hypothesized that one or more component of the CENP-H/I/K complex became polySUMOylated in SENP6-depleted cells, resulting in its STUbL-mediated degradation. One prediction of this hypothesis is that loss of both SENP6 and RNF4 would result in the accumulation of polySUMOylated forms of the targeted protein(s). Consistent with this prediction, RNF4 depletion restored the levels of CENP-H and -I in SENP6-depleted cells, and both proteins accumulated in highly SUMOylated species.³³ These findings show that CENP-H/I/K complex members are bona fide SUMO-2/3 conjugation substrates in vivo.

Consequences of CENP-H/I/K Degradation after SENP6 Depletion

In surveying kinetochore components, we did not observe the wholesale degradation of other proteins in the manner that we had found for CENP-H and -I,³³ indicating that the CENP-H/I/K may be unique among the kinetochore complexes as a target of RNF4-mediated degradation in the absence of SENP6. Previous reports have established that the CENP-H/I/K complex depends on the CENP-T/W complex and CENP-L, M and N for its localization at the kinetochore.⁵⁷ In turn, the CENP-O/P/Q/U/R⁶³ and CENP-S/X⁶⁴ complexes require CENP-H/I/K for their localization. Loss of the CENP-H/I/K complex also results in reduced levels of some outer kinetochore complexes, including the Hec1 and Mis12 complexes.⁶⁶ The CENP-O/P/Q/U/R complex recruits polo like kinase 1 (Plk1), which is responsible for phosphorylating many kinetochore proteins,^63,67–70 suggesting that Plk1 localization should indirectly depend upon the CENP-H/I/K complex.

The pattern in which other kinetochore components were re-localized after SENP6 depletion was consistent with the previously described hierarchy of kinetochore protein recruitment. Although we found no degradation of CENP-O/P/Q/U/R complex members, CENP-O was absent from kinetochores, as would be expected in the absence of CENP-H/I/K. In turn, mis-localization of CENP-O/P/Q/U/R should disrupt Plk1 recruitment.⁶⁸ We tested this prediction by examining the localization of Plk1 in cells depleted of SENP6, CENP-I and CENP-O (Fig. 3). We found that depletion of each of these proteins caused the loss of Plk1 from kinetochores, but not from centrosomes. Cells lacking either SENP6 or CENP-I frequently showed large cytoplasmic Plk1 aggregates in SENP-6 and CENP-I depleted cells [Fig. 3(ii) and (iii)] which are not present in CENP-O depleted cells [Fig. 3(iv)]. We often observed cytosolic CENP-O foci after SENP-6 knockdown with a similar appearance (data not shown), so it is possible that these aggregates reflect association of the mis-localized populations of CENP-O/P/Q/U/R complex and Plk1. Consistent with the correct recruitment of Plk1 to centrosomes, we did not find major defects in centrosomal maturation or MT nucleation (data not shown). Finally, the predicted absence of CENP-S/X complex from kinetochores after degradation of CENP-H/I/K may underlie the increase in intrakinetochore distance in metaphase SENP-6 depleted cells.⁶⁴

Quantitative immunofluorescence showed that the Hec1 complex was partially mis-localized in the absence of SENP6,³³ consistent with its pattern of mis-localization after CENP-H/I/K complex depletion.⁶⁶ Interestingly, we observed differential loading of Hec1 on the attached and unattached kinetochores within individual cells: Hec1 levels were reduced around 45% on the aligned kinetochores, whereas they were reduced around 75% on the unaligned kinetochores. This differential pattern was surprising because earlier reports had not found that Hec1 recruitment is sensitive to the attachment status of individual kinetochores.^71,72 It is interesting to speculate that loss of CENP-H/I/K may lead to inconsistent loading of the Hec1 complex onto kinetochores. In this case, our observations would imply that kinetochores require a threshold level of Hec1 for efficient MT capture, and that kinetochores that do not meet this threshold may remain unaligned indefinitely. Mis12 is a member of another key outer kinetochore complex whose localization has been reported depend partially upon CENP-H/I/K.⁶⁶ Mis12 showed decreased localization on kinetochores, particularly those of unattached chromosomes, although the extent of this decrease was somewhat less than we observed for Hec1.³³ Notably, outer kinetochore components whose recruitment has been reported to be independent of CENP-H/I/K (CENP-E and -F⁶⁶) did not show significant changes after SENP6 depletion.

Together, these patterns are consistent with the idea that CENP-H/I/K complex members are the key targets of SENP6 within the kinetochores, since the defects that result from SENP6 depletion closely mimic those caused by loss of CENP-H/I/K.

Implications for Kinetochore Assembly

Our observations do not clarify why the CENP-H/I/K complex is regulated through this elaborate mechanism that balances the activities of SENP6 with RNF4. Nor, indeed, do they reveal why CENP-H/I/K complex members are SUMOylated in the first place. It was striking that the maximum accumulation of these polySUMOylated species occurred during S phase in cells that were synchronized using a double thymidine block/release (DTB) protocol, rather than in mitosis.³³ S phase is an interval of dynamic recruitment of CENP-H/I/K complex to constitutive inner kinetochore structures,⁷³ and it is possible that maximal levels of their SUMOylation corresponds to the interval of inner kinetochore assembly. It is thus attractive to speculate that CENP-H and -I SUMOylation promotes assembly of inner kinetochores, in a manner that may be analogous to the role of SUMOylation in the assembly of other multi-protein complexes.^22,23,31 In this context, it is notable that co-depletion of RNF4 substantially rescues chromosome congression defects observed in the absence of SENP6 alone. CENP-H and -I accumulate in highly SUMOylated forms under these circumstances,³³ indicating that complete de-SUMOylation of CENP-H and -I cannot be essential for assembly of sufficiently functional inner kinetochores.

There are several possible ways in which CENP-H and -I SUMOylation could promote inner kinetochore assembly. First, SUMOylation may enhance formation of the CENP-H/I/K complex. As shown in Supp. Table 1, CENP-H contains both a SUMOylation consensus motif and SIM, whereas CENP-I and CENP-L have SIMs. Alternatively, SUMO moieties attached to the CENP-H/I/K complex could provide an interaction surface for CENP-H/I/K complex loading onto inner kinetochores or for recruitment of other SIM-containing kinetochore proteins (Suppl. Table 1). If de-SUMOylation were only achieved after inner kinetochore assembly were complete, persistent SUMOylation of assembly intermediates may allow their interactions with RNF4 and their degradation, thus disposing of stalled or superfluous assembly intermediates.

It is clearly possible that such regulated degradation of assembly intermediates could be a mechanism to prevent the formation of ectopic centromeres. Ectopic centromeres are extremely deleterious for genomic stability as they cause chromosome fragmentation during anaphase.^74,75 CENP-A loading in extra-centromeric sites is very inefficient but not unprecedented,^76–78 suggesting that its deposition is normally tightly controlled. The CENP-H/I/K complex enhances the loading of CENP-A containing nucleosomes;⁶¹ if unrestricted, such an activity could imaginably bypass normal restraints for CENP-A deposition at extra-centromeric sites. This model would predict tight temporal or spatial regulation of SENP6, RNF4 or both enzymes to prevent inappropriate deposition of centromeric components.

Implications for the SUMO Pathway

STUbLs mediate the targeted destruction of poly-SUMOylated substrates, while chain editing SUMO proteases mediate the deconjugation of the same species. Our studies show that the direct antagonism of these classes of enzymes can form a biologically significant network, regulating the abundance of their targets. It appears likely that this paradigm will also be applicable to other proteins whose degradation is controlled by RNF4, such as PML^79,80 and Hypoxia-Inducible Factor-2α.⁸¹

SUMO-2/3 modification can be reversed by six SUMO proteases, which are distributed to different regions of the nucleus.⁴ Given that nuclear subdomains are extremely dynamic,⁸² any of these SENPs might in principle be able to deconjugate SUMO-2/3 chains from CENP-H/I, whereas it appears that only SENP6 actually fulfills this function in reality.³³ The explanation for this exquisite specificity might lie in the enzymatic properties and the protein levels of individual SENPs. SENP6 shows a higher affinity for poly-SUMO chains than SENP2, and it is more active in cleavage of polySUMOylated substrates.¹⁴ An elevated affinity of SENP6 towards SUMO chains can be accomplished through its multiple SUMO interacting motifs, and RNF4 has been proposed to act preferentially on polySUMOylated substrates for the same reason.²⁹ This model predicts that other SENPs cleave SUMOylated species with one or two SUMO moieties. When the chain length exceeds this level, SUMOylated species become high affinity targets of RNF4 and SENP6, which catalyze their destruction or reduce their level of SUMOylation below the threshold required for high affinity recognition. Since SENP7 also preferentially acts on SUMO-2/3 chains,^14,15 it would be expected to function in a similar manner. However, SENP7 is much less abundant than SENP6 in HeLa cells (unpublished observation), which may explain its inability to compensate SENP6 deficiency in our experiments.

Obviously, it is also possible that SENP6 or RNF4 may have specificity for individual polySUMOylated target proteins. We have not observed concentration of SENP6 near centromeres¹³ or observed detectable amount of SENP6 in CENP-H/I immunoprecipitations (unpublished observation), so we currently have no evidence that such a mechanism is applicable in the case of the CENP-H/I/K complex.

Conclusions and Future Perspectives

Our findings provide the first case in which inner kinetochore assembly is regulated through post-translational modification. The use of SUMOylation in this context may be uniquely valuable, as it provides both a mechanism to facilitate complex assembly and a mechanism through which inappropriate structures can be destroyed. It will now be important to discover how this modification is used and what binding interfaces may be made or disrupted through SUMOylation. It will also be important to discover the E3 enzyme(s) that controls CENP-H/I/K complex modification, and to understand how this enzyme and SENP6 are temporally and spatially regulated during the process of kinetochore assembly. It appears likely to us that similar regulatory networks can be used in a variety of cellular contexts, and it will be fascinating to discover how they are adapted within different situations.

Supplementary Material