Physiological functions and roles in cancer of the proliferation marker Ki-67

The Ki-67 protein (also known as proliferation marker protein Ki-67) was discovered in a screen for nuclear antigens present only in proliferating Hodgkin's lymphoma L428 cells (Gerdes et al., 1983). One of the monoclonal antibodies employed, Ki-67, recognised a nuclear antigen that was present only in proliferating cells and absent in quiescent cells, lending its name to the unknown factor (Gerdes et al., 1983, 1984b). Because Ki-67 is expressed in all cell cycle phases and its detection is technically simple [in contrast to other cell proliferation markers, particularly proliferating cell nuclear antigen (PCNA), confusingly first known as ‘cyclin’; Mathews et al., 1984], the Ki-67 labelling index was quickly adopted for the assessment of the proliferating fraction of tumours (Lelle et al., 1987), and it is also of prognostic value in breast cancer (Cuzick et al., 2011).

Nowadays, Ki-67 is the most widely used marker of cell proliferation in cancer, but very few studies have investigated its functions and mechanisms of action. Initial research on Ki-67 was essentially descriptive and focused on its localisation and expression. Ki-67 is present in all vertebrates analysed, whereas searching for homologues by sequence alignment reveals no similar proteins in representative plant, fungi or invertebrate species (including Arabidopsis, fission yeast, budding yeast, Drosophila, Caenorhabditis elegans, starfish). Since early work on the cell cycle generally employed simple genetic organisms, such as yeast or Drosophila, that do not have Ki-67 homologues, and because cell proliferation mechanisms are generally conserved across eukaryotes, Ki-67 was not thought to be an important cell cycle factor and received little attention from the cell cycle community. Another issue is that Ki-67 is a large protein (200–400 kDa in different species), which precludes simple biochemical purification and analysis. It is predicted to be largely intrinsically disordered (Box 1) (Remnant et al., 2021), with the only structured parts found in the N terminus: the forkhead-associated (FHA) and protein phosphatase 1 (PP1)-binding regions (Box 1). This lack of conserved sequences and enzymatic folds gives no clues to the mechanisms of action of Ki-67, nor any tools with which to inhibit its functions. Finally, the available techniques for disrupting gene function in vertebrate cells are relatively recent innovations. Since the evidence for a role in cell proliferation from RNA interference (RNAi) approaches was marginal, and because no mechanisms of action were proposed, interest in Ki-67 declined. Recently, however, a flurry of new studies has revealed possible functions of Ki-67 and roles in carcinogenesis, though many questions remain unanswered. In this Review, we summarise recent findings and speculate on future directions.

Ki-67 localisation changes during the cell cycle (Fig. 1), and its levels are regulated by cell cycle-dependent transcription and protein degradation. In G1 and S phase, Ki-67 localises to discrete nucleoplasmic foci, and once the nucleoli are reformed, it appears to be concentrated at their periphery (Fig. 1), between the strongly DAPI-labelled DNA and other nucleolar components, such as Pescadillo homologue (PES1) or fibrillarin (Bridger et al., 1998; Kill, 1996; Manoir et al., 1991; Sobecki et al., 2016). In postmitotic G1 cells, Ki-67 expression is at its lowest, and its levels rise progressively in subsequent cell cycle phases. In metaphase, when its signal is at maximum intensity, Ki-67 coats the surface of the chromosomes (Fig. 1) (Starborg et al., 1996; Verheijen et al., 1989), after which its levels start to decline (Bruno and Darzynkiewicz, 1992; Gerdes et al., 1984a; Lopez et al., 1991; Sobecki et al., 2017).

At the beginning of G1 phase in human fibroblasts, Ki-67 colocalises with all satellite regions (the majority of centromeres, telomeric repeats and heterochromatic blocks of chromosomes 1 and Y). As cells progress through interphase, a progressive dissociation of specific satellite regions from Ki-67 is observed, and when Ki-67 is present in nucleoli, only centromeres and the short arms of acrocentric chromosomes colocalise with Ki-67 (Bridger et al., 1998). A recent genome-wide mapping of Ki-67 chromosome binding confirmed this dynamic localisation pattern throughout the cell cycle, which also reflects the maturation of nucleoli (Schaik et al., 2021 preprint). The pattern of Ki-67 distribution in mouse meiosis (Traut et al., 2002) is largely similar to that observed in mitotic cell cycles.

Ki-67 mRNA expression is under the control of cell cycle transcription factors and is thus highly correlated with that of cell cycle genes (Sobecki et al., 2017). In the G1 and S phases of the cell cycle, E2F1 and E2F2 promote Ki-67 transcription (Ishida et al., 2001; Ren et al., 2002), whereas B-Myb (also known as Myb-related protein B), a component of the Myb–MuvB (MMB) complex, is responsible for the increase in S phase and G2 (Miller et al., 2018). E2F family- and B-Myb-dependent transcription is regulated by cyclin-dependent kinases (CDKs) that alleviate the repression of E2F and MMB complexes by retinoblastoma (RB) pocket-family proteins (Fischer and Müller, 2017). Analysis of other cell proliferation markers, cyclin A and PCNA, reveals that Ki-67 expression is abolished upon treatment with inhibitors of CDK4 and CDK6 only in cells whose proliferation is arrested (Sobecki et al., 2017). Thus, Ki-67 staining is a good readout for the effects of CDK4 and CDK6 inhibitors, which are now employed in breast cancer therapy, on cell proliferation (O'Leary et al., 2016).

The degradation of Ki-67 protein starts in anaphase and telophase (Schlüter et al., 1993), continues in G0 and G1, and depends on CDH1 (Fizzy-related protein homologue)-mediated activity of the anaphase-promoting complex/cyclosome (APC/C) (Miller et al., 2018; Sobecki et al., 2017). Human Ki-67 isoforms have two or three KEN-box consensus motifs for recognition by the APC/C–CDH1 complex (Sobecki et al., 2016). Nevertheless, quiescent cells that have recently exited the cell cycle continue to express low levels of Ki-67 (Bullwinkel et al., 2006; Miller et al., 2018; Sobecki et al., 2017), whose degradation carries on in arrested cells. Thus, Ki-67 is not a ‘black and white’ marker of cell proliferation. More generally, the changing levels of Ki-67 throughout the cell cycle appear to account for all the variability in Ki-67 expression observed in non-transformed and cancer cells (Sobecki et al., 2017), which has implications for the interpretation of clinical data.

Several of the unstructured domains (Box 1) of Ki-67 confer localisation properties. The conserved domain (CD, amino acids 619–765; Sobecki et al., 2016) is responsible for Ki-67 localisation to the nucleolus (Saiwaki et al., 2005), whereas the leucine–arginine-rich (LR) domain in the C terminus targets Ki-67 to heterochromatin in interphase and to the periphery of the chromosomes in mitosis (Saiwaki et al., 2005). This domain directly binds to DNA, an interaction that is decreased upon phosphorylation (MacCallum and Hall, 2000).

Indeed, Ki-67 undergoes a large number of posttranslational modifications (see PhosphoSitePlus, https://www.phosphosite.org; Hornbeck et al., 2015), which include phosphorylation, ubiquitylation, acetylation and others. Such modifications alter the net charge (Ki-67 is highly basic, pI=9.5) and thus may modify electrostatic interactions of the disordered regions. Phosphorylation of Ki-67 regulates its localisation and is at least in part mediated by CDKs. Ki-67 contains ∼80 confirmed and putative CDK phosphorylation sites (Box 1; Altelaar et al., 2022 preprint); it can be phosphorylated in vitro by CDK1–cyclin B and protein kinase C, and dephosphorylated by PP1 and PP2A protein phosphatases (Endl and Gerdes, 2000; MacCallum and Hall, 1999). In cells, Ki-67 is phosphorylated from the beginning of mitosis by CDK1 (Blethrow et al., 2008; Takagi et al., 2014); abolishing its mitotic phosphorylation prevents its localisation to mitotic chromosomes, and instead, Ki-67 is found in cytoplasmic foci that colocalise with the nucleolar component nucleolin (Hégarat et al., 2020; MacCallum and Hall, 1999).

The localisation of Ki-67 in mitosis and the beginning of G1 has also been reported to depend on p150 (also known as chromatin assembly factor 1 subunit A; Matheson and Kaufman, 2017), a subunit of the chromatin assembly factor 1 (CAF-1) that has been found to interact with many nucleolar proteins and regulate their localisation to nucleoli (Smith et al., 2014). When p150 is depleted, mitotic Ki-67 association to the perichromosomal layer (PCL) is strongly reduced, and staining for Ki-67 in G1 becomes diffuse in the nucleoplasm (Matheson and Kaufman, 2017). However, altered localisation of other nucleolar proteins, such as nucleolin, suggests that depletion of p150 might generally affect nucleolar integrity. Indeed, disruption of nucleolar structure by any method causes dispersion of Ki-67 in the nucleoplasm (Sobecki et al., 2016). Since Ki-67 is required to localise nucleolar components to the PCL of mitotic chromosomes (Booth et al., 2014; Sobecki et al., 2016), these results suggest that p150 might link Ki-67, nucleolar components and the PCL.

A large portion of Ki-67 is comprised of a repeat domain (amino acids 1000–2936; Box 1). Each repeat encompasses a conserved ‘Ki-67 motif’ of 22 amino acids (Duchrow et al., 1996; Schlüter et al., 1993), which contains CDK and mitogen-activated protein kinase (MAPK) consensus phosphorylation sites (Takagi et al., 2014). The functions of this domain and its phosphorylation remain unknown.

That Ki-67 transcription, degradation and posttranslational modifications are tightly regulated by cell cycle factors explains why Ki-67 expression and localisation change during the cell cycle. However, this does not necessarily mean that Ki-67 controls cell proliferation; indeed, it is clearly not required for proliferation of most eukaryotic cells, because it is only present in vertebrates.

Multiple early studies downregulated and depleted Ki-67 using antisense, RNAi or light-induced inactivation in different cell lines (see Box 2 and Table 1 for more details). A number of these studies reported increases in cell death or reduced cell growth rates and inferred that Ki-67 is required for cell proliferation. Yet none of these publications demonstrated unequivocally that Ki-67 is essential, nor identified any mechanism by which Ki-67 might act on cell proliferation.

Recent genetic approaches in many different cell lines have now confirmed that Ki-67 is not required for cell proliferation (see Table 1). Adeno-associated virus (AAV)-mediated gene targeting in MCF-10A and DLD-1 cell lines has been used to introduce a premature nonsense mutation into exon 2 of both alleles of the gene encoding Ki-67 (Cidado et al., 2016), but this has no effect on cell growth rates. TALEN-mediated disruption of the gene encoding Ki-67 in NIH-3T3 fibroblasts does not alter cell proliferation, although mutant cells lacking Ki-67 are marginally more sensitive to serum starvation (Sobecki et al., 2016). Conversely, cell cycle arrest is not affected in mouse embryonic fibroblasts (MEFs) when Ki-67 degradation is inhibited through genetic disruption of Fzr1, which encodes Cdh1 (Sobecki et al., 2016), showing that cell proliferation can be experimentally uncoupled from Ki-67 in either direction. Targeting the gene encoding Ki-67 using CRISPR-Cas9 in HeLa cells, human triple-negative breast cancer MDA-MB-231 cells or mouse 4T1 cells does not affect cell proliferation, levels of DNA damage or cell death (Cuylen et al., 2016; Mrouj et al., 2021). Furthermore, Ki-67 has been conditionally depleted using an auxin-inducible degron system in HCT116 cells (Takagi et al., 2016, 2018), and no effects on cell proliferation were reported. The most compelling evidence for the lack of requirement for Ki-67 in cell proliferation and survival comes from analysis of the Cancer Dependency Map (DepMap; https://depmap.org/portal/) data, which have compiled gene effects from over 1000 genome-wide CRISPR-Cas9 screens. Ki-67 knockout only has effects on viability and/or proliferation for ∼1% of cell lines, compared to ∼99% for genes that encode established cell proliferation regulators, such as c-Myc, PCNA or cyclin A2 (Mrouj et al., 2021).

Finally, TALEN-mediated disruption of Ki-67 in mice has been found to not affect animal growth, development or adult homeostasis (Sobecki et al., 2016). However, low levels of Ki-67 expression remain in these animals, presumably due to leaky translation from a downstream ATG codon, which if the resulting truncated protein is functional, might suffice to fulfil Ki-67 functions (Sobecki et al., 2016). Furthermore, an independent mouse Ki-67 gene disruption resulting from a non-functional Ki-67 Cre knock-in also supports the conclusion that mice lacking Ki-67 are viable (O. Basak and H. Clevers, personal communication); no residual protein or mRNA could be detected in these mice, and they did not have any apparent developmental defects. However, a definitive answer on whether Ki-67 is essential or not for mouse development will require a genetic approach resulting in a complete deletion of the Ki-67 open reading frame (which is underway in our team).

Given that Ki-67 expression is not required for cell proliferation or, probably, for tissue development, the question arises whether Ki-67 is required for tumorigenesis. Interestingly, despite its dispensability for cell proliferation, Ki-67 is expressed at similar levels in all cancer cell lines analysed and is not subject to copy number variation, suggesting that its over- or under-expression confers a disadvantage to cancer cells (Mrouj et al., 2021; Sobecki et al., 2017). The former might be because its overexpression induces ectopic heterochromatin and arrests the cell cycle (Sobecki et al., 2016), while the latter implies that Ki-67 plays a role in carcinogenesis. This is supported by cancer patient data, in which the MKI67 gene encoding Ki-67 is rarely mutated and almost never amplified or deleted (Mrouj et al., 2021). Indeed, NIH-3T3 Ki-67-knockout cells show significantly lower ability than wild-type cells to form colonies after transduction with H-Ras oncogene (Mrouj et al., 2021). Furthermore, in mice lacking Ki-67, intestinal tumorigenesis induced chemically or genetically is largely abrogated (Mrouj et al., 2021). These results indicate a requirement for Ki-67 in chemically or oncogenically driven cell transformation.

Tumour progression from established cancer cells also appears to depend on Ki-67. In early studies, the delivery of antisense oligonucleotides against Ki-67 in vivo was found to inhibit the establishment of tumours but not growth from existing tumours. This was reported for both syngeneic models of murine cancers (Kausch et al., 2003) and for xenografts from human renal carcinoma (Kausch et al., 2004, 2005). Interpretation of these studies requires caution because the downregulation of Ki-67 levels in tumours was not verified. In a subsequent study, despite only partial depletion of Ki-67 using shRNA, xenografts of human 786-0 renal cancer cell line showed a significant inhibition in growth (Zheng et al., 2009). A more recent study has found that tumours of DLD-1 colon cancer cells lacking Ki-67 engrafted at intermediate density subcutaneously into mice grow more slowly than their control counterparts, although, confusingly, this is not observed at low or high cell density (Cidado et al., 2016). In the 4T1 mouse mammary carcinoma model, Ki-67 knockout has been found to strongly reduce tumour growth in two immunocompromised mouse models (athymic nude and NOD SCID), despite unaltered cell proliferation, apoptosis and DNA damage in vivo (Mrouj et al., 2021). Similarly, human breast adenocarcinoma MDA-MB-231 Ki-67-knockout xenografts in nude mice grow slower than controls, and apoptosis and fibrosis are increased, whereas necrosis is reduced (Mrouj et al., 2021). Tumour growth is also decreased in xenografts originating from HeLa S3 cells with stably knocked down Ki-67, and again, cell proliferation is not affected in vivo, but necrosis and apoptosis are increased (Mrouj et al., 2021). Last but not least, Ki-67 knockout almost completely eliminates metastasis in the 4T1 model (Mrouj et al., 2021).

Experiments with 4T1 cells in immune-competent mice have revealed surprising interactions between Ki-67 expression and anti-tumour immune responses, which may constitute an Achilles' heel for cancer cells. Unexpectedly, although metastasis is still abrogated by Ki-67 knockout in this system, the initial anti-tumour immune response is impaired, and this coincides with increased influx of myeloid-derived suppressor cells (MDSCs) that promote an immune-suppressive environment (Mrouj et al., 2021). This appears to be in part due to reduced expression of major histocompatibility complex I and genes involved in antigen presentation (Mrouj et al., 2021). As such, Ki-67 expression renders cancer cells more ‘visible’ to the immune system. These findings are important, since they may have implications for cancer treatment. If Ki-67 generally promotes anti-tumour immune responses, then, paradoxically, it might be advantageous to promote, rather than hinder, cell proliferation in order for immunotherapy to be optimally effective. However, this will require confirmation in other syngeneic models.

Cell plasticity may more generally underlie the requirement for Ki-67 in different steps of tumorigenesis. Cancer cells must adapt to a hostile environment, in which they evade contact inhibition and grow under hypoxic conditions, and upon metastasis, they must thrive in a heterologous tissue, which requires the ability to rapidly adapt transcriptional programmes. Cancer cells with the most highly stem-like characteristics belong to the hybrid intermediate epithelial–mesenchymal transition (EMT) states (Kröger et al., 2019; Pastushenko et al., 2018; Ye et al., 2015). Ki-67 knockout in 4T1 and MDA-MB-231 cells results in a shift towards a more epithelial or more mesenchymal phenotype, depending on the cell type, and loss of characteristics of stem cells, such as aldehyde dehydrogenase activity and spheroid-seeding ability (Cidado et al., 2016; Mrouj et al., 2021). This is also reflected in the reduced drug metabolism in 4T1 cells lacking Ki-67. Transcriptome analysis of Ki-67-knockout 4T1 cells has revealed downregulation of various xenobiotic metabolism enzymes, resulting in a significant increase in sensitivity to all drugs analysed (Mrouj et al., 2021), which has also been reported for other Ki-67-knockout cell lines (Cidado et al., 2016; Cuylen et al., 2016).

The mechanisms underlying this apparent loss of cell plasticity remain unknown, but given the genome-scale alterations in gene transcription and the altered chromatin states in cells lacking Ki-67 (Mrouj et al., 2021; Sobecki et al., 2016), it seems likely that they result from epigenetic programmes (Fig. 2 and see below).

As described above, Ki-67 is a very large and almost entirely intrinsically disordered protein (IDP). IDPs participate in a wide range of cellular processes (Van Der Lee et al., 2014) and often show extensive but moderate-affinity interactions with other proteins, nucleic acids and small molecules. Some IDPs form ‘hubs’ that can interact with many partners, where the disorder provides the required plasticity for dynamic interactions with different proteins (Mittag et al., 2010) and can be regulated by posttranslational modifications that alter electrostatic interactions. Ki-67 may well be one such ‘hub’ protein, given its plethora of binding partners (Box 3; Fig. 2; Table S1) and its extensive phosphorylation in mitosis (MacCallum and Hall, 1999; Altelaar et al., 2022 preprint).

The two structured regions of Ki-67 provide binding surfaces for specific interactions. The FHA domain of Ki-67 has been shown to interact with two partners, human kinesin-like protein-2 (HKLP2, also known as kinesin-like protein KIF15) (Sueishi et al., 2000) and MKI67 FHA domain-interacting nucleolar phosphoprotein (NIFK) (Takagi et al., 2001), both in mitosis and in a phospho-dependent manner. The phosphorylation of NIFK that is responsible for its tight interaction with Ki-67 occurs in mitosis (Byeon et al., 2005), when Ki-67 relocalises to the surface of chromosomes (Starborg et al., 1996; Verheijen et al., 1989). As such, this interaction might promote the partitioning of nucleolar material to daughter cells, which seems to be a key function of Ki-67 (see below). In vitro, Ki-67 interacts with the γ isoform of PP1 (PP1γ), and in vivo it also interacts with PP1β but not PP1α. This interaction mediates the appropriate localisation of the phosphatase to anaphase chromosomes (Booth et al., 2014; Takagi et al., 2014). The interaction with PP1 is controlled by Aurora B (also known as Aurora kinase B) phosphorylation (on Ser507), which prevents premature targeting of the phosphatase to mitotic chromosomes (Kumar et al., 2016).

The disordered part of Ki-67 has many partners. A variety of interactions of Ki-67 were uncovered in a yeast two-hybrid screen and confirmed in human cells (Schmidt et al., 2003). The Ki-67-interacting proteins identified could be classified into three families: cell cycle factors [MCM2, PP2A catalytic subunit, PAK2, S6 kinase, RanBP7 (also known as importin 7), RanBP9 and p95 (also known as nibrin)], nucleolar components (various ribosomal proteins, fibrillarin and SMN1) and proteins involved in regulating chromatin structure [clusterin, reptin 52 (also known as Ruv-B-like 1), eIF4A, DDX1, RBBP7, HP1α (also known as chromobox protein homologue 5) and HP1γ (also known as chromobox protein homologue 3)]. In mitotically synchronised NIG272 cells, Ki-67 has been found to interact with topoisomerase (Topo) IIα (also known as TOP2A) and the hCAP-H2 subunit of condensin II complex, which was not observed in asynchronous cells (Takagi et al., 2016).

The LR domain of Ki-67 binds to all three isoforms of heterochromatin protein 1 (HP1), with the highest affinity for the α isoform (Scholzen et al., 2002). This interaction has been confirmed both in vitro and in vivo (Kametaka et al., 2002; Scholzen et al., 2002). HP1 and Ki-67 colocalise within constitutive heterochromatin. Interestingly, overexpression of HP1 causes a redistribution of Ki-67 away from its nucleolar localisation (Scholzen et al., 2002). Conversely, overexpression of Ki-67 also attracts HP1 to ectopically formed heterochromatin (Sobecki et al., 2016).

More recently, affinity purification followed by mass spectrometry in asynchronous U2OS cells has revealed 406 specific Ki-67 interactors (Sobecki et al., 2016), some of which had already been well described, including CDK1, PP1, and NIFK. These interactors could be grouped into several functional classes: proteins involved in rRNA biogenesis, translation and ribosome biogenesis, splicing, transcription, and chromatin (Box 3; Fig. 2; Table S1). The two last classes encompass around 70 factors, suggesting that Ki-67 participates extensively in the control of chromatin and gene expression (Sobecki et al., 2016).

After cell division, Ki-67 leaves the chromosomal periphery and relocates to the reforming nucleoli (Fig. 1), suggesting that it might have a role in rRNA synthesis (Bullwinkel et al., 2006), which potentially explains its presumed role in cell proliferation. Indeed, rRNA synthesis was found to be inhibited upon irradiation of cells incubated with an anti-Ki-67 TuBB-9 antibody conjugated to a photoactivable agent (Rahmanzadeh et al., 2007). However, as this effect was not observed with another anti-Ki-67 antibody (MIB-1), it is likely that nucleolar integrity was non-specifically disrupted. A later study in HeLa cells concluded that Ki-67 has a role in efficient reactivation of ribosomal genes after mitosis (Booth et al., 2014), as cells depleted of Ki-67 were found to have reduced 47S and 26S pre-rRNAs, while the locus 13p, a marker for the nucleolar organiser region (NOR) of chromosome 13, was observed to relocate from the interior of the nucleolus to more external regions. However, altered pre-rRNA synthesis and processing were not observed upon Ki-67 knockdown in HeLa, U2OS or HCT116 cells, which also did not change nucleolar structure (Sobecki et al., 2016). The reasons for these divergent results are not clear, but what is clear is that Ki-67 is not required for nucleolar functions in non-vertebrates, and, as vertebrate cells can proliferate without Ki-67, it cannot be required for rRNA biogenesis.

Nevertheless, Ki-67 is required for the relocalisation of nucleolar components, including nucleolin, PES1 and NIFK, to the perichromosomal region upon disassembly of nuclei at the beginning of mitosis (Fig. 1), which most likely serves to symmetrically distribute them into daughter cells (Booth et al., 2014; Sobecki et al., 2016). Ki-67 is itself recruited relatively early to the perichromosomal compartment and thus might act as a scaffold for other proteins.

The chromosome periphery, or PCL, is composed of various proteins and RNA, mostly derived from nucleoli, that associate with condensed chromosomes in metaphase and account for more than a third of the mitotic chromosome volume (Booth and Earnshaw, 2017; Booth et al., 2016). It has been suggested that the function of the chromosomal periphery is to organise or structure the chromosomes, prevent them from sticking to one another, and/or provide a platform for transferring nucleolar components to the emerging daughter cells, and in all of these functions, Ki-67 seems to play a central role. Ki-67 accounts for 1.6% of the chromosome protein mass (Booth et al., 2016) and appears to have a scaffolding role in the assembly of the PCL, as in cells lacking Ki-67, all components tested [PES1, nucleolin, NIFK, PP1, cPERP-B (coiled-coil domain-containing protein 137), cPERP-C (Pumilio homologue 3), cPERP-D (ATP-dependent helicase DDX18), cPERP-F (probable ATP-dependent RNA helicase DDX27)] fail to localise to the chromosomal periphery, whereas loss of other PCL proteins does not result in such a phenotype (Booth et al., 2014; Sobecki et al., 2016). This corresponds to the disappearance of the ‘fuzzy’ transition between the electron-dense chromosome and the cytoplasm in cells lacking Ki-67 that has been observed by electron microscopy (Booth et al., 2014). There exists a small Ki-67-independent portion of the chromosomal periphery (Booth et al., 2016).

In the absence of Ki-67, chromosomes appear to be ‘swollen’, with a tendency to clump together, while the chromatin itself is slightly less compacted (Booth et al., 2014, 2016; Sobecki et al., 2016; Takagi et al., 2016), which might result from the loss of association of TopoIIα and the hCAP-H2 subunit of condensin II complex (Takagi et al., 2016). When Ki-67 is acutely depleted specifically in mitotic cells (using the auxin-inducible degron system), chromosomes are observed to lose their rod-like shapes, and their structure becomes distorted (Takagi et al., 2016). These effects have not, however, been seen in other studies (Booth et al., 2014, 2016; Sobecki et al., 2016), arguing against a role of Ki-67 or the PCL in the maintenance of intrinsic chromosome structure.

To explain the clumping of mitotic chromosomes observed in the absence of Ki-67, a role for Ki-67 as a biological surfactant has been proposed (Cuylen et al., 2016). In an siRNA screen in HeLa cells, Ki-67 was identified as the only factor whose absence resulted in mitotic chromosomes clustering together. Loss of spatial separation was accompanied by decreased mobility of chromosomes and a marginally delayed progression from nuclear envelope breakdown to anaphase. This phenotype was dependent on the DNA-binding C-terminal LR domain. The high electrical charge and amphiphilic structure of Ki-67 was proposed to explain its surfactant function, with the LR-rich C-terminus binding DNA, and the long remainder of the protein excluded from the chromatin. To confirm the hypothesis, histones, as positively charged chromosome-binding proteins, were overexpressed in Ki-67-lacking cells, partially rescuing the phenotype (Cuylen et al., 2016). These conclusions are difficult to reconcile. First, the positive net charge of Ki-67 is most likely largely neutralised by its extensive phosphorylation by CDK1 in mitosis. Second, overexpressed histones appear to be incorporated into chromatin, and do not form the PCL, which is dependent on Ki-67. It is thus difficult to appreciate how overexpressed histones, which cannot form additional nucleosomes and thus only be exchanged into existing ones, could create a surfactant layer needed to repel individual chromosomes. Furthermore, an excess of core histones is toxic to cells (Singh et al., 2010), and their levels are transcriptionally and translationally regulated, and are dependent on DNA replication (Heintz, 1991; Zhao et al., 2000). Instead, perhaps the protein- and RNA-rich chromosomal periphery created by Ki-67 serves to spatially separate mitotic chromosomes. However, the significance of such a role remains unclear given that cells lacking Ki-67 proliferate normally and show no obvious mitotic defects (Booth et al., 2014, 2016; Sobecki et al., 2016).

Another more recent proposition is that, in addition to its requirements for organising the PCL, Ki-67 mediates the exclusion of cytoplasmic components upon nuclear reformation after mitosis (Cuylen-Haering et al., 2020), at least in HeLa cells. This is a surprising claim given that nuclear envelope reassembly occurs directly on postmitotic chromatin and nuclear components are subsequently concentrated by active nuclear transport (Güttinger et al., 2009); thus, no active mechanisms to exclude cytoplasm should be required. Besides, Ki-67 is not present in non-vertebrate cells. Thus, a possible explanation for these findings might be that the absence of the PCL in Ki-67-depleted cells hinders formation of the nuclear envelope. However, this would not be consistent with the apparent observation that overexpressing histones (with the caveats mentioned above) partially rescues the chromosomal clumping phenotype but not cytoplasmic exclusion upon Ki-67 depletion. Another possibility is that defects in nuclear reformation after mitosis in cells lacking Ki-67 might be due to the significantly decreased concentration of PP1 phosphatase at chromatin in anaphase (Booth et al., 2014), when the process of nuclear envelope reassembly begins. PP1 is required at the exit of mitosis to dephosphorylate, and dissociate from chromatin, nucleoporins, nuclear envelope proteins and chromatin-associated factors (Güttinger et al., 2009). Therefore, at present, these recent findings (Cuylen-Haering et al., 2020) remain conceptually and experimentally enigmatic and will require corroboration.

Perhaps the most striking phenotype observed in cells lacking Ki-67 is the global reorganisation of their transcriptional programmes. shRNA or siRNA targeting of Ki-67 in HeLa, U2OS and hTERT-RPE1 cells leads to deregulation of expression of hundreds of genes (Sobecki et al., 2016; Sun et al., 2017). Genetic targeting of MKI67 in NIH-3T3 fibroblasts, mouse mammary carcinoma 4T1 cells or MDA-MB-231 cells results in global but cell type-specific changes in gene expression (Mrouj et al., 2021). For the most downregulated genes in 4T1 Ki-67-knockout cells, a strong correlation with an increase in the repressive histone mark histone H3 lysine 27 trimethylation (H3K27me3) is observed. This histone mark is mediated by the PRC2 complex, whose essential component SUZ12 interacts with Ki-67 (Sobecki et al., 2016). However, the deregulation of PRC2 only explains a fraction of the changes observed, and other gene expression alterations correlate with an increase in either repressive histone H3 lysine 9 trimethylation (H3K9me3) marks or activating histone H3 lysine 4 trimethylation (H3K4me3) marks surrounding the transcription start site (Mrouj et al., 2021). Furthermore, additional knockout of the gene encoding SUZ12 in 4T1 cells lacking Ki-67 does not overcome their reduced ability to form tumours in nude mice (Mrouj et al., 2021).

This raises the question of how such extensive gene expression changes can be achieved by a single protein with no known function in regulation of transcription (Fig. 1). A potential clue is provided by another observation made in cells lacking Ki-67, in that perinucleolar DAPI staining appears less intense, suggesting changes in heterochromatin compaction, which have been confirmed by fluorescence resonance energy transfer (FRET) experiments that assay inter-nucleosome proximity (Sobecki et al., 2016). Downregulation of Ki-67 also leads to loss of long-range interactions between pericentromeric and perinucleolar heterochromatin. These phenotypes are associated with a redistribution of the repressive histone marks H3K9me3 and histone H4 lysine 20 trimethylation (H4K20me3), which are more dispersed within nuclei rather than colocalising with DAPI-dense regions (Sobecki et al., 2016). Interestingly, HP1 localisation is not affected, and thus is uncoupled from that of H3K9me3. Similarly, major satellites are still retained within DAPI-dense regions but no longer show prominent H3K9me3 staining (Sobecki et al., 2016). Conversely, overexpression in human cells of Xenopus Ki-67, human full-length Ki-67 or the LR DNA-binding domain causes formation of ectopic heterochromatin that is marked by HP1 proteins (Kametaka et al., 2002; Sobecki et al., 2016). Thus, Ki-67 organises heterochromatin, with potential knock-on effects on gene expression. A recent study provides further insight; Ki-67 and lamin B1 bind to similar chromatin regions, but the two proteins show anti-correlation, and all late-replicating domains are covered by either Ki-67 or lamin B1 (Schaik et al., 2021 preprint). Upon Ki-67 depletion using the auxin-inducible degron system, lamin B1 partially relocates to regions previously occupied by Ki-67, while replication timing of pericentromeric regions is marginally delayed. Thus, Ki-67 appears to prevent certain genomic regions from lamin B1 binding and formation of lamina-associated domains (LADs), which are known to be associated with gene repression and late replication timing (Gonzalez-Sandoval and Gasser, 2016). Ki-67 loss might therefore be expected to induce ectopic LADs and alter gene expression. Surprisingly, however, acute depletion of Ki-67 in this system has few immediate effects on gene expression (Garwain et al., 2021; Schaik et al., 2021 preprint). Taken together, these considerations suggest that chromatin modifications and gene expression alterations might be a longer-term adaptation of cells to a loss of Ki-67.

The question remains, what is the molecular mechanism by which Ki-67 organises interphase chromatin? One possibility is that this depends on the physico-chemical nature of Ki-67 and its binding partners, many of which are chromatin and transcription regulators (Box 3; Fig. 2; Table S1). Transient multivalent interactions between IDPs or intrinsically disordered protein regions, RNA and DNA molecules are responsible for liquid–liquid phase separation (LLPS). LLPS is becoming increasingly recognised for its roles in various cellular processes, as it allows organisation of three-dimensional compartments with high concentrations of functionally related molecules, such as nucleoli, Cajal bodies, nuclear speckles, stress bodies, promyelocytic leukaemia (PML) bodies, nuclear pores and others (Banani et al., 2017; Shin and Brangwynne, 2017). Recently, heterochromatin has also been proposed to behave as a phase-separated compartment (Sanulli et al., 2019; Strom et al., 2017). HP1, a prominent Ki-67 interactor, has been suggested to drive the nucleation process of heterochromatin phase separation. HP1 binds to the histone H3 marks H3K9me2 and H3K9me3 that are deposited by suppressor of variegation 3-9 homologue 1 (SUV39H1), and both proteins are considered essential for formation and maintenance of heterochromatin (Cheutin et al., 2003). Although initially thought a compact and rigid entity, the heterochromatin compartment is highly dynamic and accessible to regulatory factors, as demonstrated by the transient binding of HP1 and its dynamic exchange with the nucleoplasm (Cheutin et al., 2003), a characteristic of liquid-phase condensates (Shin and Brangwynne, 2017). HP1 itself can phase separate in vitro (Strom et al., 2017), but this capacity is limited in cells (Erdel et al., 2020). However, HP1 interacts with MeCP2, a methyl-CpG-binding protein, which can phase separate in cells (Li et al., 2020), and both proteins interact with Ki-67 (Sobecki et al., 2016). Some controversy exists in the field about whether heterochromatin is liquid-like. According to a more recent study, it behaves rather as collapsed polymer globules, entirely independent of HP1, where interactions are stronger and mobility of molecules is reduced (Erdel et al., 2020), whereas another paper demonstrates that heterochromatin is solid-like, providing a scaffold for liquid-like assemblies of proteins (Strickfaden et al., 2020). Nevertheless, irrespective of whether heterochromatin is liquid and HP1 the driving force behind its phase separation, Ki-67 might participate in the formation of this compartment given its size, intrinsically disordered nature and its role as a hub for multi-protein interactions.

Importantly, a phase-separation model of transcriptional control has been proposed, which would explain the formation, characteristics and function of superenhancers (Hnisz et al., 2017). Coactivators that are enriched within superenhancers (BRD4 and MED1) have been demonstrated to phase separate (Sabari et al., 2018). RNA polymerase II transcription factories have also been shown to form phase-separated condensates, mediated by the activation domains of transcription factors interacting with the Mediator complex (Boija et al., 2018). In this context, the intrinsically disordered nature of activation domains would explain how hundreds of divergent transcription factors can interact with a very limited number of coactivators, such as the Mediator complex or p300 (histone acetyltransferase p300) (Allen and Taatjes, 2015).

It will be important to test experimentally whether Ki-67 phase separates, both in vitro and in vivo, and to investigate what role its various regions, in particular the repeats, play in this process. The repeats can be extensively phosphorylated, and this modification provides a potential means for a dynamic regulation of phase separation propensity. In-depth structural and biochemical studies will provide further mechanistic insights into the role of Ki-67. It will also be essential to study the contribution of Ki-67 to the liquid properties of heterochromatin and transcription compartments. Based on the available data, we hypothesise that Ki-67 participates in the formation of the liquid-like heterochromatin-binding protein compartment, thus controlling accessibility of chromatin regulators, potentially by itself undergoing LLPS (Fig. 1).

Why is it important to understand Ki-67 function? Although for many years it was assumed that Ki-67 is required for tumorigenesis as a result of its presumed roles in cell proliferation, the notion that it promotes cell proliferation has been comprehensively refuted. Nevertheless, recent results have shown that Ki-67 is required for tumorigenesis, probably because it allows adaptation of transcriptional programmes that are needed for cancer cells to thrive in a hostile environment – in other words, cellular plasticity. At the same time, the functions of Ki-67 are linked to cell proliferation by virtue of the regulation of its expression by canonical cell cycle regulators, for many years a ‘red herring’ that led to incorrect inferences about its roles. The reasons for this linkage are currently unclear, but one may speculate that over evolutionary timescales, mechanisms allowing genome compartmentation and plasticity provided important advantages for proliferating cells that eventually became hard-wired. An illustrative example is the prominent heterochromatin of cancer cells (Zink et al., 2004).

Whether Ki-67 is involved in pathologies other than cancer remains to be seen but appears likely. One of its binding partners, the methyl-CpG-binding protein MECP2, localises to heterochromatin, and its disruption is causal in the severe genetic neurological disorder Rett syndrome. The likelihood that vertebrate development does not require Ki-67 opens the previously unrecognised possibility that loss of Ki-67 may trigger subtle phenotypes that are not immediately apparent in mutated mice.

Our understanding of Ki-67 functions has expanded considerably, with solid demonstrations that it is not required for cell proliferation but instead involved in subcellular organisation, especially of the mitotic PCL and interphase heterochromatin. These functions may well be related and rely on the same biochemical properties of Ki-67, namely, its intrinsic disorder and charge distribution, which undoubtedly underlie its extensive protein–protein interactions. Whether these interactions serve the purpose of sequestering factors or increasing their local concentration in the heterochromatin compartment (or both) remains to be determined. Given the recent explosion of interest in IDPs and liquid-like phases in subcellular compartmentalisation, the localisation of Ki-67 to perinucleolar heterochromatin and the PCL makes it tempting to suggest that Ki-67 might more generally serve as a scaffold for formation of phase-separated compartments. One might speculate that the PCL, like nucleoli, is either liquid-like or more solid, similar to heterochromatin, but currently, the answer to this question remains open and will require further studies to test this hypothesis.