From Zero to Many: Control of Centriole Number in Development and Disease

Authors

  • Inês Cunha-Ferreira,

    1. Cell Cycle Regulation Lab, Instituto Gulbenkian de Ciência, Rua da Quinta Grande, 6P-2780-156 Oeiras, Portugal
    Search for more papers by this author
  • Inês Bento,

    1. Cell Cycle Regulation Lab, Instituto Gulbenkian de Ciência, Rua da Quinta Grande, 6P-2780-156 Oeiras, Portugal
    Search for more papers by this author
  • Mónica Bettencourt-Dias

    Corresponding author
    1. Cell Cycle Regulation Lab, Instituto Gulbenkian de Ciência, Rua da Quinta Grande, 6P-2780-156 Oeiras, Portugal
      Mónica Bettencourt-Dias, mdias@igc.gulbenkian.pt
    Search for more papers by this author

Mónica Bettencourt-Dias, mdias@igc.gulbenkian.pt

Abstract

Centrioles are essential for the formation of microtubule-derived structures, including cilia, flagella and centrosomes. These structures are involved in a variety of functions, from cell motility to division. In most dividing animal cells, centriole formation is coupled to the chromosome cycle. However, this is not the case in certain specialized divisions, such as meiosis, and in some differentiating cells. For example, oocytes loose their centrioles upon differentiation, whereas multiciliated epithelial cells make several of those structures after they exit the cell cycle. Aberrations of centriole number are seen in many cancer cells. Recent studies began to shed light on the molecular control of centriole number, its variations in development, and how centriole number changes in human disease. Here we review the recent developments in this field.

Centriole Structure and Function

The centriole is a eukaryotic organelle involved in a variety of processes. When tethered to the cytoplasmic membrane, the centriole sets up the foundations for the axoneme, the skeleton of cilia and flagella. As an axoneme organizer, the centriole is termed the basal body (1). Cilia and flagella are indispensable in a variety of cellular and developmental processes including cell motility, propagation of morphogenic signals and sensory reception (2–4). Within the cytoplasm, the centriole is part of the centrosome, the major microtubule (MT) organizer in animal cells, which coordinates cell division, motility and polarity (2). Several cancers display cells with supernumerary centrosomes (5–10), which commonly cluster in two poles upon cell division (11–13). Perhaps this process is selected to avoid cell death induced by aberrant cell division (11–13). Although many cells can divide in the absence or with an excess of centrosomes, others depend on a correct number of those structures for accurate chromosome segregation, asymmetric cell division and survival (12,14–18), suggesting the presence of tissue-specific constraints in the dependency of centrosomes for correct cell division (14).

The canonical centriole/basal body is a very stable structure, ∼0.5-μm long and 0.2μm in diameter, made of nine MT triplets composed mainly of heterodimers of α and ß tubulin. Centrioles/basal bodies are polarized along the proximo-distal axis (19–21). Electron microscopy has revealed that at their proximal end, immature centrioles, and in some species even mature ones, have a feature called the cartwheel, a structure made of a central hub linked by spokes to the inner tubule of each triplet (22–24) (Figure 1A). Older centrioles have sub-distal and distal appendages, which dock cytoplasmic MTs and anchor centrioles to the cell membrane (19,20) (Figure 1A). Basal bodies have additional appendages including rootlets at their base and a basal foot formed along precise basal body MT triplets (21,25). Those structures organize basal body position and orientation in relation to other cellular components, which are important for coordinating the movement of cilia and organization of the intricate cytoskeleton in many protists (21,25). Basal bodies also have a transition zone at their distal end, contiguous with the axoneme, which is important for the nucleation of ciliary MTs, as a docking site for protein transport into the cilia, and for organized disassembly/assembly of the cilia upon entry into mitosis (26).

Figure 1.

The centriole duplication cycle is coupled with the chromosome replication and segregation cycle. A) Centrosome/centriole structure. Structure of the centrosome with two centrioles. The mother centriole is the one showing appendages. Inset shows the detail of cartwheel, which exists in human immature centrioles. B) Centriole duplication cycle (upper yellow panel) and chromosome replication and segregation cycle (lower blue panel) (see main text for references). A1) Procentriole formation begins at the G1/S transition upon recruitment of SAS-6, which is needed to form the cartwheel, a structure that helps in defining the centriole nine-fold symmetry. CDK2 activity may be necessary for speeding up procentriole formation and elongation, hence coordinating this event with DNA replication. B1) In late G2, the daughter centriole reaches full elongation and maturation with the recruitment of several molecules to the PCM. CDK1 activity increases in G2 regulating a variety of molecules and processes needed for entry into mitosis, such as changes in MT dynamics. C1) Through the concerted action of molecules such as the kinase Nek2, the two centrosomes separate. D1) When a cell exits mitosis, the centrioles within the centrosome disengage through the action of Separase. That process may allow recruitment or activation of molecules necessary for duplication. The substrates of Separase in this process are still unknown. One molecule that may play a role is SAS-6, as its levels are highest at this stage, being degraded later on in mitotic exit/G1 by the APC/Cdh1. CDK1 activity prevents any reduplication of the centrosome and rereplication of the DNA. SAK/PLK4 is essential and can trigger procentriole formation. Although the regulation of SAK/PLK4 is not fully described, we show it in the figure as mitotic activity, as its levels are highest at this cell cycle stage. Upon proper alignment of the sister chromatids, the spindle assembly checkpoint is satisfied, Cyclin B is degraded and Separase is activated. Sister chromatid separation and centriole disengagement are triggered. The mitotic spindle segregates the chromosomes equally to the two daughter cells.

The centrosome is composed of two centrioles which recruit and organize an electron-dense and protein-rich matrix, the pericentriolar material (PCM; Figure 1A). Even centrioles that are not fully formed, which occur in animals with mutations in the centriole duplication machinery (27–29), can recruit PCM, but centriole loss leads in general to PCM dispersal (15,30–32). The PCM anchors and nucleates cytoplasmic MTs in interphase and mitosis. Principal among PCM components are members of the pericentrin and AKAP450 family of proteins, and other components, such as centrosomin (CNN)/CDK5RAP2 and SPD2/CEP192, coiled-coil molecules that dock regulatory components and molecules mediating the nucleation of MTs, such as γ-tubulin (2,20,33,34).

Variations on a Theme: Centriole Number in Different Cell Types

Morphological events and timing in the canonical cycle

The number of centrioles in a cell is normally controlled through a canonical duplication cycle in coordination with the chromosome cycle (Figure 1B). ‘One and only one’ new centriole forms orthogonally to each existing one in a conservative fashion (35). Four consecutive steps in the centrosome cycle have been defined through electron microscopy (36–40): disengagement of the centrioles, nucleation of the daughter centrioles (also called procentrioles before acquiring full length), elongation of the procentrioles and separation of the centrosomes. Nucleation of daughter centrioles is coordinated with DNA synthesis, leading to the formation of two distinct centrosomes. Thus, when the cell enters mitosis, it is equipped with two centrosomes, which participate in mitotic spindle assembly (Figure 1B). Each centrosome in a G2 cell has a different age: one has a mother and a daughter, the other has a grandmother and a daughter centriole. These differences provide variation in the competence for MT nucleation and anchoring, which are important for example in stem cell asymetric divisions and may be important for cell polarity (16,41–43).

The exact timing of events in the canonical cycle may depend on tissue- and species-specific constraints. For example, in Drosophila syncytial embryos, where there are no cell cycle gap phases, centrioles separate in telophase after disengagement, before pro-centriole formation becomes visible in interphase. Effectively, each nucleus in the embryo is left with two centrosomes, each with one centriole, right after mitosis (44). In sea urchin embryos and Chlamydomonas, procentriole formation starts during mitosis (26,45,46). It is possible that procentriole formation also starts at the end of mitosis in mammals, but that is not visible by conventional electron microscopy techniques. Studies on the cell cycle regulation of the essential activities in centriole duplication should clarify this issue.

Diversity in centriole number

Loss of the ability to nucleate MTs and/or disappearance of centrioles are common features associated with cell specialization. Notably, oocytes and spermatozoa lose parts of the centrosome in a complementary manner. Centrioles, but not centrosome proteins, are lost in the oocyte. On the other hand, in sperm, the PCM material is lost, but not the centriole/basal body, which is necessary for flagella formation. Fertilization brings together the basal body from the sperm, which becomes a centriole within the zygote, and the PCM from the egg. In Drosophila oogenesis, centrioles disappear before female meiosis. Here, a cyst of 16 connected cells is generated, one of which is the presumptive oocyte, the other ones being nurse cells. All the centrioles in the 16 cells migrate to the presumptive oocyte, where complex MT rearrangements take place. Centrioles have been seen as late as stage 9, in an Microtubule Organizing Center (MTOC) associated with the nucleus (47), but seem to have disappeared by meiosis (Figure 2A) (48–50). In other species, centrioles disappear later. In oocytes of starfish Asterina pectinifera, there is an extrusion of the first polar body containing two of the four centrioles after meiosis I. After meiosis II, there is an extrusion of a second polar body containing one centriole. The oocyte is left with one centriole which is not able to duplicate or nucleate MTs, and the sperm brings in the active basal body/centriole upon fertilization (51) (Figure 2B).

Figure 2.

Centriole biogenesis and loss in different cell types. See main text for references. A) Centrioles disappear during oogenesis in many animal species, including Drosophila. Drosophila oogenesis starts with a stem cell that divides asymmetrically. The germline cyst cell that is born divides four times giving rise to a cyst of 16 connected cells. One of the two cells that has more ring canals will undergo meiosis and will become the presumptive oocyte. The other 15 cells undergo DNA endoreplication cycles and become the nurse cells. All the centrioles in the nurse cells migrate to the presumptive oocyte, where complex MT rearrangements take place. Centrioles have been seen as late as stage 8/9, in an MTOC associated with the nucleus (47), but disappear soon after (48–50). Female meiosis is acentriolar. B) Centrioles disappear during oogenesis in starfish A. pectinifera. In Metaphase I the spindle is oriented perpendicularly to the oocyte membrane. The centriole outer pair is extruded into polar body 1 (PB1) and the inner centriole pair is inherited by the secondary oocyte. In meiosis II, the pair of centrioles within the secondary oocyte splits into two single centrioles and forms a meiosis II spindle. The mature mother centriole is then extruded into PB2, and the nonreproductive daughter is left in the mature egg. The remainder centriole decays shortly after completion of meiosis. The four centrioles are already non-equivalent in their nucleation and reproductive capacity before meiosis (see Ref. (52) for more detail). C) Inactivation of supernumerary centrosomes. Cancer cells often show centrosome amplification, clustering at the poles of the mitotic spindle. Other centrosomes may show less nucleation activity and do not cluster with the remainder. D) Formation of multiple centrioles during ciliogenesis of epithelial cells. Hundreds of centrioles are formed either around a pre-existing mother centriole (1) or a deuterosome (2). These centrioles migrate and dock to the cell membrane, as basal bodies, where they nucleate cilia. E) De novo centriole formation during oogenesis of parthenogenic insect species. After egg activation, multiple centrioles arise de novo and join the female pronucleus resulting from meiosis. In the absence of fertilization, those MTOCS set up the first mitotic spindle in the unfertilized egg. The remaining MTOCs disappear. F) Centriole duplication is uncoupled from the DNA replication cycle during human spermatogenesis. Primary spermatocytes duplicate the two pairs of centrioles in the transition from Meiosis I to Meiosis II, with no DNA replication. Secondary spermatocytes give rise to two spermatids, each with two centrioles. During sperm differentiation the mother centriole matures into a basal body and nucleates the flagella while the daughter centriole is retained near the nucleus within the sperm head and may degenerate.

Centriole inactivation and/or loss associated with the formation of non-radial, non-centrosomal MT arrays is commonly seen upon differentiation (53). For example, in the differentiation of skeletal muscle, upon fusion of myoblasts to give rise to the syncytial myotubes, the centrosome disassembles and most of the MT nucleating capacity is taken by the nuclear envelope. Proteins such as γ-tubulin, pericentrin and ninein localize to the nuclear envelope, and longitudinal MT bundles are formed along the long axis of the cell (54–56). Centrosome differential inactivation/activation is also seen in the context of asymmetric stem cell division (57) and when an excess of centrosomes is present in a dividing cell (Figure 2C). In the latter, some centrioles can cluster at the spindle poles, whereas others are inactivated and neither participate in, nor disturb, normal mitotic spindle formation (11–13).

An amazing system of centriole biogenesis is seen in mammalian epithelial multiciliated cells, where 200–300 basal bodies are formed in each cell after differentiation (1,58–60) (Figure 2D). This arises in the developing vertebrate respiratory and reproductive tracts and in the ventricular system of the brain. Multiple centrioles form around a mother centriole, breaking the control of ‘only one’ centriole forming per mother. Centrioles can also form around less characterized, non–MT-based dense structures of heterogeneous size, called deuterosomes (1). It is not clear why there are two morphologic pathways. Perhaps the roles of centrioles and deuterosomes are the same, recruiting and concentrating the components needed for centriole biogenesis. Further studies of the composition and function of those structures are needed.

Many centrioles can also be formed in the absence of pre-existing centriolar structures, called de novo biogenesis. Particularly impressive is centriole biogenesis in insect species which can develop with no fertilization, i.e. parthenogenesis (Figure 2E). After egg activation, many foci of MTs containing γ-tubulin appear. Two of these MTOCs join the female pronucleus to set up the first mitotic spindle in unfertilized eggs and drive their embryonic development. The other MTOCs disappear(61–63).

Strikingly, chromosome and centrosome cycles are uncoupled in certain divisions. For example during human male meiosis, centrioles duplicate from meiosis I to meiosis II, while there is no DNA replication at that stage (Figure 2F) (64,65). In contrast, in Drosophila there is no centriole duplication between meiosis I and II. Very little is known about the regulation of centriole duplication in male meiosis and why it is so diverse in different species. Uncoupling of the cycles can also be seen in certain zygotic systems and in cancer cells, where several cycles of centriole formation are observed during S phase arrest or after DNA damage, a phenotype we will discuss later in this review (66–73).

Centriole biogenesis: one molecular pathway fits all?

The recent sequencing and annotation of several genomes made it feasible to characterize proteomes and transcriptomes of centrioles and related structures, such as basal bodies and cilia, and to do large loss of function screens (74–86). Some proteins involved in assembling a centriole and in controlling that process are listed in Table 1.

Table 1.  Proteins involved in centriole regulation
MoleculeOrganismAssayCentrosome PhenotypeRefs
  1. This Table is divided in three sections: Trigger, Structural and Unknown Functions in Duplication, Cell Cycle/Checkpoint. Proteins categorized as Trigger are essential for centriole biogenesis and can induce it even in the absence of a mother centriole. Proteins categorized as Structural and Unknown Functions in Centrosome Duplication are involved in centriole duplication. Most of those proteins localize to the centrosome and their best characterized function is a centrosome-related one. While we have placed SAS-6 in this category because it may be a structural component of the cartwheel, high levels of SAS-6 lead to abnormal centriole numbers and therefore SAS-6 might also be considered a ‘trigger’. Proteins categorized as Cell Cycle/Checkpoint have been demonstrated to perform diverse roles during the cell cycle and/or are involved in cell cycle checkpoints. Little is known on how these proteins regulate centrosome number, whether they impinge directly on the centriole biogenesis machinery, or indirectly through changes in cell cycle timings.

  2. Ce, Caenorhabditis elegans; Cr, Chlamydomonas reinhardtii; Dm, Drosophila melanogaster; Hs, Homo sapiens; Mm, Mus musculus; Pt, Paramecium tetraurelia; Sc, Saccharomyces cerevisiae; Sp, Schizosaccharomyces pombe; Tt, Tetrahymena thermophila; Gg, Gallus gallus. Inhibition is used here for different means of inhibiting the function of a protein: dominant-negative, antibodies, chemical compounds. Mutation refers usually to complete or partial loss of function. PCM, pericentriolar material; RNAi, RNA interference; Reduplication refers to centrosome amplification in the context of cells arrested during S-phase.

Trigger
SAK/PLK4Dm; HsRNAi; mutationsNo duplication (Dm; Hs); no reduplication (Hs); no formation of basal bodies (Dm)(12,14,30,87–91)
  OverexpressionOverexpression: amplification (Dm;Hs); De novo formation (Dm) 
ZYG-1CeMutationsNo duplication (Ce)(92–94)
Structural and unknown functions in duplication
SPD2/CEP192Ce; Dm; HsRNAi; mutationsNo duplication (Ce); less recruitment of PCM (Ce, Dm and Hs); no duplication of basal body (Dm); no duplication in Hs (95) (differs (96)).(95–100)
Asterless/CEP152Dm; Dr; HsMutations; morpholinosProblems in PCM recruitment (Dm), no duplication (Dm; Dr)(101,102)
SAS6/DSAS6/Bld12Hs;Ce;Dm;CrRNAi; mutationsNo duplication (Hs; Dm; Ce); no reduplication (Hs)(29,87,88,92,94,103–106))
  OverexpressionAmplification (Dm; Hs)(105,106)
SAS5CeRNAi; mutationNo duplication(107)
SAS4/DSAS4/CPAPCe;Dm;Cr;HsRNAi; mutationsNo duplication (Hs;Ce;Dm;Cr); no reduplication (Hs)(15,27,28,92)
Centrin/Cdc31/VLF2/ CEN2/3/CEN1Hs;Sp;Sc;Cr;Pt;TtRNAi; mutationsNo duplication (Hs;Sp;Sc;Tt); differing duplication results (Hs); aberrant centriole segregation (Cr); aberrant duplication (Pt)(108–113)
SFi1ScMutationNo SPB duplication(114)
CP110/DCP110Hs; DmRNAiNo reduplication or amplification (Hs); no duplication (Dm)(85,88,115)
γ-Tubulin/TBGHs;Dm;Pt;Tt;CeRNAi; mutationsAberrant centriole duplication (Ce;Hs;Tt); problems in centriole structure, and separation (Pt;Dm)(88,103,116–119)
  Overexpression (Tt)De novo formation, amplification of basal bodies (Tt)(118,120)
Δ-TubulinHs;Cr;PtMutationsCentrioles with less C tubules; Abnormal transition zone(121–125)
ε -Tubulin/Bld2Xl;Hs;Cr;PtMutations; immunodepletionCentriole stability disrupted, only singlets (Cr); no duplication (Xl;Pt) problems in PCM organization (Xl)(121,123,126,127)
Ana-1DmRNAiNo duplication(84,85)
Ana-2DmRNAiNo duplication(84,85)
Ana-3DmRNAiNo duplication(84,85)
CentrobinHsRNAiNo duplication(128,129)
Cep135/ BLD10/DBld10Hs, Cr, DmRNAi; mutationsNo reduplication upon SAK/PLK4 overexpression (Hs); no duplication (Dm); disorganized microtubules (Hs); no duplication of basal bodies (Cr)(85,88,130–132)
  OverexpressionAccumulation of particles (Hs)(131)
Cell cycle and checkpoints
CDK1Dm, Sc (mitoticcyclins), GgMutations inhibitionAmplification with no disengagement (Dm); no duplication in the absence of both CDK1 and CDK2 activity (Gg)(133–135)
 Sc (mitoticcyclins)OverexpressionInhibition of centriole duplication 
CDK2/cyclin EHs; Xl; GlInhibitionNo reduplication; no centriole separation; no problems in duplication; needed for duplication in the absence of CDK1(136–141)
Nucleophosmin/B23Hs; MmInhibition; mutationAmplification(142)
CAMKIIXlInhibitionBlocks early steps in duplication(140)
Skp1,Skp2,Cul-1, Slimb (SCF Complex)Mm;Xl;Hs;DmMutations; inhibitionBlocks separation of M-D pairs and blocks reduplication (Xl); increased centrosome number (Dm; Mm)(143–147)
p53Hs, MmMutationsAmplification (reports differ) (Hs; Mm)(148–151)
SeparaseXlInhibitionBlocks centriole disengagement; impaired duplication (Xl)(136)
ShugoshinHsInhibitionCentriole disengagement (Hs)(152)
CHK1Hs; GgRNAi; mutationsNo centrosome amplification upon DNA damage (Hs; Gg)(153)
PLK1HsRNAiNo reduplication in S phase arrested cells(154)
PLK2HsRNAiNo reduplication in S phase arrested cells(155)
MPS1Hs;Mm;ScRNAi; mutationsNo reduplication (Hs; Mm: reports differ); normal duplication (Dm); no spindle-pole-body duplication (Sc)(156–158)
BRCA1Hs (mammaryderived cell lines; breast tissue); MmInhibition; RNAi; mutationsPremature centriole separation and reduplication in S-G2 boundary (Hs); amplification (Mm)(159–161)
Cdc14BHsRNAiAmplification(162)
PP2DmInhibitionCentrosome amplification(163)
  OverexpressionPrevents reduplication 
GemininHsRNAiCentrosome amplification (Hs)(164–166)
  OverexpressionCentrosome amplification impaired in S-phase arrested cells (Hs)(166)
Orc1HsRNAiCentriole amplification(167)
  OverexpressionInhibits reduplication 
Mcm5HsOverexpressionInhibits reduplication(168)

The core structure of the centriole, with its nine-fold symmetry, is notably conserved. The first described intermediate in centriole assembly showing nine-fold symmetry is the cartwheel (Figure 1A). In Paramecium and Chlamydomonas the cartwheel forms associated with a less characterized electron dense material (22,169). It is possible that components of the PCM, such as γ-tubulin, play a role early in the centriole biogenesis process and help forming the electron dense material and defining the number of centrioles formed (63,103,116–118,120,170–172). Another protein, asterless, which is associated with the centriole and the PCM, has recently been shown to be necessary for PCM recruitment and to play an early role in centriole duplication in Drosophila(85,101,102). Perhaps the PCM plays a role in recruiting the components needed to form the cartwheel and/or the centriole. Strikingly, the cartwheel can self-organize in vitro in a solution with basal body components (173), suggesting the intrinsic properties of its molecular components dictate its nine-fold symmetry. Bld10/CEP135 and SAS-6 are two components of the cartwheel (Table 1) (2,104,130,174). Bld10/CEP135 may also play a role at this stage because they are known to modify MT stability (131). Mutations in those molecules most often cause failure to form centrioles, with those that do form showing abnormal symmetry (2,29,87,88,92,103,104,130,175,176). The assembly and/or stabilization of centriole MTs is dependent on SAS-4 and γ-tubulin (Table 1) (15,88,89,92,119,177). Post-translational modification of MTs may also play a role (178).

Rare tubulin isoforms are involved in elongation and stabilization of centriole MTs, although they are not present in all species (Table 1, (121)). Centrins are members of a conserved subgroup of the EF-hand superfamily of Ca2+-binding proteins. In Saccharomyces cerevisiae, centrin is part of the structure that connects both Spindle Pole Bodies (SPBs), called half-bridge, and is essential for SPB duplication (108,109). Centrin mutants in Chlamydomonas reinhardii and Paramecium tetraurelia suggest that they play an important role in linking mother and daughter centrioles, and perhaps in determining the site of new centriole formation. In humans, centrin2 localizes to the distal end of centrioles but it is not clear whether it is involved in centriole duplication (88,110), (Table 1). Another component, CP110, which binds centrin, may be essential for capping the centriolar structure, regulating its assembly and transition to basal body (88,115,179,180).

SAK/PLK4, a protein kinase of the Polo-like protein kinase family, is essential for centriole biogenesis in flies (30,90) and human cells (30,91). This kinase is mutated in hepatocellular carcinomas, and SAK/PLK4 heterozygotic mice have elevated levels of cancer incidence (181). Moreover, mutation and overexpression of this kinase in fruitflies lead to tumorigenesis (12,182). Overexpression of SAK/PLK4 leads to an increase in the number of centrioles (30,88,89,91). Most strikingly, this kinase can trigger centriole formation de novo, in Drosophila eggs or tissue culture cells depleted of centrioles (87,89). Bld10, SAS-6 and SAS-4 all act downstream of SAK/PLK4 in canonical centriole biogenesis (88,89). SAS-6 and SAS-4 are also required downstream of SAK/PLK4 in de novo centriole formation (89). SAS-6 is also necessary for multiple centriole formation in multiciliated cells (60). Together those results suggest a unique pathway for centriole biogenesis in a variety of developmental contexts and species, triggered by SAK/PLK4 and executed by other molecules, such as SAS-6 and SAS-4 (14,24,89). The identification of a common trigger and common downstream components opens new avenues into the understanding of the control of the timing and number in centriole biogenesis, which we discuss below.

Control of Centriole Number, Localization and Timing of Biogenesis

The role of the mother centriole

Different types of control may operate in diverse cell types. In general, de novo formation leads to a high and variable number of centrosomes, suggesting a lack of control in their biogenesis. Procentrioles appear faster in the presence of a mother centriole than in de novo biogenesis in experimental systems, such as Chlamydomonas mutants (183), in vertebrate somatic cells after laser ablation of pre-existing centrioles (172,184) or in Drosophila melanogaster eggs after overexpression of SAK/PLK4 (89). In those three experimental systems the presence of a single centriole is also sufficient to prevent de novo biogenesis. Taken together, this suggests that the mother centriole might provide some catalytic/seeding role in daughter biogenesis. Perhaps the mother behaves as a platform for recruitment or stabilization of centriole assembly promoting factors, and integration of other cellular signals, and thus serves to regulate daughter centriole assembly (14,24). The catalytic role of the mother centriole may also be involved in inhibiting de novo formation elsewhere in the cell, but little is known about this process. Through these characteristics, the localization and timing of biogenesis of the new centriole within the cell is defined by the mother centriole, which integrates a variety of signals. This control can be very important for cell morphogenesis and polarity (21,25) and for coordination with cell cycle events through many cell cycle regulators that localize to the centrosome. The centrosome thus controls its own biogenesis, in response to a variety of signals from the cell.

How is the control of centriole number exerted within the canonical cycle?

First, there is control on how many daughters form close to each mother any time point (Figure 3A, B) (185). With exception of the multiciliated cells referred to above, this number is restricted to one. Second, there is only one time point in the cell cycle in which daughter centrioles can form, i.e. only one centriole cycle occurs within each chromosome cycle (Figure 3A, C). The two cycles are coordinated so that when one is delayed, the other stops, avoiding mistakes.

Figure 3.

Centrosome number controls. See text for references. A) Pro-centriole formation occurs during S phase, with the formation of a daughter centriole (D1) orthogonally to a mother centriole (M), reaching full elongation by the end of G2. At least two controls govern the centriole duplication cycle so that it occurs once and only once per cell cycle (185): B) Progeny number control-limits the number of daughters formed per mother centriole. Violation of this control leads to the formation of multiple daughter centrioles (D1-D6) per mother centriole simultaneously leading to the formation of multiple centrosomes. This is commonly observed in multiciliated cells and upon overexpression of SAK/PLK4 and SAS-6 in cycling cells. C) Centriole engagement is the tight association between mother and daughter centriole in S, G2 and mitosis (pre-anaphase). Violation of this control in embryonic systems and in certain transformed cells arrested in S-phase may lead to the reduplication of the S-phase centrosome (M and D1) and results in the formation of two new centrosomes in one reduplication round (M,D3) and (D1/D2). This also leads to centrosome amplification. Upon exit from the arrest, these cells may form abnormal mitotic spindles. Although two centrosomes are present in S and G2 phases of the cell cycle, only one is represented in this figure for simplicity.

How is the number of procentrioles formed per mother controlled?

Overexpression of SAK/PLK4 in humans and Drosophila and SAS-6 in humans leads to the formation of ‘flower-like structures’(88,91,105,147,186), a single mother surrounded by several daughters, similar to what occurs in multiciliated cells (Figure 3B). SAK/PLK4 and SAS-6 are highly expressed in multiciliated cells (60,187). To avoid the occurrence of flower-like structures, there must exist a very tight control of the protein levels and activity of SAS-6 and SAK/PLK4 in cycling cells. Indeed, interfering with SAS-6 degradation, via APC/C/Cdh1 (105) or with SAK/PLK4 degradation, via SCF (Skp, Cullin, F-box containing complex) /Slimb (143,147,186), two major ubiquitin ligase complexes involved in protein degradation in the cell cycle, leads to the formation of flower-like structures (Figure 3B and Table 1). SAK/PLK4 regulation by the SCF/Slimb complex in Drosophila is likely to be a conserved mechanism because the Slimb phosphodegron in SAK/PLK4 is conserved in both mice and humans (144,147). Moreover, knockout or depletion of the orthologue of Slimb, β-Trcp1, in mice leads to centrosome amplification (188). Skp1 and Cul-1, two members of the complex, have also been localized to the centrosome (145), and chemical inhibition of the proteasome was shown to lead to the formation of more than one daughter per mother centriole in mammalian cells (189). The biogenesis of many centrioles leads to the presence of many centrosomes and problems in cell division (12,89,105). It will be crucial in the future to better understand how levels and activity of SAK/PLK4 and SAS-6 are regulated in time and space.

How are the centrosome and chromosome cycles coordinated?

In the last 20 years, a complex system has been unravelled that controls DNA synthesis, restricting it to once per cell cycle (190,191). The DNA-licencing mechanism divides the cell cycle in two phases: a licenced state (mitotic exit/G1) where the cycle re-sets and initial steps for DNA replication are taken but progression cannot take place; and an unlicenced state where the initial steps cannot be taken and progression takes place (S phase). Could there be a similar licencing event in centrosome duplication? To address this, Wong and Stearns (192) carried out experiments similar to the classical study of Rao and Johnson (193) by fusing cells in different cell cycle stages. They found that centrosome duplication is controlled extrinsically, as S-phase cytoplasm advances the duplication of a G1 centrosome. However, there is a centrosome-intrinsic block to re-duplication, as the centrosome of a G2 cell does not duplicate in an S-phase cytoplasm. This is not because of an inhibitory effect from the cytoplasm because, when G2 cells were fused to G1 cells, the G1 centrosomes duplicated. So once centrioles have duplicated in S phase, they cannot duplicate again until the next S phase (192). One possibility is that the same factors that licence or inhibit DNA replication regulate the centriole duplication cycle. One such factor is Geminin, an inhibitor of the licencing DNA-replication factor Cdt1 (190). Indeed Geminin localizes at the centrosome (164) and Geminin RNAi leads to centrosome amplification (165). The mechanism of action of geminin in the centrosome cycle is not known. Another, non-exclusive possibility, is the control of centriole duplication by their disengagement, as this constitutes one prerequisite for growth of daughter centrioles (194) (Figure 1B). Indeed, mother and daughter disengagement is seen in centriole overduplication in transformed cells arrested in S phase, where there is uncoupling of both cycles (171,195) and daughter centrioles may still form in proximity of the mother (Figure 3B). Vidwans et al. (196) showed that centriole disengagement is dependent on the APC/C-Cdc20. Tsou and Stearns (136) suggested that Separase, a protease that is activated by APC/C-Cdc20 at the metaphase-to-anaphase transition to trigger separation of sister chromatids, triggers disengagement. This may licence the centrioles for a new round of duplication, perhaps by relieving a pre-existing block to duplication or exposing sites where new centrioles can bud (Figure 1B). A requirement for disengagement may explain why centrosomes in G2 cannot normally undergo further rounds of duplication.

Other cell cycle regulators have been shown to play a role in synchronizing the two cycles. In S phase, Cyclin Dependent Kinase 2 (CDK2), a kinase involved in promoting DNA replication, may also be involved in centriole duplication (137,138) (Table 1). One evidence of CDK2 involvement in this process comes from studies in C.elegans where it was shown that inhibition of CDK2 may be important for centriole disappearance in oogenesis.(197). Although CDK2 activity in somatic cycles is not essential, because CDK2 mutant mice are viable, with no ciliary or cell division defects. Centrioles can duplicate in CDK2 null cells (139) and Xenopus extracts depleted of CDK2 (140). However, CDK2 might set the right pace in centriole elongation (136). CHO (chinese hamster ovary) transformed cells (p53–/–), and nontransformed Chinese hamster embryo fibroblast (CHEF) cells arrested in G1 with low CDK2 activity can duplicate centrioles, but this process is slower than in Wild type (WT) cells with higher CDK2 activity (198). These results suggest that CDK2 might accelerate the centriole assembly process, to ensure that centrosomes rapidly complete their duplication as the cell proceeds through S phase. More detailed studies are needed to fully understand its role. Mitosis is most likely a non-permissive state for procentriole formation and elongation. When mitosis is prolonged in a variety of cell types, centrosome duplication cannot proceed (196,199). In yeast and flies, increase of mitotic CDK activity has an inhibitory role over spindle pole body and centrosome duplication, but the mechanism of this inhibition and whether it may enforce any of the controls discussed in Figure 3 remain unclear (2,133,134).

As discussed above, the coordination between the centrosome and chromosome cycle may emerge through the existence of cell cycle regulators that have distinct inputs in the centriole cycle (e.g. CDK2, Separase, CDK1, SCF/Slimb, APC/C). However, it is not known how many of those inputs are read by the centriole assembly machinery to confine their activity to a particular window of time. As a possibility, SAS-6 levels are highest in mitosis prior to its degradation by the APC-Cdh1 (105), perhaps SAS-6 acts in this small window of time, after Separase activation and before SAS-6 degradation.

Controlling centriole inheritance

At the end of mitosis mother and daughter centrioles disengage, but remain connected until the beginning of mitosis in the next cycle. Centrosomes have to be well separated during mitosis, with one centrosome at each pole of the bipolar spindle, so that each cell inherits two centrioles (194). C-Nap, a molecule that is recruited to the proximal end of both mother centrioles, binds rootletin, a fibre-forming molecule. Rootletin and another molecule, CEP68, may provide a dynamic link between mother centrioles (200–202). Nek2-mediated phosphorylation of C-Nap1 at the entry of mitosis releases that link. Overexpression of Nek2 leads to premature centrosome disjunction (200,201). Nek2 is normally counteracted by PP1 in interphase (194,203).

The connection between centrioles and the spindle is also very important, ensuring that centriole segregation to daughter cells occurs normally. This connection seems to be dependent on the centrosomal MT nucleating activity. Mutations or depletion of CNN/CDK5RAP2, a molecule needed for γ-tubulin attachment, leads to a looser centriole–PCM connection (32,204) and errors in centriole segregation (32).

What if it all goes wrong… abnormal regulation of centriole number?

Cell cycle checkpoints ensure that critical steps in the chromosome cycle only occur when previous events have finished with no errors. If there are problems, such as in DNA replication, the cycle is halted until they are corrected. Chromosome and centrosome cycles are also coordinated in this respect. When S phase is arrested in non-transformed cells, both cycles stop. How the centrosome cycle is halted in those cases is not clear, but it may be through the checkpoint machinery feeding into the cell cycle/centrosome regulators referred above. This cross talk may be affected in fast early embryonic cycles and in some transformed cells, such as CHO and U2OS osteosarcoma cells, which show centrosome reduplication when arrested in S phase (66–70). It is interesting that CDK2 is necessary for centrosome reduplication induced by the oncogenic papilloma virus HPV16-E7 (139) and for centrosome reduplication in Xenopus and human cells arrested in S phase (137,138,140,141). There is clearly diversity in requirement for CDK2 in centriole biogenesis in arrested cells when comparing duplication in non-transformed cells and reduplication in transformed/embryonic cells. Perhaps the role of CDK2 in setting the pace in centriole duplication discussed above reflects a function in providing centriole precursors which may become limiting when reduplication cycles are involved. Alternatively, it may reflect a different regulation of the first events in centriole biogenesis in non-transformed and transformed/embryonic cells where cell cycle regulation might be different. This difference in requirement in normal cycles and reduplication cycles is also seen with ORC1, a subunit of the origin recognition complex that prevents centriole reduplication but not canonical centriole duplication. It has been suggested that ORC1 acts through the inhibition of cyclinE activity at the centrosome at the G1/S boundary (167). Minichromosome maintenance 5 (MCM5), another DNA prereplication complex factor, also interacts with CyclinE, independently of CDK2 activity, and inhibits reduplication cycles (168). More experiments are needed to further understand this link between the DNA replication licencing machinery and centriole biogenesis.

Which molecules lie downstream of the requirement for CDK2 in centriole reduplication? Several candidate molecules have been shown to play a role in centriole duplication and reduplication, and to be regulated by CDK2. That is the case of E2F, MPS1, nucleophosmin (NPM)/B23 and CP110 (115,137,142,205). The kinase Mps1, which is phosphorylated and stabilized by CDK2 (206,207), has an important role in the spindle assembly checkpoint, and a controversial role in centriole reduplication (Table 1) (156,157,205). NPM is a CDK2 substrate, associated with oncogenesis (208), that shuttles between the nucleus and the centrosome (142). When present on the centrosome, NPM can prevent unscheduled centrosome amplification (142,209).

Other tumour suppressors and checkpoint proteins have been associated with abnormal centrosome numbers and centrosome inactivation. For example, loss of p16INK4a, which is associated with cancer, can lead to centrosome amplification through centriole splitting (210). Loss or mutational inactivation of p53 can lead to centrosome amplification, but this is not the case for all cell types (148,149,211). It is not clear how p53 could prevent centrosome amplification, but it may induce the death of cells with amplified centrosomes (212), some of which arise through failed cell division (148). It has also been suggested that p53 might downregulate SAK/PLK4 (213), hence controlling centrosome number. BRCA1, an ubiquitin ligase, is another tumour suppressor associated with centrosome amplification, which predisposes women to ovarian and breast cancer. It is not clear how BRCA1 prevents centriole amplification, but it has been suggested that BRCA1 ubiquitinates γ tubulin (214) and NPM (215), two regulators of centriole biogenesis.

Finally, DNA damage is associated with changes in centrosome number in transformed cells and embryos (66,69,71–73). Disruption by gene targeting or pharmacological inhibition of checkpoint kinase 1 (Chk1), an established transducer of Ataxia telangiectasia and Rad3 related (ATR) and Ataxia telangiectasia mutated (ATM) dependent signalling in response to DNA damage, suppres DNA-damage-induced centrosome amplification (153). Because Chk1 localizes to the centrosome, it may act indirectly through its function in cell cycle regulation by interfering with CDKs, or alternatively, impinging directly on players in centriole duplication (153). In Drosophila syncytial embryos, DNA damage leads to centrosome inactivation, resulting in the loss of damaged nuclei. That process is dependent on Chk2 (216). In human cancer cells, DNA damage may also lead to Chk2 dependent centrosome fragmentation, which has been suggested to eliminate cells with damaged DNA (217). It is not known whether those pathways are also involved in centriole elimination and inactivation in the developmental systems discussed in this review. In the future it will be important to dissect the relevance of those mechanisms in non-transformed and transformed cells and define how they regulate the centriole biogenesis machinery.

Ultimately, it is likely that the mechanisms involved in centriole biogenesis discussed here, all contribute to a robust and stable homeostatic control of centriole number (218). For example, if errors occur and cells are born without centrioles, these structures can be formed de novo. More strikingly, at least in Chlamydomonas, if a cell is born with supernumerary centrioles, centriole biogenesis is inhibited in those cells and centriole number returns to normal levels (218). What could this control system be? Because the PCM has a role in controlling centriole assembly (63,116,170,171,117,120,118), it is possible that the amount of PCM is limited in cells with more of these structures, resulting in fewer centrioles being formed. Further studies are needed to understand the cellular and molecular mechanisms involved in this homeostasis.

Implications for Centrosome/Centriole-Associated Disease

Centrosomes are associated with tumorigenesis. It was Theodor Boveri who proposed that abnormal centrosome numbers could lead to multipolar cell division (219,220). Indeed, numerical and structural centrosome aberrations are an early event in many cancers such as colorectal, uterine cervix, prostate and female breast carcinomas, and haematological malignancies that correlate with poor prognosis (5–10). The origin of those changes is not clear. Tampering with some of the biogenesis controls referred above is one possibility, because many tumour suppressors are involved in centriole amplification. However, it is also possible that in some tumours, changes in centrosome number arise through defects in cytokinesis, splitting of centriole pairs or cell fusion (5). Although Boveri predicted high centrosome numbers would result in abnormal chromosome segregation and high chromosome instability, cancer cells might evolve mechanisms, such as the one of centrosome clustering and/or inactivation (Figure 2C), of avoiding abnormal cell division. If aneuploidy is rare, how do supernumerary centrosomes lead to cancer? Recent papers suggest an alternative model based on stem cell homeostasis (12,182). Because asymmetric cell division may be affected when centrosome number and function is changed, perhaps this could lead to an expansion of cells with higher proliferative potential in the stem cell compartment. Indeed, recent evidence suggests alterations in adult stem cells could be at the origin of certain tumour types (221).

It might prove rewarding to explore the potential of centrosome-related processes in therapeutic approaches, as they may show more specificity towards dividing cells. Other drugs that target cell division, such as taxanes and vinca alkaloids, also interfere with the dynamics of MT filaments in normal cells, such as neurons, leading to peripheral neuropathy. Drugs that inhibit centrosome duplication or separation induce the formation of monopolar spindles. Drugs that target centrosome clustering in cancer cells displaying multiple centrosomes cause the formation of aberrant mitotic spindles. Severe defects in mitosis such as monopolar or multipolar spindle formation may cause checkpoint arrest or non-viable daughter cells (mitotic catastrophe) (11,222).

Several other diseases have been associated with centrosome function. Recently, a variety of genes involved in microcephaly/dwarfism have been identified, all shown to play an important role in centriole assembly and centrosome function. That is the case of pericentrin, SAS-4/CENPJ/CPAP, Centrosomin/CDK5RAP2, ASPM/Asp and MCPH1. Mutations in these genes are associated with reduction in brain size and in certain cases with dwarfism. It is not yet clear what is the precise cell biology mechanism associated with the disease phenotype, but the importance of the centrosomes in mitotic spindle organization, in asymmetric cell division and in DNA damage signalling have been proposed to play an important role (223–225).

Perspectives

In recent years there has been a great improvement in our understanding of the molecular mechanisms regulating centrosome formation. The knowledge of the molecules involved in triggering centriole biogenesis and in defining centriole architecture should allow us to unravel their regulation in the normal cell cycle of non-transformed cells, after DNA damage, and in the abnormal cycles of cancer cells. Model organisms and non-transformed human cells should play an important role in the definition of the basic centriole biogenesis machinery and its controls. This will be relevant for understanding important processes in cell differentiation and embryonic development, such as centrosome inactivation and loss, and asymmetric cell division. Finally, understanding the complex regulation of centriole biogenesis in different cell types and in tumours should help define how to better target cancer cells.

Acknowledgments

We thank Z. Santos, A. Rodrigues-Martins and J. Borrego-Pinto for their help with images and critical reading of manuscript. We thank R. Kuriyama and C. Morrison for their critical reading and very helpful comments on the manuscript. We thank an anonymous reviewer for very helpful criticisms. We are grateful to grants from Fundação Calouste Gulbenkian, Fundação para a Ciência e Tecnologia (FCT; PTDC/SAU-OBD/73194/2006; PTDC/BIA-BCM/73195/2006), Crioestaminal (Viver a Ciência) and to an EMBO Installation Grant to MBD. I.C.F. and I.B. are recipients from scholarships from FCT.

Ancillary