A primary microcephaly protein complex forms a ring around parental centrioles.

Autosomal recessive primary microcephaly (MCPH) is characterized by a substantial reduction in prenatal human brain growth without alteration of the cerebral architecture and is caused by biallelic mutations in genes coding for a subset of centrosomal proteins. Although at least three of these proteins have been implicated in centrosome duplication, the nature of the centrosome dysfunction that underlies the neurodevelopmental defect in MCPH is unclear. Here we report a homozygous MCPH-causing mutation in human CEP63. CEP63 forms a complex with another MCPH protein, CEP152, a conserved centrosome duplication factor. Together, these two proteins are essential for maintaining normal centrosome numbers in cells. Using super-resolution microscopy, we found that CEP63 and CEP152 co-localize in a discrete ring around the proximal end of the parental centriole, a pattern specifically disrupted in CEP63-deficient cells derived from patients with MCPH. This work suggests that the CEP152-CEP63 ring-like structure ensures normal neurodevelopment and that its impairment particularly affects human cerebral cortex growth.

We ascertained a consanguineous Pakistani family in which three members had primary microcephaly at birth (head circumference −4SD to −6SD) and by five years showed mild/moderate mental retardation (Fig. 1 A and Supplementary Note). They were microcephalic and, to a lesser degree, had proportionate short stature, leading to a diagnosis of MCPH with growth retardation or mild Seckel syndrome [16]. There was no genetic linkage to known MCPH or Seckel syndrome loci. Autozygosity mapping revealed a single, concordant homozygous locus of 24Mb on chromosome 3 shared by the patients (Fig. 1B). Within this genomic region we identified a pathogenic mutation in the centrosomal gene CEP63 [17]; c.129G>A causing an immediate STOP codon W43X in exon 4. Alternative splicing of CEP63 produces four products that are expressed in foetal brain (Fig. 1C, Supplementary Fig. 1A). Similar to other MCPH genes[18-21], CEP63 underwent positive Darwinian selection during the evolutionary period that higher ape and human brains have grown in relative size (Supplementary Fig. 1B). An antibody recognising CEP63 stained centrosomes of apical neural precursors in mouse brain (Fig. 1D). CEP63 was enriched in centrosomes of B-lymphocytes derived from parent-of-patient (CEP63+/−) (Fig.1E and Supplementary Fig. 2A). A weak CEP63 signal was detectable in centrosomes of patient cells (CEP63−/−), implying that c.129G>A is a hypomorphic mutation. Indeed, CEP63 transcripts were present in patient cells (Supplementary Fig. 2B).

Fig.1

Identification of a MCPH-causing mutation in the CEP63 gene

(A) Schematic diagram shows simplified pedigree. Filled in circles indicate individuals with MCPH. Only the closest link between family members is depicted. Pictures of three afflicted individuals are shown on right. (B) The three affected individuals share only one region that was concordantly homozygous in chromosome 3. The region was defined by heterozygous allele results for the markers D3S3513 at 137cM/121Mb and D3S1569 at 158cM/145Mb. The three candidate genes sequenced are shown within the linkage interval. (C) Exons of CEP63 gene are shown as bars; solid bars are translated. There is a single CpG island for the gene with transcription starting in exon 3. The position of the nonsense mutation c.129G>A is located in exon 4. A predicted Kozak sequence downstream of the mutation is shown (labelled as ‘K’). Alternatively spliced exons are shown in blue and red (see Supplementary Fig. 1A). CEP63 protein is shown with six predicted coiled-coil domains marked in green. Framed areas correspond to differentially spliced regions. The CEP63 antibody was raised against a recombinant fragment of CEP63 shown. (D) CEP63 protein expression in mouse cerebral cortex neuroepithelium at E13.5. Mitotic cell is marked with asterisk. Magnified views of framed areas are shown on right. Samples were co-stained with antibodies against CEP63 (red in merge) and the centrosomal protein γ-tubulin (green in merge). DNA is in blue. (E) CEP63 protein expression in parent-of-patient (CEP63+/−) and patient (CEP63−/−) lymphocytes. Mitotic cells are marked with asterisks. Magnified views of centrosomes are shown on right. Cells were co-stained with antibodies against CEP63 (red in merge) and the centriolar protein, centrin-3 (green in merge). DNA is in blue. Scale bars=10μm.

MCPH-linked genes encode proteins that function in mitotic spindle assembly and/or centriole biogenesis [10]. Consistently, CEP63 has been implicated in mitotic entry and spindle formation [22,23]. To address the role of CEP63 in the centrosome, we generated a CEP63 knockout in DT40 chicken B-lymphocytes using homologous gene-targeting. Exons 5-9 of CEP63 were replaced with drug resistance genes (CEP63KO; Supplementary Fig. 3). CEP63 antibody stained centrosomes of wild-type (WT), but not CEP63KO cells (Fig. 2A). Since the antibody epitope was largely disrupted in CEP63KO cells, these results were confirmed by labelling gene products with a GsTAP tag (comprising protein G and streptavidin-binding protein[24]) inserted in-frame to the 3′ ends of the CEP63 alleles of WT and CEP63KO cells (TAP-WT and TAP-CEP63KO cells, respectively; Supplementary Fig. 3). TAP-WT cells expressed a single protein product (TAP-CEP63) that localised to centrosomes throughout the cell cycle (Fig. 2B,C and Supplementary Fig. 4A). No product was detected in TAP-CEP63KO cells, indicating exons 6-16 were not translated. Homozygous CEP63KO clones displayed increased population doubling times (Supplementary Fig. 4B), but cell cycle profiles were similar between WT (G1:20±3%, S:51±5%, G2/M:25±6%) and CEP63KO (G1:20±2%, S:50±2%; G2/M:23±3%). However, CEP63KO cells exhibited mitotic spindle defects such as monopolar and multipolar spindles [22] (Fig. 3A). Time-lapse microscopy confirmed these phenotypes (Supplementary Fig. 5 and Videos 1,2).

Fig.2

Disruption of the centrosomal gene CEP63 in vertebrate cells

(A) WT and CEP63KO DT40 cells were co-stained with antibodies against CEP63 (red in merge) and γ-tubulin (green in merge). Magnified views of framed areas are shown in insets. DNA is blue. Scale bar=5μm. (B) Western blots of cytoplasmic cell extracts (CCE) from WT and gene-disrupted DT40 cell lines were probed with protein G antibodies. TAP cell lines contain in-frame GsTAP tag in their CEP63 alleles. α-tubulin serves as loading control. (C) Summary of the localisation and expression of tagged and truncated CEP63 products in the DT40 cell lines specified. n/a: the antibody epitope is largely destroyed in predicted protein product.

Fig.3

CEP63 is required for maintaining normal centrosome numbers

(A) Examples for mitotic spindle defects of CEP63KO DT40 cells are shown. Cells are stained with antibodies against the centrosome marker CDK5RAP2 (red in merge) and α-tubulin (green in merge). DNA is in blue. Graph shows quantification of spindle phenotypes (n=3). Scale bars=5μm. (B) Graph shows number of centrioles in interphase WT and CEP63KO DT40 cells (n=3). Antibodies against centrin-3 were used for experiment. (C) Graph shows number of centrioles in monastrol-induced monopoles of WT and CEP63KO DT40 cells (n=3). Note that centriole numbers 1 and 3 could sometimes reflect insufficient spatial resolution of engaged centriole pairs. Examples for CEP63KO monopoles with different centriole numbers are shown; cells were co-stained with antibodies against the spindle pole marker TACC3 (red in merge) and centrin-3 (green in merge). DNA is in blue. Magnified views of centrin-3 staining in framed areas are shown in insets. Similar results were obtained using polyglutamylated tubulin as centriole marker; 0-2 centrioles were present in 42%, 3-4 centrioles in 52% and >4 centrioles in 6% of cells (100 CEP63KO mitotic cells). (D) Graph shows number of centrioles in untreated monopolar CEP63KO spindles (n=3). Antibodies against centrin-3 were used for experiment. (E) Graph shows number of centrioles in mitotic cells with multipolar spindles (n=3). On the left an example for CEP63KO multipolar spindle is shown with two singlet centrioles (asterisks) at spindle poles. Cell was co-stained with antibodies against the spindle pole marker TACC3 (red in merge) and centrin-3 (green in merge). DNA is in blue. Magnified views of centrin-3 staining in framed areas are shown in insets. Scale bars=5μm. Bars in graphs correspond to mean±standard deviation (s.d.). p values were obtained by two-tailed, unpaired Student t test. N.S.=not significant.

After mitosis a cell inherits a centrosome comprising a mother and daughter centriole embedded in the pericentriolar matrix (PCM). The centrosome duplicates in S-phase when a procentriole forms adjacent to each parental centriole engaging in an orthogonal configuration [25] (see schematic in Supplementary Fig. 6). The two centrosomes then mature by PCM expansion and separate to form the mitotic spindle poles. Monopolar spindles could therefore arise from defects in centrosome maturation, separation or duplication. Since PCM components γ-tubulin (Fig. 2A), CDK5RAP2 and TACC3 [26] (Fig. 3) were present in CEP63KO spindle poles, CEP63 deficiency does not preclude centrosome maturation. Furthermore, centrosome separation appeared normal in CEP63KO cells (Supplementary Fig. 7, Videos 3-8). Thus, monopolar spindles must reflect a defect in centrosome duplication. Indeed, interphase CEP63KO cells contained fewer centrioles than WT (Fig. 3B). Since interphase cells normally contain 2-4 centrioles and in some cells multiple small centrin foci were visible, a more accurate quantification of centriole numbers was achieved in mitotic cells arrested with monastrol [27]. Monastrol prevents centrosome separation, thereby converging centrioles into a monopole. Consistent with the presence of 2 duplicated centrosomes 80% of WT but only 40% of CEP63KO monopoles contained 4 centrioles (Fig. 3C). 40% of monastrol-induced and 70% of untreated CEP63KO monopolar spindles contained only 1-2 centrioles (Fig. 3 C,D).

If a mitotic cell contains only 2 centrioles, it could have defects in procentriole assembly or disengagement, a process whereby the tight link between parental and procentrioles is removed, allowing procentrioles to duplicate in the next cell cycle [28,29] (Supplementary Fig. 6). CEP63 is not a core centriole assembly factor, since CEP63KO cells do not become depleted of centrosomes over time (Supplementary Fig. 8A). Moreover, CEP63KO amplify centrosomes when treated with a DNA replication poison (Supplementary Fig. 8B). Since overduplication depends on recurring cycles of procentriole assembly and disengagement[30], CEP63 is dispensable for these steps during an extended cell cycle arrest. Yet, during normal cell cycles CEP63 deficiency could cause a delay in procentriole assembly leading to a stochastic failure in centrosome duplication and thus monopolar spindles. Such a delay, however, does not explain the increase in multipolar spindles in CEP63KO cells. Since centriole pairs segregate with mitotic spindle poles, aberrant engagement between mother and procentrioles could cause asymmetric centriole segregation and multipolarity [31]. Consistent with a potential defect in centriole engagement, nearly half of CEP63KO multipolar spindles contained 4 centrioles in total that were distributed as singlets between 3 or 4 spindle poles (Fig. 3E). This extreme phenotype occurred in only 7% of CEP63KO cells, but distances between mother and daughter centrioles were elevated in mitotic CEP63KO centrosomes (Supplementary Fig. 8C). Since a marker of centriole distal ends was used to evaluate distances (centrin-3), abnormal centriole structure or length could distort the data. However, transmission electron microscopy revealed no such anomalies (Supplementary Fig. 8D). In summary, abnormal centrosome numbers in CEP63KO cells could result from delayed procentriole assembly and impaired engagement.

To examine the molecular mechanism of CEP63 action, interactors of CEP63 were identified by affinity-purifying GsTAP-CEP63 from biallelically tagged TAP-WT cells (Fig. 4A,B and Supplementary Fig. 9). WT and TAP-WT cells were differentially labelled with light (WT) or heavy (TAP-WT) forms of amino acids. Streptavidin-bound fractions were analysed by mass-spectrometry and ratios of light to heavy peptides (L/H) were determined using SILAC [32]. Two proteins exhibited L/H ratios lower than 0.25: CEP63 and CEP152 (Supplementary Tables 1 and 2). Mammalian CEP63 was among nine proteins reported as CEP152 interactors by the Mitocheck consortium [33]. Binding between CEP152 and FLAG-tagged CEP63 was confirmed in human cells (Fig. 4C), an interaction mediated by the C-terminal part of CEP63 (Fig. 4D).

Fig.4

CEP63 forms a protein complex with CEP152

(A) Diagram illustrates experimental design to identify CEP63-interacting proteins. (B) Western blot shows an example for single step purification of GsTAP-tagged CEP63 protein. Cell lysates of WT or TAP-WT DT40 cells (Input=I) were incubated with streptavidin-agarose resin. Bound proteins were eluted from resin with biotin (Eluate=E). Immunoblot was probed with protein G antibodies to detect GsTAP-tagged CEP63. (C) Cytoplasmic cell extracts (CCE) were prepared from HeLa cells that were mock transfected (−) or transfected with FLAG-CEP63 (+). Anti-FLAG antibodies were used for immunoprecipitation (IP). Immunoblots were probed with antibodies against CEP63 and CEP152, as marked. (D) Cytoplasmic cell extracts (CCE) were prepared from HeLa cells that were mock transfected (−) or transfected with FLAG-CEP63-CT (+). Anti-FLAG antibodies were used for immunoprecipitation (IP). Immunoblots were probed with antibodies against CEP152 and FLAG, as marked.

CEP152 is a key regulator of centriole biogenesis and centrosome function [12-15,34,35] and is mutated in MCPH and Seckel syndrome [9,36]. We thus wondered if abnormal CEP152 function underlies the phenotypes of CEP63KO DT40 cells. Existing antibodies did not recognise chicken CEP152, so CEP152 localisation in DT40 cells was assessed with ectopically expressed STREP-tagged human CEP152. While STREP-CEP152 accumulated in WT centrosomes, it was weak in interphase and absent from mitotic CEP63KO centrosomes (Fig.5A). To restore CEP152 levels in CEP63KO centrosomes CEP152 was fused to PACT [37], a centrosome localisation signal that mediates CEP63-independent targeting of GFP (Fig. 5B, Supplementary Fig. 7). Indeed, unlike STREP-CEP152, STREP-CEP152-PACT localised to CEP63KO centrosomes throughout the cell cycle (Fig. 5C). To address whether phenotypes of CEP63KO cells were due to diminished centrosomal CEP152, stable CEP63KO clones were generated by introducing the following transgenes: puromycin resistance gene only or in combination with FLAG-CEP63, STREP-CEP152 or STREP-CEP152-PACT (Fig. 5B-D). Remarkably, expression of STREP-CEP152-PACT complemented the mitotic spindle phenotypes and restored centriole numbers in CEP63KO cells (Fig. 5E, F). STREP-CEP152-PACT complemented the phenotypes as efficiently as FLAG-CEP63, implying that CEP63 maintains centrosome numbers via centrosomal recruitment of CEP152.

Fig.5

CEP63-dependent centrosomal accumulation of CEP152 maintains normal centrosome numbers

(A) Localisation of STREP-tagged human CEP152 in transfected WT and CEP63KO DT40 cells. Magnified views of framed areas are shown in panels below. Cells are co-stained with antibodies against centrin-3 (green in merge) and CEP152 (red in merge). (B) Outline of constructs that were transfected into CEP63KO DT40 cells to derive cell lines stably expressing transgenes. (C) Cytoplasmic cell extracts of CEP63KO-derived DT40 cell lines (nomenclature as in B) were immunoblotted with antibodies against CEP152 and as a loading control, α-tubulin. Below mitotic cells co-stained with antibodies against centrin-3 (green in merge) and strep-tag II (red in merge) are shown. Magnified views of framed areas are shown. DNA is in blue. (D) Cytoplasmic cell extracts of CEP63KO-derived DT40 cell lines (nomenclature as in B) were immunoblotted with antibodies against CEP63 and as a loading control, α-tubulin. Below mitotic cells co-stained with antibodies against centrin-3 (green in merge) and CEP63 (red in merge) are shown. Magnified views of framed areas are shown. DNA is in blue. (E) Graph depicts quantification of spindle phenotypes in CEP63KO-derived DT40 cell lines (nomenclature as in B; n=2; >350 mitotic cells per clone). (F) Graph depicts number of centrioles in monastrol-induced monopoles of CEP63KO-c152 or CEP63KO-c152-pact DT40 cell lines (nomenclature as in B; n=2). Bars in graphs correspond to mean±s.d. p values were obtained by two-tailed, unpaired Student t test. Scale bars=5μm.

Super-resolution microscopy (3D-SIM) revealed that CEP63 formed a discrete centrosomal ring in TAP-WT DT40 cells (Fig. 6A). The diameter of the ring was approximately 280 nm, similar to centriole diameter. Its position relative to centrin-3 indicated a proximal location for CEP63. Partial overlap was seen with polyglutamylated tubulin antibody marking centriole walls (Supplementary Fig. 10A). The PCM component γ-tubulin was organised around the circumference of the ring (Supplementary Fig. 10B). CEP63 thus occupies a position between the proximal end of parental centriole wall and the PCM. From G1 until mitotic exit only parental centrioles carried a ring. Procentrioles acquired CEP63 signal in late mitosis, but their rings were completed only after cytokinesis, coinciding with centriole disengagement and PCM recruitment (Fig. 6A) [28,31]. The ring was also present in humans (Supplementary Fig. 10C) and appeared lateral to SAS-6 found at the proximal end of procentrioles [38] (Fig. 6B).

Fig.6

CEP63 and CEP152 form a ring around parental centrioles, a structure disrupted in CEP63−/− patient cells

(A) 3D-SIM images are shown of TAP-WT DT40 cells co-stained with antibodies against protein G (recognises TAP-CEP63; red in merge) and centrin-3 (green in merge) antibodies. Each centrosome is shown at high magnification with separate images depicting protein G (top panels) and centrin-3 (middle panels) stainings. Specific cell cycle stages are stated. All images are maximum projections apart from those referred to as 3D rotations. The latter represent alternative views of centrioles (generated by rotations of 3D volumes) to highlight particular features. Schematic shows relative positions of centrin and CEP63 staining. (B) 3D-SIM image of CEP63+/− human lymphocyte co-stained with antibodies against CEP63 (green in merge) and the daughter centriole marker, SAS-6 (red in merge). Schematic shows relative positions of SAS-6 and CEP63 staining. (C) HeLa cells stably expressing FLAG-CEP63 were co-stained with antibodies against FLAG (green in merge) and CEP152 (red in merge). 3D-SIM images of centrosomes are shown. (D) 3D-SIM images of interphase (top) and mitotic (bottom) CEP63+/− and CEP63−/− human lymphocytes are shown. Cells were co-stained with antibodies against CEP152 (red in merge) and centrin-3 (green in merge). DNA is in blue. Framed areas are shown in higher magnification with corresponding CEP152 (left panels) and centrin-3 (middle panels) signal. Graphs on the right show percentages of CEP63+/− and CEP63−/− centrosomes with specific CEP152 localisation pattern (n=2). (E) CEP63+/− and CEP63−/− human lymphocytes were treated with DMSO, PLK1 inhibitor BI-2536 (BI) or monastrol (mon) for 90 minutes prior to fixation. Cells were co-stained with antibodies against CEP152 (red in merge) and centrin-3 (green in merge). Examples for centrosomes from individual mitotic cells are shown. Note that mon and BI induce monoastral spindle, whereas those in DMSO remain bipolar. Scale bar=0.5μm. Distribution of mean intensities of centrosomal CEP152 signal in patient cells is illustrated in graph. Number of centrosomes scored: CEP63+/−: n=30 DMSO- and n=38 BI-treated; CEP63−/−: n=26 DMSO- and n=24 BI-treated. Background CEP152 signal was measured in areas of CEP63−/− cells that do not contain centrosomes (n=26). Box plots: length of whiskers is at 5th and 95th percentiles, the box shows interquartile (25-75) range and horizontal line represents the median. p values were obtained by two-tailed, unpaired Student t test. N.S.=not significant.

Reminiscent of the CEP63 ring, CEP152 was found to localise to the proximal end of the mother centriole wall [13,15]. Indeed, CEP152 and FLAG-CEP63 co-localised in HeLa cells (Fig. 6C). Similar to CEP63KO DT40 cells, siRNA-mediated depletion of CEP63 prevented centrosomal accumulation of CEP152 in human cells (Supplementary Fig. 11). This prompted us to investigate the localisation and sub-centrosomal distribution of CEP152 in CEP63+/− and CEP63−/− patient lymphocytes. Despite CEP63+/− and CEP63−/− cells expressing similar amounts of CEP152, protein levels were reduced in CEP63−/− centrosomes (Supplementary Figs. 2A and 12A, B). While most interphase CEP63+/− centrosomes contained CEP152 rings, these were rare in CEP63−/− centrosomes (Fig. 6D). Like CEP63KO DT40 centrosomes, mitotic CEP63−/− centrosomes were devoid of CEP152, a defect complemented by expression of FLAG-CEP63 (Supplementary Fig. 12C). CEP63−/− cells exhibited no mitotic spindle abnormalities, but a small increase in centriole distances was detected in mitotic cells (Supplementary Fig. 12D).

Why is CEP152 localisation particularly affected by CEP63 deficiency during mitosis? We hypothesised that CEP63 opposes an activity that removes/modifies centrosomal CEP152. A good candidate is POLO-LIKE KINASE 1 (PLK1) that links centriole biogenesis to cell cycle progression[28,30,31,39]. Indeed, treatment with the PLK1 inhibitor BI-2536[40] dramatically increased centrosomal CEP152 levels in mitotic CEP63+/− cells (Fig. 6E). This was not due to mitotic arrest imparted by BI-2536, since monastrol had no effect. While we cannot exclude that PLK1 modifies the epitope of the CEP152 antibody, our results nonetheless implicate PLK1 in regulating CEP152. BI-2536 treatment produced a small increase in CEP152 at CEP63−/− centrosomes, indicating that PLK1 inhibition can only partially override the defect in CEP152 recruitment caused by CEP63 deficiency.

CEP152 scaffolds procentriole formation by promoting centrosomal accumulation of CPAP and POLO-LIKE KINASE 4 (PLK4) [13-15], a kinase that controls procentriole formation [25,29,41-43]. In both chicken CEP63KO and patient CEP63−/− cells CEP152 was absent from mitotic centrosomes, but residual CEP152 protein was detectable in interphase. Since PLK4 activity peaks in mitosis [44], depending on the window of PLK4 activity and the kinetics of CEP63-independent accumulation of CEP152, a delay in CEP152 recruitment could preclude procentriole formation especially in rapidly proliferating cells like DT40 with 8 or apical progenitors with 10-14 hour cycles [45]. In contrast, CEP63−/− patient lymphocytes have a 24-hour doubling time that could explain lack of centrosome duplication defects in these cells. Our finding that a CEP152-PACT fusion product complemented centrosome phenotypes of CEP63KO cells suggests that the role of CEP63 in centrosome duplication is to maintain CEP152 at centrosomes. How can PACT mimic the action of CEP63? While PACT is derived from the PCM component AKAP450[37], it accumulates on centrioles[46] and therefore STREP-CEP152-PACT is probably recruited near the centriole wall in CEP63KO cells.

The position of the CEP63-CEP152 ring coincides with the site where a cohesive structure is predicted to engage parental and procentrioles. A limited role for CEP63 in engagement is supported by the co-existence of mono- and multipolar spindles in CEP63KO populations. This function is consistent with findings in Xenopus, where DNA damage-induced removal of CEP63 from the centrosome caused multipolar spindle formation[22]. Further support for a link between MCPH and centriole engagement emerged from studies implicating MCPH7/STIL/ana2 [47] and MCPH3/CDK5RAP2 [48,49] in this process. During neurogenic mitoses of the developing mouse neocortex, progenitors retain the centrosome containing the old mother centriole [50], whereas newborn neurons inherit the new mother centriole. Since this asymmetry contributes to cell-fate determination[50], delayed procentriole assembly and impaired engagement between mother and daughter centrioles can alter stem cell fate. In summary we show that the CEP63-CEP152 MCPH complex is an essential part of the molecular machinery controlling centrosome numbers, and defects in either component result in a diminished pool of neural precursors that cannot provide for the enlarged human brain.