Cyclin-dependent kinase 2 (CDK2) has been proposed to function as a master regulator of centrosome duplication. Using mouse embryonic fibroblasts (MEFs) in which Cdk2 has been genetically deleted, we show here that CDK2 is not required for normal centrosome duplication, maturation and bipolar mitotic spindle formation. In contrast, Cdk2 deficiency completely abrogates aberrant centrosome duplication induced by a viral oncogene. Mechanistically, centrosome overduplication in MEFs wild-type for Cdk2 involves the formation of supernumerary immature centrosomes. These results indicate that normal and abnormal centrosome duplication have significantly different requirements for CDK2 activity and point to a role of CDK2 in licensing centrosomes for aberrant duplication. Furthermore, our findings suggest that CDK2 may be a suitable therapeutic target to inhibit centrosome-mediated chromosomal instability in tumor cells.
Centrosomes serve as major microtubule-organizing centers in animal and human cells (Bornens, 2002). The single centrosome of a cell duplicates precisely once before mitosis in synchrony with the cell division cycle (Hinchcliffe and Sluder, 2001; Delattre and Gonczy, 2004). Abnormal centrosome numbers are detected in virtually all human cancers where they can contribute to abnormal mitotic spindle polarity and chromosomal instability (Salisbury et al., 1999; Brinkley, 2001; Doxsey, 2002; Nigg, 2002; Duensing, 2005). The molecular mechanisms that govern normal and aberrant centrosome duplication are incompletely understood. Cyclin-dependent kinase 2 (CDK2) has been reported to control centrosome duplication in various model systems in which repeated centrosome reproduction was induced by a prolonged S phase arrest (Hinchcliffe et al., 1999; Lacey et al., 1999; Matsumoto et al., 1999; Meraldi et al., 1999). Several targets downstream of CDK2 relevant for the control of centrosome duplication have been identified including nucleophosmin/B23 (Okuda et al., 2000), Mps1 (Fisk and Winey, 2001) and CP110 (Chen et al., 2002). However, Cdk2-null mice are viable and develop essentially normal (Berthet et al., 2003; Ortega et al., 2003). Cells derived from these animals have no apparent proliferation defect in culture and re-enter the cell division cycle after serum deprivation without significant delay (Ortega et al., 2003). In addition, cancer cells were found to proliferate despite CDK2 inhibition (Tetsu and McCormick, 2003). Based on these results, we designed the present study to reassess the role of CDK2 in normal and aberrant centrosome duplication.
To determine the role of CDK2 in the regulation of centrosome duplication, we used mouse embryonic fibroblasts (MEFs) in which Cdk2 has been genetically deleted. Primary MEFs derived from Cdk2−/− mice (Ortega et al., 2003) did not have any detectable CDK2 protein expression (data not shown). MEFs were analysed for centrosome numbers using immunofluorescence microscopy for γ-tubulin, a commonly used marker for centrosomes. The γ-tubulin staining pattern in Cdk2−/− MEFs was normal (Figure 1a) and indistinguishable from cells wild-type for Cdk2. Quantification of centrosome numbers per cell revealed that 14.1% of Cdk2−/− MEFs had one centrosome compared to 37.3% in Cdk2+/+ cells (Figure 1b). In Cdk2−/− MEFs, 70.1% of cells showed duplicated centrosomes (i.e., two per cell) compared to 54.9% in Cdk2+/+ cells. The percentage of cells with three, four or more than four centrosomes was 4.7, 7.7 and 1.9% in Cdk2−/− cells compared to 3.4, 3.8 and 0.6% in Cdk2+/+ MEFs. The increased fraction of Cdk2−/− MEFs with two centrosomes and the simultaneous decrease of cells with one centrosome is likely related to the fact that Cdk2-deficient cells have the tendency to enter culture crisis earlier than Cdk2 wild-type cells (Ortega et al., 2003). Senescent cells have been reported to undergo growth arrest in a tetraploid state (Mason et al., 2004) and may therefore contain two centrosomes. The mitotic index, however, was not found to differ significantly between Cdk2−/− MEFs and wild-type MEFs (2.6% in both populations). Importantly, more than 90% of mitoses showed a normal bipolar spindle pole arrangement in both Cdk2−/− and Cdk2+/+ MEFs indicating that centrosomes functioned normally with respect to their role in spindle pole formation (Figure 1a).
Besides duplication, centrosomes undergo a process of maturation in order to become fully functional. Rodent fibroblasts grow a primary cilium specifically from the older, mature centriole of a centrosome under conditions of serum deprivation. A single, unduplicated centrosome therefore usually carries one primary cilium. Two primary cilia can be found at later stages of the G2 phase of the cell division cycle, when two centrosomes are present and each centrosome contains a centriole that has reached a certain level of maturation (Wheatley et al., 1996; Albrecht-Buehler and Bushnell, 1980; Alvarez-Salas et al., 1998). The analysis of primary cilium formation hence allows to evaluate the maturation status of centrosomes that are present in a cell (Guarguaglini et al., 2005). Primary cilia can be detected by immunostaining for acetylated α-tubulin. To analyse primary cilia in Cdk2−/− and Cdk2+/+ MEFs, cells were serum-deprived for 24 h to stimulate cilium formation followed by coimmunofluorescence analysis for γ-tubulin and acetylated α-tubulin (Figure 1c). A single centrosome associated with one primary cilium was detected in 46.2% of Cdk2−/− and 40% of Cdk2+/+ MEFs, respectively (Figure 1d). Duplicated centrosomes with one cilium-carrying centrosome were detected in 60.3% of Cdk2−/− cells and 65.2% of Cdk2+/+ MEFs, whereas two centrosomes each carrying a primary cilium were found in 3.9% of Cdk2−/− and 4.3% of Cdk2+/+ cells, respectively. Taken together, these results show that normal centrosome duplication and maturation are not affected by Cdk2 deficiency.
We next explored the possibility that ablation of Cdk2 may affect aberrant centrosome duplication. To stimulate centrosome overduplication, we used the E7 oncoprotein of human papillomavirus type 16 (HPV-16). HPV-16 E7 disrupts the retinoblastoma protein (pRB) tumor-suppressor pathway through binding and degradation of pRB and the related p107 and p130 proteins as well as inactivation of the CDK inhibitors p21Cip1 and p27Kip1 (Munger and Howley, 2002). HPV-16 E7 rapidly stimulates aberrant centriole duplication and therefore predisposes cells to centrosome-related mitotic defects and chromosomal instability (Duensing et al., 2000). It has recently been shown that the HPV-16 E7 oncoprotein stimulates centriole overduplication through the increased formation of immature daughter centrioles (Guarguaglini et al., 2005).
Cdk2-deficient and Cdk2 wild-type MEFs transiently transfected with HPV-16 E7 did not differ significantly in the expression level of the E7 oncoprotein (Figure 2a). To determine whether expression of HPV-16 E7 increases cyclin E- or cyclin A-associated kinase activities, an in vitro kinase assay was performed using a recombinant C-terminal fragment of pRB as a substrate. A modest increase of mainly cyclin E-associated kinase activity was detected in HPV-16 E7-transfected Cdk2+/+ cell populations but not in Cdk2-deficient MEFs (Figure 2b). The relatively small changes in kinase activity are most likely related to the fact that experiments were performed after transient transfection of MEF populations that yielded transfection rates of approximately 5–10%.
We next determined numerical centrosome aberrations in cells from both genotypes following transient transfection with HPV-16 E7 or empty control vector. This analysis was performed on a cell-per-cell basis using DsRED as marker for transfected cells (Figure 2c). Fold-changes of the percentage of cells with abnormal centrosome numbers (i.e., more than two per cell) were calculated (Figure 2d). HPV-16 E7-transfected Cdk2−/− MEFs showed no increase of abnormal centrosome numbers (0.9-fold), whereas Cdk2 wild-type MEFs showed a significant 1.6-fold increase of the proportion of cells with more than two centrosomes after transfection with HPV-16 E7 (P<0.05, Student's t test for independent samples; Figure 2d). Cotransfection of Cdk2−/− MEFs with HPV-16 E7 and a CDK2 expression plasmid compensated for Cdk2 deficiency and resulted in a significant 1.4-fold increase of the proportion of cells with numerical centrosome anomalies compared to controls (P<0.05). This increase was similar to the increase detected in Cdk2+/+ MEFs transfected with HPV-16 E7 either alone or HPV-16 E7 in combination with CDK2 (1.7-fold; P<0.05; Figure 2d). Transient transfection of MEFs with CDK2 alone did not stimulate numerical centrosome aberrations in both Cdk2−/− or Cdk2+/+ cells (data not shown).
To determine whether cyclin E or cyclin A can rescue the HPV-16 E7-induced centrosome phenotype as CDK2 does, we transiently overexpressed cyclin E or cyclin A in Cdk2−/− and Cdk2+/+ MEFs either alone or in combination with HPV-16 E7 and analysed centrosome numbers in transfected cells (Figure 2e). Expression of cyclins was confirmed by immunoblotting (data not shown). Overexpression of cyclin E or cyclin A alone as well as coexpression of cyclin E or cyclin A with HPV-16 E7 did not lead to an increase of the proportion of cells with abnormal centrosome numbers in Cdk2−/− MEFs (Figure 2e). A 1.3-fold increase detected in Cdk2−/− cells transfected with HPV-16 E7 and cyclin A did not reach statistical significance. In Cdk2 wild-type MEFs, cyclin A overexpression alone caused a statistically significant 1.6-fold increase of cells with abnormal centrosome numbers (P<0.001) and coexpression of cyclin A and HPV-16 E7 led to a 1.5-fold increase of cells with abnormal centrosome numbers (P<0.005). Overexpression of cyclin E alone or in combination with HPV-16 E7 in Cdk2 wild-type MEFs stimulated a 1.6-fold and 1.5-fold increase of cells with abnormal centrosome numbers, respectively, that was in the range of cyclin A but did not reach statistical significance. Coexpression of cyclin A or cyclin E with HPV-16 E7 did not cause more centrosome abnormalities than overexpression of HPV-16 E7 alone (2.05-fold; P<0.05). Collectively, these results show that overexpression of cyclin A or cyclin E cannot compensate for CDK2 loss with respect to HPV-16 E7-induced aberrant centrosome duplication.
To rule out the possibility that Cdk2−/− MEFs have adapted to the loss of CDK2 expression by complementing mutations, small-interfering RNA (siRNA) experiments were performed using Cdk2 wild-type MEFs and mouse-specific siRNA duplexes targeting CDK2 (Figure 3). Depletion of CDK2 protein (Figure 3a) was found to abolish the ability of HPV-16 E7 to induce abnormal centrosome numbers (Figure 3b) without stimulating an increased cell death or growth arrest. These findings further support the notion that CDK2 is required for the induction of abnormal centrosome duplication.
We next sought to determine whether aberrant centrosome numbers induced by HPV-16 E7 in Cdk2+/+ MEFs was caused by an increased generation of immature centrosomes. Cdk2+/+ MEFs transiently transfected with HPV-16 E7 were serum-deprived in order to stimulate the formation of primary cilia (Figure 4). In a representative experiment, Cdk2+/+ MEFs with abnormal centrosome numbers after transient transfection with the HPV-16 E7 oncoprotein contained an average of 3.3 centrosomes in the presence of an average of 1.0 cilium-carrying centrosome per cell. In Cdk2+/+ MEFs without centrosome overduplication (one or two centrosomes per cell), an average of 0.8 cilium-carrying centrosomes was detected. These results underscore that HPV-16 E7 induces a bona fide centrosome overduplication in Cdk2+/+ MEFs as opposed to an accumulation of centrosomes, which would result in an increase of mature centrioles carrying primary cilia.
Collectively, our findings provide the first definitive demonstration of CDK2-independent centrosome duplication in somatic mammalian cells. Although CDK2 is not required for normal centrosome duplication, we demonstrate that CDK2 is a mediator of centrosome overduplication induced by the HPV-16 E7 oncoprotein.
In previous studies, small molecule CDK2 inhibitors were successfully used to interfere with centrosome duplication and it is important to mention that one inhibitor (Purvalanol A) was found to block normal centrosome duplication (Wong and Stearns, 2003), whereas another CDK inhibitor (indirubin-3′-oxime) inhibited centrosome overduplication while leaving normal duplication unaffected (Duensing et al., 2004b). CDK inhibitors can differ significantly in both the spectrum of inhibited protein kinases and the inhibitory activity against individual kinases (Meijer et al., 1999). As our results suggest that the activity of other kinases including cyclin A-associated CDK activity may compensate for the loss of CDK2 to allow normal centrosome duplication (Figure 1), it is likely that differences in the activity spectrum of the small molecule inhibitors account for the observed discrepancies.
Cyclin A has previously been reported to be more effective than cyclin E in promoting aberrant centrosome duplication (Meraldi et al., 1999). We find a similar trend, although it is notable that cyclin E can induce abnormal centrosome numbers in Cdk2 wild-type cells to a similar extent as cyclin A, albeit without reaching statistical significance (Figure 2e). Overexpression of cyclin E has been reported to stimulate normal centrosome duplication and aberrant centrosome numbers in p53-deficient cells (Mussman et al., 2000). We have previously shown that MEFs deficient of cyclins E1 and E2 maintain normal centrosome duplication (Duensing et al., 2004a) suggesting that CDK2-independent functions of cyclin E (Matsumoto and Maller, 2004; Aleem et al., 2005) may play a role in the observed effects on the normal centrosome duplication cycle.
Centrosomes do not have unlimited capacity to reproduce during a normal duplication cycle. In fact, duplicated centrosomes are endowed with an intrinsic block to reduplicate even in a cellular milieu that normally supports their reproduction (Wong and Stearns, 2003). Therefore, our finding that CDK2 is necessary for centrosome overduplication may point to a role of CDK2 in licensing centrosomes for aberrant duplication. This is supported by the presence of multiple immature centrosomes in cells with HPV-16 E7-induced centrosome amplification (Figure 4). Our study does not disagree with previous reports showing CDK2 to promote multiple rounds of centrosome duplication upon S phase arrest (Hinchcliffe et al., 1999; Lacey et al., 1999; Matsumoto et al., 1999; Meraldi et al., 1999) as it is conceivable that under such conditions the intrinsic block to reduplicate also needs to be relaxed. Whether and to what extent deregulated CDK2 activity drives centrosome overduplication and chromosome missegregation in human cancers remains to be determined. If such a function can be substantiated, our results suggest that CDK2 would be a suitable drug target to inhibit centrosome-mediated chromosomal instability in cancer patients.
We are grateful to K Munger, P Hinds, Y Chang and P Moore for sharing reagents. This work was supported by NIH/NCI Grant CA112598 (to SD).
About this article
Cell Division (2011)