|Home | About | Journals | Submit | Contact Us | Français|
Centrosome amplification (CA) contributes to carcinogenesis by generating aneuploidy. Elevated frequencies of CA in most benign breast lesions and primary tumors suggest a causative role for CA in breast cancers. Clearly, identifying which and how altered signal transduction pathways contribute to CA is crucial to breast cancer control. Although a causative and cooperative role for c-Myc and Ras in mammary tumorigenesis is well documented, their ability to generate CA during mammary tumor initiation remains unexplored. To answer that question, K-RasG12D and c-Myc were induced in mouse mammary glands. Although CA was observed in mammary tumors initiated by c-Myc or K-RasG12D, it was detected only in premalignant mammary lesions expressing K-RasG12D. CA, both in vivo and in vitro, was associated with increased expression of the centrosome-regulatory proteins, cyclin D1 and Nek2. Abolishing the expression of cyclin D1, Cdk4 or Nek2 in MCF10A human mammary epithelial cells expressing H-RasG12V abrogated Ras-induced CA, whereas silencing cyclin E1 or B2 had no effect. Thus, we conclude that CA precedes mammary tumorigenesis, and interfering with centrosome-regulatory targets suppresses CA.
Overexpression of Ras and Myc proto-oncogenes in breast cancers is associated with poor prognosis (Berns et al., 1992; Shackney et al., 2004). Ras is constitutively active in breast cancers through deregulated Her2, Erb4 and Egfr tyrosine kinase receptors (Janes et al., 1994; Hoadley et al., 2007), overexpression of H-Ras, K-Ras and N-Ras in 69% breast cancers or mutational activation of K-Ras in 6.5% breast cancers (Clair et al., 1987; Miyakis et al., 1998). Amplification of c-Myc results in its overexpression in 15–70% breast cancers (Chrzan et al., 2001; Rummukainen et al., 2001; Blancato et al., 2004). Direct evidence that Ras and c-Myc are involved in mammary cancers was obtained by the coexpression of c-Myc, H-RasG12V, telomerase and SV40 T antigens in primary human mammary epithelial cells, resulting in transformation (Elenbaas et al., 2001). In addition, MMTV (mouse mammary tumor virus) transgenic mice expressing c-Myc (Sinn et al., 1987; D'Cruz et al., 2001), H-RasG12V (Sinn et al., 1987; Sarkisian et al., 2007), N-Ras (Mangues et al., 1990) and K-RasG12D (Omer et al., 2000; Podsypanina et al., 2008) developed mammary tumors.
Concurrently, apoptosis and cell-cycle arrest are transient barriers to Myc and Ras mammary carcinogenesis (D'Cruz et al., 2001; Sarkisian et al., 2007). For example, overexpression of H-RasG12V in mouse mammary epithelial cells results in transient cell-cycle arrest between 14 and 32 days after induction (Sarkisian et al., 2007). Indeed, an active p53 pathway is a major obstacle to H-RasG12V-initiated mammary tumors (Hundley et al., 1997b; Sarkisian et al., 2007), and c-Myc triggers activating K-Ras mutations to induce nonregressing mammary tumors (D'Cruz et al., 2001). Chromosome instability (CIN) is a potential mechanism used by oncogenes to abrogate transient barriers to mammary cancers; consistent with this, mammary tumors expressing H-RasG12V and c-Myc are genomically unstable (Hundley et al.,1997b; Weaver et al.,1999). However, the source, timing and relevance of oncogene-dependent CIN to mammary tumorigenesis are unknown.
Centrosome amplification (CA), the acquisition of three or more centrosomes within a cell, is one of the major contributors to CIN in human cancers (Doxsey, 2002). CA is frequently observed in human cancers—including prostate, colon, breast and cervical cancer—which suggests an involvement in tumorigenesis (Pihan et al., 1998; Carroll et al., 1999; Ghadimi et al., 2000; Lingle et al., 2002). CA is a potential initiator of mammary tumorigenesis, as a significant fraction of benign breast lesions (Berman et al., 2005; Guo et al., 2007) and most breast cancers display CA (Lingle et al., 2002; Schneeweiss et al., 2003; Guo et al., 2007). Centrosomes ensure equal segregation of chromosomes by directing the bipolarity of the mitotic spindle (Fukasawa, 2005). Thus, CA generates multipolar spindles, merotelic attachments (attachment of single kinetochores to microtubules emanating from different poles), chromosomal lagging and aneuploidy, a major type of CIN (Ganem et al., 2009). As CA and multipolar mitoses are potentially transforming, they are suppressed by various mechanisms, including mitotic catastrophe, centrosomal clustering during mitosis, and genomic convergence (Oikawa et al., 2005; Quintyne et al., 2005; Ganem et al., 2009).
The most direct evidence showing the involvement of CA in tumorigenesis is that ectopic expression of centrosome-regulatory proteins in transplanted Drosophila neuronal stem cells resulted in tumors (Basto et al., 2008; Castellanos et al., 2008). In mammalian cancers, aneuploidy is ubiquitous (Duesberg and Rasnick, 2000). In contrast to mammalian cells, in which low-level aneuploidy initiates and sustains various mouse tumors (Weaver and Cleveland, 2007; Schliekelman et al., 2009), and in contrast to mammalian tumors, which are aneuploid, tumorigenesis in Drosophila was not accompanied by CIN. In fact, of the five independent genetic alterations required to transform primary human mammary epithelial cells (Elenbaas et al., 2001), three (namely H-RasG12V, inactive Rb and p53) trigger CA and CIN (Fukasawa et al., 1996; Saavedra et al., 1999; Iovino et al., 2006), whereas c-Myc triggers aneuploidy and chromosome recombinations (Weaver et al., 1999); this suggests a close relationship between CA, CIN and mammary tumor initiation.
Showing that CA is involved in mammary tumor initiation requires establishing that oncogene-driven CA occurs during premalignancy and identifying single or cooperating oncogenes responsible for CA. The identification of centrosome-regulatory proteins deregulated by oncogenes would allow future therapeutic interventions to abrogate CA and aneuploidy that drive breast tumors. We show the ability of Ras to signal CA in premalignant mouse mammary lesions and human mammary epithelial cells through cyclin D1/Cdk4 and Nek2.
Doxycycline-inducible MMTV transgenic mice expressing K-RasG12D and/or c-Myc for 5 days, or until mammary tumors developed, were used to address various abnormal phenotypes involved in mammary tumor initiation (D'Cruz et al., 2001). Real-time PCR using transgene-specific primers (Supplementary Table S1) showed that K-RasG12D and/or c-Myc expressed robustly in the corresponding transgenic groups, which was undetectable in controls (Supplementary Figures S1a and b). Western blots detecting endogenous and transgenic Ras and c-Myc (Supplementary Figures S1c and d) showed that levels of K-RasG12D (sevenfold over the control) are within the average Ras expression in human breast tumors, which are 2–10-fold relative to the nonaffected mammary epithelium (Clair et al., 1987; Miyakis et al., 1998). In contrast, c-Myc levels are much higher (50–70-fold over controls) than the average c-Myc levels in human breast tumors, which are 1.8–4-fold relative to the nonaffected mammary epithelium (Chrzan et al., 2001; Rummukainen et al., 2001; Blancato et al., 2004).
As reported previously for H-RasG12V and K-RasG12D (Sinn et al., 1987; Podsypanina et al., 2008), mammary tumors initiated by K-RasG12D occurred much faster relative to c-Myc (Supplementary Figure S1e), and coexpression of K-RasG12D and c-Myc induced mammary tumors faster than either transgene did separately. These results allowed us to select a time point of 5 days to investigate events associated with early premalignancy, as it precedes tumorigenesis by a few weeks.
We assessed various abnormal phenotypes associated with the expression of K-RasG12D and c-Myc that have been thoroughly studied in tumors, but are poorly understood in early premalignancy; these include histological changes, ectopic proliferation and apoptosis (Hundley et al., 1997a; D'Cruz et al., 2001; Bearss et al., 2002; Blakely et al., 2005; Podsypanina et al., 2008). Mammary glands expressing oncogenes displayed distinct histopathological changes at premalignancy: c-Myc led to mild hyperplasia of ducts and lobules, with single-layered acini adjacent to each other. In contrast, K-RasG12D, or K-RasG12D and c-Myc severely altered the normal structure of the mammary gland; specifically, ducts and lobules were hyperplastic, epithelial cells occupied the lumen of the acini, and had invaded into the stroma (Figure 1a). Such distinctions were obscured in tumors caused by K-RasG12D or c-Myc, as both harbored numerous malignant epithelial cells and scanty stroma (Figure 1a).
Ki-67 immunostaining showed that all oncogenes enhanced proliferation of mammary epithelial cells, and that K-RasG12D and c-Myc cooperated to increase those frequencies in premalignant lesions (Figures 1a and b). Cleaved caspase-3 showed that although K-RasG12D and c-Myc increased cellular apoptosis during premalignancy, only c-Myc signaled apoptosis in tumors. K-RasG12D suppressed c-Myc-signaled apoptosis at both stages (Figures 1a and c).
Thus, K-RasG12D and c-Myc are triggering malignant phenotypes during premalignancy, and their synergistic nature is obvious as soon as premalignancy because they cooperate to modulate frequencies of proliferation and apoptosis, as well as to accelerate mammary tumor formation.
Frequencies of CA were assessed in mammary premalignant lesions and tumors initiated by K-RasG12D and/or c-Myc using immunostainings against γ-tubulin and pericentrin, proteins within the pericentriolar material of centrosomes essential to the nucleation of microtubules (Figure 2a). In spite of the universal CA found in tumors (Figure 2c), only mammary glands expressing K-RasG12D or K-RasG12D and c-Myc displayed elevated frequencies of CA at premalignancy (Figure 2b). However, K-RasG12D and c-Myc did not significantly increase frequencies of CA compared with K-RasG12D alone. Thus, CA occurs during tumor initiation and is specific to the K-RasG12D pathway.
One of the major mechanisms generating CA is the deregulation of the centrosome duplication cycle in the late G1/S phase. Deregulation may arise as a consequence of the downregulation of negative regulators of cell and centrosome cycles, including p53 (Fukasawa et al., 1996), NPM (Grisendi et al., 2005), p21Waf–1 (Duensing et al., 2006), p16INK4A (Berman et al., 2005; McDermott et al., 2006), Brca1 (Xu et al., 1999), Brca2 (Tutt et al., 1999) and E2F3 (Saavedra et al., 2003). Various checkpoint controls activated in response to overexpressed H-RasG12V are also involved in the negative regulation of the centrosome cycle; for example, protein expression of p21Waf–1, p16INK4A and p19ARF plateau at 8 days after induction, whereas p53 is activated at 4 days after induction (Sarkisian et al., 2007). A second major mechanism leading to the deregulation of the centrosome cycle is the overexpression of cell and centrosome-regulatory molecules; these include E2F2 and E2F3 (Meraldi et al., 1999), cyclin D1 (Nelsen et al., 2005), cyclins E and A in p53-null cells (Hanashiro et al., 2008), Plk4 (Kleylein-Sohn et al., 2007), Mps-1 (Fisk and Winey, 2001) or Nek-2 (Hayward et al., 2004).
Taking these mechanisms into account, we screened the steady-state transcriptional levels of various molecules involved in cell and centrosome cycles using quantitative real-time PCR (Table 1). Supplementary Table S1 lists the primer sequences used to amplify differentially regulated genes. None of the CKIs were significantly downregulated; rather, we observed significant overexpression of some of those transcripts, including p16INK4A, p19ARF and p27Kip1. Similarly, Nek2, E2F2, E2F3a, cyclin D1, cyclin B2 and Plk4 were upregulated.
We selected a subset of differentially expressed gene products from Table 1 and assessed their steady-state protein levels with western blots (Figure 3). In general, K-RasG12D and c-Myc led to a more robust deregulation of most target genes relative to either single oncogene. The results from real-time PCR were not always consistent with western blots, perhaps because real-time PCR can detect minuscule amounts of mRNAs. For example, western blots did not detect upregulated p19ARF or cyclin B2. In addition, even though p16INK4A mRNA was upregulated by K-RasG12D, the p16INK4A protein was only robustly upregulated in mammary glands coexpressing K-RasG12D and c-Myc. In other instances, upregulated mRNA corresponded to upregulated proteins; for example, p27Kip1 was upregulated by all oncogenes, and p21Waf–1 was upregulated by K-RasG12D and c-Myc. Another important checkpoint, p53, was hyperphosphorylated in mammary glands expressing K-RasG12D, or K-RasG12D and c-Myc.
Various gene products associated with CA were upregulated; for example, Nek2 was equally upregulated by all combinations of oncogenes, and cyclin E1 was only upregulated by K-RasG12D and c-Myc. More importantly, cyclin D1 was upregulated at the mRNA and protein levels in mammary epithelial cells expressing K-RasG12D, or K-RasG12D and c-Myc. Consistent with upregulated cyclins D1 or E1 was the increased phosphorylation of Rb in mammary glands expressing K-RasG12D, or K-RasG12D and c-Myc.
Taken together, the data indicated that rather than downmodulating CKIs or p53, K-RasG12D or K-RasG12D and c-Myc-dependent CA may arise from their ability to upregulate targets that are critical in regulating both the centrosome cycle (such as Nek2) and the cell cycle (such as cyclins D1 and E1).
We have described the ability of K-RasG12D, either alone or coexpressed with c-Myc, to deregulate various key regulators of cell and centrosome duplication cycles. Of those, Nek2 is a centrosome separase normally active at mitosis (Fry et al., 1998). Nek2 is overexpressed in breast cancers and exhibits centriole-splitting activity when expressed in interphase (Hayward et al., 2004). Deregulated cyclin E/Cdk2 signals CA (Tarapore et al., 2001; Saavedra et al., 2003; Hanashiro et al., 2008). Similarly, cyclinD1/Cdk4, is associated with CA (Nelsen et al., 2005) and is a key regulator of centrosome duplication (Adon et al., 2010). Cyclin B is localized in the centrosome (Bailly et al., 1992). Thus, it is reasonable to assume that the upregulation of Nek2, cyclin D1 cyclin E1 or B2 may mediate Ras and Ras- and c-Myc-dependent CA. Hence, MCF10A cell lines stably expressing H-RasG12V, or H-RasG12V and c-Myc were generated. MCF10A is a nontransformed human mammary epithelial cell line with intact p53 (Neve et al., 2006). The MCF10A system showed minor differences relative to the transgenic mice; for example, although H-RasG12V, or H-RasG12V and c-Myc caused upregulation of Nek2 and cyclin D1 as observed in vivo, the expression of cyclins E1 or B2 was unchanged, but nevertheless highly expressed (Figure 4a). As ectopic expression of H-RasG12V, or H-RasG12V and c-Myc results in CA (Figure 4c), Nek2, cyclinD1, Cdk4, cyclin E1 and cyclin B2 were silenced (Figure 4b) using small interfering RNAs presented in Supplementary Table S2. Silencing Cdk4, cyclin D1 or Nek2 abrogated oncogene-triggered CA (Figure 4b). These data showed that Nek2 and cyclin D1/Cdk4 are critical to oncogene-triggered CA.
An explanation for the ability of silenced cyclin D1, Cdk4 and Nek2 to suppress CA is that their down-regulation causes cell-cycle arrest. Cell-cycle analysis by flow cytometry (Table 2) showed that silencing cyclin E1, cyclin B2 and Nek2 in MCF10A cells expressing H-RasG12V significantly increased cell population in the G1 phase and correspondingly decreased cell population in the G2/M phases, whereas inhibition of Cdk4 or cyclin D1 did not. In MCF10A cells expressing vector control, or coexpressing H-RasG12V and c-Myc, silencing Nek2, cyclins D1, B2 or E1 significantly elevated cells accumulating in the G1 phase. 5-Bromo-2-deoxyuridine incorporation assays (Figure 4d) showed that knockdowns of cyclin D1 and Nek2 significantly inhibited cell proliferation induced by ectopically expressing H-RasG12V, or H-RasG12V and c-Myc. Thus, deregulation of the cell cycle is not the only cause of CA.
This study addresses some important questions regarding the relationship between CA and mammary tumorigenesis. First, we show that CA precedes tumorigenesis and is oncogene specific, as K-RasG12D initiated CA in mammary precursor lesions. In contrast, c-Myc was unable to induce CA in early premalignancy, but induced CA in tumors. These findings place K-RasG12D among a group of oncogenes, including Aurora A and Pin-1, causing CA in premalignant mammary lesions (Suizu et al., 2006; Wang et al., 2006). In contrast, c-Myc falls in the category of genetic alterations, including ablated p53, that does not display CA during premalignancy (Goepfert et al., 2000; Fleisch et al., 2006). As some oncogenic stimuli lead to CA at premalignancy, whereas others do so later, suggests that some oncogenes directly trigger CA to rapidly initiate tumors, whereas others require additional genetic or epigenetic changes to induce CA. The capacity of K-RasG12D to induce CA and lower apoptotic frequencies may contribute to faster times to tumors relative to c-Myc, as both oncogenes are similarly efficient in triggering ectopic proliferation in premalignant lesions and tumors. Another major finding was the identification of a subset of K-RasG12D-specific centrosome-regulatory targets mediating CA in premalignant mammary lesions, including Nek2 and cyclin D1/Cdk4, and that their silencing abrogated CA in human mammary epithelial cells. Nevertheless, the question remains as to whether they mediate CA in vivo.
Interestingly, c-Myc resulted in Nek2 upregulation without causing CA. An explanation for this is that Nek2 is necessary, but not sufficient to trigger CA without cooperating with other altered centrosome-regulatory molecules. There is precedent for the cooperative nature of Nek2: For example, the ectopic expression of Nek2 cannot induce CA unless mammary epithelial cells are preimmortalized with the SV40 T-antigen (Hayward et al., 2004). In addition, we showed that c-Myc enhanced proliferation without causing CA during premalignancy, showing that CA and ectopic proliferation arise independently. In contrast, c-Myc mammary tumors harbored CA. This suggests that c-Myc cooperates with secondary alterations to cause CA; one of the alterations might be K-Ras, as it is a common hotspot in nonregressing mouse mammary tumors initiated by c-Myc (D'Cruz et al., 2001).
Evidence suggests that ablated E2F3 and p53 upregulate Cdk2 to trigger CA (Fukasawa et al., 1996; Saavedra et al., 2003). In fact, our recent work showed that Cdk2 and Cdk4 are key mediators of CA in p53-null cells (Adon et al., 2010). In contrast, the ectopic expression of cyclin E (the catalytic partner of Cdk2) in hepatocytes only results in mild CA, whereas overexpression of cyclin D1 (the catalytic partner of Cdk4) results in more severe CA (Nelsen et al., 2005), suggesting that in some cell/tissue types, cyclin D1/Cdk4 is more potent than cyclin E/Cdk2 in signaling CA. This is also evidenced by our observations that silencing cyclin D1/Cdk4 significantly inhibits H-RasG12V, or H-RasG12V and c-Myc-dependent CA. In contrast, inhibition of cyclin E1 or cyclin B2 severely alters cell-cycle profiles without affecting CA.
Although our studies clearly showed that Cdk4 is involved in Ras-induced CA, it is unknown how it leads to CA. One explanation is that Cdk4 phosphorylates targets required for regulating the centrosome cycle. For example, there is a strong correlation between hyperactive Cdk2, hyperphosphorylation and inactivation of NPM—a major negative regulator of the centrosome cycle (Saavedra et al., 2003). Our recent work showed that NPM is phosphorylated by Cdk4 during the G1 phase, and that expressing NPMT199A, a mutant lacking the Cdk2/Cdk4 phosphorylation site prevented CA in p53-null cells (Adon et al., 2010). Similarly, deregulated Cdk4 may also use canonical Cdk2 phosphorylation sites in molecules involved in other steps in the centrosome cycle, including CP110, Mps-1 and Plk4—regulators of centriole duplication that are direct Cdk2 targets or that require Cdk2 for their optimal activity (Chen et al., 2002; Fisk et al., 2003; Habedanck et al., 2005; Kleylein-Sohn et al., 2007). Another explanation is that hyperactive cyclin D/Cdk4 hyperphosphorylates Rb, leading to increased E2F activity and the deregulation of centrosome-regulatory targets. We will identify transcriptional and posttranscriptional targets of cyclin D1/Cdk4 in the future.
As ablation of cyclin D1 or Cdk4 abrogates mammary tumorigenesis in MMTV-Ras or MMTV-Neu (Her2) mice (Yu et al., 2001; Reddy et al., 2005), Cdk4/Cdk6-specific inhibitors reduce ectopic proliferation in human Her2+ breast cancer cells (Finn et al., 2009), and inhibitors of Nek2 decrease the tumorigenic potential of breast cancer cells (Wu et al., 2008; Tsunoda et al., 2009), we propose that centrosomal-regulatory targets downstream of Ras would represent important future targets for intervening with breast tumorigenesis.
We thank Drs Rene Opavsky, Paul W Doetsch, Ya Wang and Hui Wang for manuscript discussions. We also thank Ms Carla G Saavedra and Meredith Roberts for editing; Dr Harold Varmus for providing tetO-K-RasG12D mice; Dr J Brugge for nontransformed MCF10A cells; and Jana Opavska, Joi Carmichael and Stacy Sannem for technical assistance. We thank Dr Adam Marcus (from the Emory Imaging Core) and Mr Alan Bakaletz, for imaging advice. Lewis A Chodosh was funded by NIH R01CA98371, DOD BCRP W81XWH-05-1-0405 and NIH U01 CA105490, Gustavo Leone by R01CA85619, R01HD042619, R01CA121275, R01HD047470 and P01CA097189, Harold Saavedra by K01CA104079, and a Georgia Cancer Coalition Distinguished Scholar Award.
Conflict of interest
The authors declare no conflict of interest.