Oncogenes induce senescence with incomplete growth arrest and suppress the DNA damage response in immortalized cells

Summary

Activation of the Her2 (ErbB2) oncogene is implicated in the development of breast, ovary and other cancers. Here, we show that expression of NeuT, a mutant-activated rodent isoform of Her2, in immortalized breast epithelial cells, while promoting senescence associated morphological changes, up-regulation of senescence associated β-galactosidase activity, and accumulation of the cyclin-dependent kinase inhibitor p21, failed to trigger the major senescence end-point, i.e. permanent growth arrest. Similar senescence-associated phenotype with incomplete growth arrest, which we dubbed senescence with incomplezXte growth arrest (SWING), could also be triggered by the expression of the Ras oncogene. SWING phenotype was stable, and persisted in tumor xenografts established from NeuT-transduced cells. Furthermore, a significant population of cells in SWING state was found in tumors in the MMTV/NeuT transgenic mouse model. SWING cells showed downregulation of histone H2AX, critical for repair of double-stranded DNA breaks, and impaired activation of Chk1 kinase. Overall, SWING cells were characterized by increased DNA instability and hypersensitivity to genotoxic stresses. We propose that the SWING state could be a stage in the process of cancer development.

Introduction

Normal human somatic cells have a limited proliferation lifespan and undergo replicative senescence characterized by irreversible growth arrest after a certain number of cell divisions (Hayflick & Moorhead, 1961). This senescence response was shown to be driven by telomere shortening that triggered a DNA damage response (DDR), culminating in irreversible cell cycle arrest (Harley et al., 1990; Bodnar et al., 1998). Alternatively, telomere length-independent senescence could be precipitated in cells that have not yet reached their proliferative limit by exposure to various DNA damaging stresses (Schmitt et al., 2002; Campisi & d’Adda di Fagagna, 2007). For example, oncogene imbalance that can induce DNA damage is a potent inducer of a senescence response (McDuff & Turner, 2010). In fact, various oncogenes such as KRAS, BRAF, Mek, or Mos were shown to stimulate senescence both in fibroblasts and in transformed cell lines (Serrano et al., 1997; Lin et al., 1998; Zhu et al., 1998; Bartkova et al., 2006). Moreover, activation of E2F, Ras, BRaf, or loss of PTEN induced senescence in mouse models (Braig et al., 2005; Chen et al., 2005; Collado et al., 2005; Lazzerini Denchi et al., 2005; Michaloglou et al., 2005). Altogether these observations led to a hypothesis that oncogene-induced senescence (OIS) represents a major barrier toneoplastic transformation (Yaswen & Campisi, 2007). In line with this idea, early preneoplastic lesions like human nevi which carry activating BRafV600E mutation harbor senescent melanocytes and rarely progress to melanomas (Pollock et al., 2003).

On the other hand, once cell have been immortalized with reactivation of endogenous telomerase activity, their response to oncogenes changes. For example, expression of the Raf-1 or Ras oncogene in immortalized human mammary epithelial cells (HMEC) not only fails to produce growth arrest, but confers malignancy associated properties (Olsen et al., 2002; Cipriano et al. 2011). In vivo, reports show that cells with high level of activated Ras that normally triggers senescence eventually form aggressive tumors (Sarkisian et al., 2007). Furthermore, hyperplastic colorectal polyps with endogenous activated KRAS or BRAF despite showing major signs of senescence, e.g. senescence-associated-β-galactosidase (SA-β-gal) staining, continue to grow (Minoo & Jass, 2006). Similarly, early hyperplastic lesions in the PTEN knockout model of the prostate cancer show extensive SA-β-gal staining. These lesions give rise to slowly growing tumors that stained positively for SA-β-gal, and which upon inactivation of p53 eventually become fast-growing and SA-β-gal-negative (Chen et al., 2005). Also, acute somatic activation of ErbB2 in mammary glands led to the development of precancerous lesions where 75% of cells expressed SA-β-gal. Importantly, these lesions rapidly developed into tumors, where 20–40% of cells were still showing positive staining for SA-β-gal (Reddy et al., 2010). Therefore, at present, the question regarding the role of OIS in preventing tumor development remains open.

One interpretation of the presence of cells with senescent phenotype in hyperplastic lesions and tumors is that there is an ongoing transition to senescence of a fraction of tumor cells, which become SA-β-gal positive, while the rest of the cells remain transformed and proliferative. Alternatively, it is possible that morphological and biochemical signs of senescence are not rigidly linked to permanent growth arrest. In fact, owing to the technical difficulties in many cancer models where senescence was studied, monitoring of SA-β-gal and other senescence-associated biochemical parameters is usually not followed by assessment of cell cycle arrest. Here, we hypothesize that activation of oncogenes in immortalized or cancerous cells, which express endogenous telomerase activity and commonly lack critical regulators of cell cycle progression, may be sufficient to trigger morphological and biochemical manifestations of senescence (e.g. SA-β-gal), but cannot ensure full growth arrest. We call this condition senescence with incomplete growth arrest (SWING). In contrast, activation of oncogenes in finite lifespan cells can cause true senescence resulting in block of tumorigenesis. The latter condition could reflect appearance of senescence markers in nontumorigenic nevi. These different responses of finite and immortalized cells to oncogenes support the concept that cellular senescence represents a set of subprograms with independent end points. Accordingly, in response to activation of oncogenes, immortalized cells may trigger some of the sub-programs (e.g. cell enlargement and SA-β-gal activity), while having impaired capability to execute the permanent growth arrest because of the multiple checkpoint deficiencies.

Here, we investigate responses of immortalized HMEC to Her2 and Ras oncogenes. The proto-oncogene Her2 is overexpressed in approximately 30% of aggressive breast carcinomas (Slamon et al., 1989). Furthermore, expression of this oncogene in mammary epithelium in transgenic mice triggers the appearance of breast tumors (Muller et al., 1988). However, as with Ras or Raf-1 oncogenes, expression of the rodent oncogenic form of Her2 (NeuT) in the human breast carcinoma line MCF7 and low-passage murine embryonic fibroblasts failed to stimulate proliferation but instead provoked accumulation of enlarged SA-β-gal-positive cells, which were considered senescent (Trost et al., 2005). Therefore, NeuT oncogene along with Ras represents a relevant disease-related model to investigate OIS in immortalized mammary epithelium cells.

It has been suggested that DNA damage triggered by oncogene imbalance underlies OIS growth arrest. Activated components of the DDR, such as phATM, phChk2, or 53BP1 were detected in various untreated preneoplastic lesions associated with bladder, lung, breast, and colon tumors (Bartkova et al., 2005; Gorgoulis et al., 2005; Campisi, 2010). Furthermore, inhibition or knocking down of the DDR mediators prevented OIS and facilitated transformation of melanocytes (Di Micco et al., 2006; Mallette et al., 2007). Unexpectedly, in the process of this investigation, we uncovered that SWING cells with activated oncogenes, when challenged with various genotoxic stresses, have an impaired ability to activate Chk1 kinase, or carry out phosphorylation of the histone H2A isoform H2AX, which are involved in promoting double-stranded DNA breaks (DSB) repair as part of the DDR. H2AX is a critical factor implicated in early stages of major DSB repair pathways (Thiriet & Hayes, 2005). Phosphorylated H2AX molecules, called γH2AX, form irradiation-induced foci in the vicinity of DSBs, which serve as platform for recruitment of DNA repair factors, thus promoting DSB repair. Through assisting in DNA repair and facilitating chromatin remodeling, γH2AX is involved in the maintenance of DNA stability (Zha et al., 2008; Chanoux et al., 2009). Accordingly, suppression of γH2AX in SWING cells led to the appearance of chromosomal aberrations because of impaired DNA repair. Here, we connect the downregulation of H2AX and Chk1 upon development of SWING state to increased DNA instability in neoplastic transformation.

Results

NeuT oncogene induces true senescence in finite lifespan HMEC and senescence-associated properties without growth arrest in immortalized HMEC

To assess OIS in finite and immortalized HMEC, we compared effects of the oncogenic rodent activated form of Her2 (NeuT) on proliferation of normal finite lifespan 184 HMEC (further marked HMEC) and two distinct immortalized HMEC lines, MCF10A and 184B5, the latter derived from normal 184 HMEC. Finite and immortal HMEC were infected with control or NeuT-expressing retrovirus, and NeuT expression was confirmed by immunoblotting (Fig. 1A). At day 5 postinfection, HMEC/NeuT and MCF10A/NeuT cells acquired typical senescent morphology, where more than 90% of cells became enlarged, flattened, and highly vacuolized (Fig. 1B). 184B5/NeuT cells also acquired senescent morphology but later, on day 12 postinfection (Fig. 1B). Furthermore, approximately 90% of HMEC/NeuT, approximately 50% of MCF10A/NeuT, and approximately 70% of 184B5 cells became SA-β-gal positive (Fig. 1B). In contrast, finite HMEC, the MCF10A, and 184B5 lines expressing control virus retained normal morphology, and fractions of SA-β-gal positive cells were 12%, 3%, and 15% respectively (Fig. 1B). Therefore, by morphological and biochemical criteria, almost all the finite HMEC and the majority of the immortalized MCF10A and 184B5 cells acquired properties associated with senescence upon expression of the NeuT oncogene. Despite these similarities, we observed dramatic difference in growth properties of HMEC/NeuT compared with MCF10A/NeuT and 184B5/NeuT cells. To assess growth status of control and NeuT-expressing cell populations, we monitored expression of the proliferation marker, Ki67. More than 80% of both finite and immortal HMEC populations infected with control virus demonstrated Ki67 positivity (Fig. 1C). In line with the senescent appearance, almost no HMEC/NeuT cells (< 5%) stained positive for Ki67, indicating that NeuT oncogene triggered bona fide senescence in these cells. In contrast, the MCF10A/NeuT and 184B5/NeuT populations contained Ki67-positive cells, similar to controls (> 70%), suggesting that these cells were not cell cycle arrested. Noticeably, typically senescent looking large and flattened cells show strong Ki67 fluorescence (Fig. 1C, bottom panel). In agreement with the Ki67 staining, direct measurement of growth rates demonstrated that HMEC/NeuT cells stopped dividing by day 5 postinfection, while MCF10A/NeuT and 184B5/NeuT cells continued proliferating, albeit slower than cells infected with control virus (Fig. 1D). Thus, in contrast to finite HMEC, expression of NeuT oncogene in immortalized HMEC results in the development of a senescence morphology but not growth arrest.

Fig. 1.

Effects of oncogenes on senescence and growth properties of MCF10A, finite 184 human mammary epithelial cells (HMEC) and 184B5cells. (A) Retroviral expression of NeuT in MCF10A, HMEC, and 184B5 cells. Samples were collected 5 days postinfection (MCF10A and HMEC) or 12 day postinfection (184B5) and probed by immunoblotting with anti-NeuT antibody. (B) β-Gal staining of cells infected with either NeuT- or Ras-expressing viruses. Tests were performed with cells on same days postinfection, as described earlier. (C) Ki67 immunostaining of cells expressing NeuT oncogene. Nuclei are stained with DAPI. Bottom panels represent enlarged images of the corresponding areas on the figure. (D) NeuT expression blocks growth of HMEC cells but not MCF10A cells. Cells were plated in triplicates at 10% confluence, and growth rate was measured by cell counting over the period of 3 days. (E). NeuT expression leads to increase in sizes of MCF10A cells. Cells were removed from plates with trypsin, dropped onto the slide, and pictures were taken. Image analysis was carried out using AxioVision software.

To further characterize the unusual phenotype of the MCF10A/NeuT and 184B5/NeuT cells, we assessed other senescence-associated properties. Direct measurement of cellular diameters using AuxioVision (Carl Zeiss, Oberkochen, Germany) software (see Experimental Procedures) demonstrated that, on average, diameters of MCF10A/NeuT expressing cells were 60% larger than those in control population (Fig. 1E), which is typical for senescent cells. Importantly, the senescence morphology in MCF10A/NeuT and 184B5/NeuT cells was associated with the accumulation of the CDK inhibitor p21 (Fig. 4A), which is implicated in OIS in various cell lines (Quereda et al., 2007). Of note, we did not detect senescence-associated heterochromatic foci, SAHF, in MCF10A/NeuT cells. On the other hand, SAHF formation is more commonly observed in senescent fibroblast than epithelial cells, where it depends on the CDK inhibitor p16, absent from MCF10A (Narita & Lowe, 2004).

Fig. 4.

SWING phenotype depends upon p21 and exists in cancerous cells. (A) NeuT expression causes induction of p21. p21 Levels were measured by immunoblotting in MCF10A cells at day 6 postinfection and in 184B5 at day 12 postinfection. (B) Development of SWING phenotype is p21- and p53-dependent. Fractions of SA-β-gal-positive cells were measured in NeuT-expressing WT, p21KO, and p53 knockdown cells. Three independent fields were scored. (C) SA-β-gal staining of a focus in culture formed by MCF10A cells transformed by NeuT. (D) SA-β-gal staining of a focus formed byMCF10A cells isolated from tumor xenograft. Tumor xenografts were established from NeuT-transformed MCF10A cells, (E) SA-β-gal and Ki67 double staining of NeuT-induced tumor from the MMTV neu transgenic mouse. Newly emerged tumor was excised and stained for β-gal activity; stained tumor was sectioned and processed for immunohistochemical detection of Ki67. Right panel shows section stained for both SA-β-gal and Ki67; sections stained only for SA-β-gal (left panel) or Ki67 (middle panel) are shown for comparison.

As MCF10A growth was slowed following NeuT expression, there was a possibility that with each round of division, a fraction of the MCF10A/NeuT becomes truly senescent (i.e., growth arrested and SA-β-gal positive), while in the rest of the population, the senescence program is not activated. To assess growth status of the entire MCF10A/NeuT population, we used FACS analysis. Cell cycle distribution of MCF10A/NeuT cells was almost undistinguishable from that of control MCF10A cells (Fig. 2A), suggesting that the total populations proliferate similarly. To further validate the lack of growth arrest in a subpopulation of NeuT-expressing cells, we employed a method based on staining cells with a fluorescent lipophylic dye PKH67. This dye accumulates in cellular membranes and remains there without defusing into the medium. Therefore, nondividing cells retain PKH67 fluorescence, while PKH67 is diluted in proliferating cells with each round of division. This method allows detecting even a minor population of growth-arrested cells. Control and NeuT-expressing MCF10A cells were loaded with PKH67, the excess of the dye was washed out, and fluorescence was monitored daily for 3 days. We captured images and analyzed fluorescence in three independent fields. As seen in Fig. 2B, upon loading of PKH67 at day 0, control and NeuT-expressing cells showed similar high fluorescence, with the median fluorescence intensity around 2600 relative units. Decrease of average fluorescence over time was 20% slower in MCF10A/NeuT cells as compared to control MCF10A cells (Fig. 2B), which was in agreement with a decrease in growth rate of these cells. Importantly, at day 3, median fluorescence of the population of MCF10A/NeuT cells was reduced to < 1000 relative units and no brightly stained cells with fluorescence > 2000 relative units (corresponding to permanently growth arrested cells) were detected. These results indicate that almost every cell in MCF10A/NeuT population underwent divisions during the 3 days period. In contrast, in MCF10A/NeuT cells forced to exit the cell cycle by treatment with a low dose of doxorubicin, median fluorescence after 3 days was 2400 units, which was not significantly different from that of cells at day 0 (Fig. 2B).

Fig. 2.

Effects of NeuT expression on cell cycle progression ofMCF10A cells. (A). Cell cycle distribution of control and NeuT-expressing MCF10A cells. FACS analysis of cells harvested on day 6 postinfection. (B) Divisions of cells measured by the PKH67 loading technique. Control, NeuT- and Ras-expressing cells were loaded with PKH67 and then plated at 10% confluency. Pictures were taken for three consecutive days; fluorescence intensity of cells was analyzed in three random fields using AxioVision software. As positive control, we treated cells with 40 nm doxorubicin, which caused minor growth inhibition in control cells and complete growth inhibition in oncogene-expressing cells. (C). S-phase imaging in MCF10A/NeuT cells. MCF10A/NeuT at day 4 postinfection were infected with cell-cycle reporter lentiviruses, i.e., mKO2-hCdt1(30/120) [G1/S; red] and mAG-hGem(1/110) [S/G2; green], and images captured. (D) Time-lapse analyses of division of a senescence-looking flat MCF10A/NeuT cell. (Arrowhead points the cell undergoing mitosis).

Further, to directly observe that flat, enlarged senescent-looking cells can enter S-phase, we infected MCF10/NeuT cells with lentiviruses expressing the cell-cycle reporters (Sakaue-Sawano et al., 2008). In line with the FACS data, we observed that approximately 15% of the MCF10A/NeuT population expressed both G1/S and S/G2 markers indicative of cells in S-phase (Fig. 2C). Finally, we directly visualized cell division by monitoring MCF10A/NeuT cultures under the microscope for 24 h and capturing images every 30 min. As seen in movie frames in Fig. 2D, a randomly chosen flat enlarged senescent-looking cell underwent a division. Altogether these experiments indicate that in the immortal MCF10A line, NeuT expression precipitated an unusual phenotype, i.e. Senescence With Incomplete Growth arrest, which we called SWING. These results indicate that while NeuT expression promotes OIS in finite lifespan HMEC, in immortal HMEC, its expression triggers the SWING state.

Further we investigated what component of the immortalization process promotes a switch from OIS to the SWING state. Accordingly, we restored expression of p16 in MCF10A/NeuT cells, which lack this cell cycle inhibitor. The expression level of p16 that we achieved in this system was similar to the endogenous level of p16 in normal HMEC cells passage 4 (Fig. 3A). As MCF10A/NeuT cells, MCF10A/NeuT/p16 cells had SA-β-gal activity in about 70% of the population (Fig. 3B). At the same time, these cells continued proliferating and about 60% displayed Ki67 staining (Fig. 3C). Therefore, reconstitution of p16 did not reverse NeuT-induced SWING state to OIS.

Fig. 3.

Effects of p16 and hTERT on the development of the SWING. (A) Retroviral expression of recombinant p16 in MCF10A cells; comparison with p16 levels in finite human mammary epithelial cells. (B) Effects of p16 and NeuT on fractions of SA-β-gal-positive MCF10A cells. Three independent fields were scored. (C) Effects of p16 and NeuT on fractions of Ki67-positive MCF10A cells. Three independent fields were scored. (D) Effects of NeuT on fractions of SA-β-gal-positive 184BTERT and 184DTERT cells. Three independent fields were scored. (E) Effects of NeuT on fractions of Ki67-positive 184BTERT and 184DTERT cells.

To test for the role of telomerase in establishing the NeuT-induced SWING state, we utilized two derivatives of the finite lifespan 184 HMEC: (i) 184DTERT cells, which were obtained by ectopic expression of hTERT in 184 HMEC, and (ii) 184BTERT cells which were obtained by the expression of recombinant hTERT in 184 HMEC line that spontaneously lost p16 because of promoter methylation (Brenner et al., 1998; Stampfer et al., 2001; Garbe et al., 2009). Both 184DTERT and 184BTERT cells were infected with retrovirus expressing NeuT. By day 8 postinfection, both cell lines acquired typical senescence morphology, including enlargement and flattening, and about 80% of cells became SA-β-gal positive (Fig. 3D), similar to the effect of NeuT on finite lifespan HMEC (Fig. 1B). Importantly, in contrast to finite HMEC, most of NeuT-expressing 184DTERT and 184BTERT cells remained Ki67-positive (Fig. 3E), although certain decrease in the population of Ki67-positive cells upon NeuT expression was seen with 184DTERT (Fig. 3E). These data indicate that expression of telomerase was sufficient to switch the response of epithelial cells to oncogene from OIS to SWING. Surprisingly, this switch was independent on the p16. Next, we investigated whether oncogenes other than NeuT could trigger transition to the SWING state. MCF10A cells were infected with control retrovirus or retrovirus expressing the oncogenic form of Ras, H-RAS V12. Microscopic examination of cells on 5 days postinfection showed that similar to MCF10A/NeuT cells, approximately 70% of MCF10A/Ras cells attained enlarged, flattened, and highly vacuolized morphology and became SA-β-gal positive (Fig. 1B). These data indicate that MCF10A/Ras cells acquired senescent characteristics. To elucidate whether activation of Ras oncogene triggers permanent growth arrest in MCF10A cells, we loaded control and MCF10A/Ras cells with PKH67 and monitored cell divisions for three consecutive days, as described earlier. We observed that the pace of PKH67 reduction was only slightly diminished in MCF10A/Ras cells compared with control cells, indicating that flat, enlarged, highly vacuolized MCF10A/Ras cells continued dividing (Fig. 2B). Therefore, expression of Ras and NeuT oncogenes in immortal MCF10A cells failed to promote growth arrest but did promote transition into the SWING state, where cells attained some senescence- associated sub-programs (e.g. morphological changes and expression of SA-β-gal).

Acquisition of SWING phenotype depends on p21

CDK inhibitors, p21, p16, and p15, have been reported to be mediators of OIS in some cell types (Quereda et al., 2007). MCF10A, like most in vitro immortalized HMEC lines, does not express p16. As noted, expression of p21 increased approximately two-fold in NeuT-exposed MCF10A and 184B5 (Fig. 4A). Therefore, we investigated whether p21 is implicated in the establishment of the SWING state, using a p21 knockout derivative of MCF10A cells (Karakas et al., 2006). Parental and p21KO MCF10A cells were infected with control or NeuT-expressing virus, as described earlier. Microscopic observation of cells for 12 days starting on day 5 postinfection revealed that in contrast to parental cells, NeuT expressing p21KO cells had minimal senescence-associated changes in morphology. Furthermore, SA-β-gal assay performed on days 5 and 10 postinfection showed that < 10% of MCF10A p21KO/NeuT cells became SA-β-gal positive as compared to approximately 50% SA-β-gal positive parental MCF10A/NeuT cells (Fig. 4B). Next, we demonstrated that p53, which regulates p21 at the level of transcription, is similarly critical in establishing the SWING state. In fact, knockdown of p53 using shRNA almost completely (to < 5%) blocked development of SA-β-gal activity and senescence morphology following NeuT expression (Fig. 4B). Thus, the p53-p21 pathway is required for the NeuT-induced SWING state.

SWING cells could be malignantly transformed

Expression of NeuT in MCF10A cells obtained from ATCC (see Experimental Procedures) established the SWING phenotype, but did not transform them to malignancy as determined by the lack of formation of foci on plates and tumors in nude mice. To address whether the senescence features of NeuT-expressing cells affect tumorigenesis, we assessed malignancy- associated properties of the SWING cells. In these experiments, we used a clone of MCF10A cells obtained from Dr. B. Park that spontaneously acquired the ability to form foci on plates and tumors in nude mice in response to expression of NeuT, suggesting malignant transformation (Fig. 4C). Notably, similar response to NeuT was demonstrated previously in certain MCF10A clones (Giunciuglio et al., 1995). Surprisingly, there was no apparent selection against the senescence properties, and foci formed by MCF10A/NeuT cells stain positive for SA-β-gal (Fig. 4C), indicating that they retain the SWING phenotype. To analyze the stability of the SWING phenotype, we established xenografts in nude mice. Accordingly, 0.5 × 10⁶ control and MCF10A/NeuT cells were injected subcutaneously into front flanks of nude mice, and tumor growth was monitored by caliper. As expected, control MCF10A cells did not form tumors in mice, while MCF10A/NeuT cells readily formed tumors that became palpable 7 days postinjection. Further, we excised tumors and re-established them in cell culture. Cells in these cultures formed foci, indicating that they retain this transformed property. Importantly, these cells stained positively for SA-β-gal, demonstrating that SWING state is stable and persists in tumors (Fig. 4D).

Next, we evaluated whether SWING cells can be detected in a transgenic mouse model of NeuT-induced mammary tumor. Accordingly, tumors that appeared in 6–7-month-old transgenic MMTV/NeuT mice were excised, and processed for immunohistological analysis. Slides were double stained for the senescence marker SA-β-gal and the proliferation marker Ki67. Of note, Ki67 stains nuclei, while SA-β-gal stains cytoplasm, and therefore, the presence of the SA-β-gal cytoplasmic staining adjacent to Ki67 nuclear staining indicate the double-stained cell. Tumor sections contained significant cell populations that were positive for both markers (although the percentage of double-stained cells varied significantly between the sections), thus confirming that a population of cells in NeuT-initiated tumors retains SWING phenotype in vivo (Fig. 4E). To clarify the appearance of double staining, Fig. 2E shows panels single-stained with either Ki67 or SA-β-gal, and the double-stained panel.

These experiments establish that (i) biochemical properties associated with senescence (SA-β-gal) in immortalized cells and oncogene-expressing tissue may develop independently of growth arrest, (ii) development of the SWING state instead of OIS associates with reactivation of telomerase, (iii) cells in SWING state may be malignantly transformed, and (iv) the SWING state is stable, is not lost after passage of cells through xenografts, and is retained in early tumors.

SWING cells have impaired DNA damage response and increased DNA instability

Recently, we showed that ectopic expression of mediators of the senescence response (e.g. p21 or p16) increases DNA instability and impairs activation of certain branches of the DDR, i.e. accumulation of γH2AX and phosphorylation of Chk1 kinase (Gabai et al., 2008, 2010) in several cancer cell lines. These data were consistent with prior observation that cellular senescence suppresses certain DNA repair pathways (Seluanov et al., 2004). Here, we investigated whether these components of the DDR are also suppressed in SWING cells.

γH2AX serves as an early and critical step in sensing and subsequent repair of DSB. Therefore, we next investigated effects of transition to SWING state on formation of γH2AX foci at the sites of damaged chromosomes. Control and MCF10/NeuT cells were treated with doxorubicin, a potent inducer of DSBs, and γH2AX foci were visualized by immunofluorescence. Incubation of control cells with 200 nm of doxorubicin for 3 h led to the appearance of approximately 30% of cells with more than 10 γH2AX foci per nucleus, and 22% of cells with 1–10 foci per nucleus. Under these conditions, there was 50% reduction in foci formation in NeuT- expressing MCF10A cells, i.e. < 15% of nuclei with more than 10 foci and approximately 10% nuclei with 1–10 foci (Fig. 5A). Thus, activation of NeuT oncogene strongly suppressed γH2AX foci formation.

Fig. 5.

Downregulation of Chk1 and H2AX in SWING cells. (A) Effect of NeuT oncogene on γH2AX foci formation following doxorubicin treatment. Control and NeuT-expressing cells were treated with 200 nM doxorubicin for 3 h; cells were fixed and immunostained with anti-γH2AX antibody. (B) Effects of Ras and NeuT oncogenes on γH2AX levels following treatments with genotoxic stresses. Treatment conditions were the same as in Fig. 4A. Levels of γH2AX were assessed by immunoblotting. Of note, in p21KO and shp53 cells (upper panel), expression of NeuT did not lead to suppression of γH2AX. (C) Effects of oncogenes on the expression levels of histone H2AX, as assessed by immunoblotting with anti-H2AX antibody. Samples were taken on day 6 postinfection forMCF10A cultures and on day 12 postinfection for 184B5 cultures. (D) Expression of NeuT upregulates p21 and downregulates H2AX in mammary tissue of 3-month-old mice. Mammary glands were collected from control and MMTV/NeuT animals and sections stained for p21 and H2AX with the corresponding antibodies. (E) Oncogene-mediated suppression of Chk1 phosphorylation in response to genotoxic treatments. Cells were infected with control, NeuT- or Ras-expressing retroviruses. At day 6 postinfection, cells were treated with either 100 J/m⁻² UV or 0.5 μm doxorubicin. Samples were taken at indicated time points after the treatments and levels of phosphorylated Chk1 were assessed by immunoblotting. Bottom panel – samples were taken 1 h after UV irradiation or 5 h treatment with doxorubicin. (F) Effects of NeuT oncogene on γH2AX and phChk1 levels in 184B5 following treatments with indicated doses of doxorubicin.

In parallel, we used immunoblotting to assess levels of γH2AX in oncogene- expressing MCF10A cells exposed to genotoxic stresses. In control cells, both UVC irradiation and doxorubicin treatment led to the rapid appearance of γH2AX (Fig. 5B). Under these conditions, accumulation of γH2AX was significantly reduced in MCF10A/NeuT and MCF10A/Ras cells (Fig. 5B), further indicating that establishment of the SWING state associates with impairment of DDR signaling. At least in part, reduction of γH2AX foci formation and γH2AX levels was associated with the overall downregulation of H2AX expression. We detected two-fold reduction in H2AX levels in both MCF10A/NeuT and MCF10A/Ras cells as compared to control cells; in MCF10A/NeuT cells, the reduction was seen as early as on the second day post-infection (Fig. 5C). Thus, reduction of γH2AX foci formation and γH2AX levels was an early response to oncogene activation and was associated with overall downregulation of H2AX expression.

To investigate effects of NeuT on H2AX in vivo, we utilized the MMTV/NeuT transgenic model of the Her2-positive cancer. Expression of NeuT in this model starts at the age of about 2 months because of elevation of estrogen levels. Cancers develop much later in about 12-month-old animals. To assess early effects of NeuT, mammary tissue was collected from the 3-month-old MMTV/NeuT and control mice, and sections were immunostained with anti-p21 and anti-H2AX antibody. As seen on Fig. 5D, NeuT caused induction of p21 and strong downregulation of H2AX, thus replicating effects in cell culture. These biochemical changes coincide with the development of SA-β-gal activity, as reported previously (Reddy et al., 2010).

As establishment of SWING state critically depends upon accumulation of p21, to link H2AX downregulation with the development of SWING phenotype, we utilized the p21KO derivative of MCF10A cells. Control and NeuT-expressing MCF10A p21KO cells were treated with doxorubicin, and levels of γH2AX and total H2AX were assayed by Western blotting with the corresponding antibodies. In contrast to parental MCF10A cells, in p21KO cells, expression of NeuT neither caused suppression of γH2AX following genotoxic treatments (Fig. 5B), nor led to reduction of H2AX levels (Fig. 5C). Similarly, we did not observe NeuT-induced suppression of γH2AX in p53 knockdown cells (Fig. 5B). Thus, downregulation of H2AX following NeuT expression depends on p53 and p21, thus linking the DDR suppression to the development of the SWING state.

Similarly, we observed that expression of either NeuT or Ras oncogenes reduced activation of Chk1. In control cells, UVC irradiation or treatment with doxorubicin led to a strong activation of Chk1; however, in both MCF10A/NeuT and MCF10A/Ras cells activation of Chk1 was strongly suppressed (Fig. 5E). Similar to MCF10A/NeuT, we observed strong suppression of both γH2AX and phChk1 in 184B5/NeuT cells challenged with doxorubicin (Fig. 5F). In these cells, impairment DDR may also be mediated by downregulation of overall levels of H2AX and Chk1. (Figure 5C and data not shown).

Suppression of phChk1 and γH2AX triggered by oncogenes as part of the SWING state could contribute to the increased DNA instability and karyotype abnormalities seen in cancers. Accordingly, we performed micronucleus (MN) assay. Micronucleus arise when extra-nuclear chromosomal fragments induced by DNA damage fail to incorporate into the nucleus, and therefore, MN frequency correlates with genome instability. There was only a minor difference in MN frequencies inMCF10A/NeuT and MCF10A control cells. However, NeuT expression significantly potentiated MN formation inMCF10A cells upon exposure to doxorubicin (Fig. 6A).

Fig. 6.

Chromosome instability and sensitivity toward genotoxic stress in SWING cells. (A) Fractions of micronuclei in control and NeuT-expression cells with or without genotoxic treatments. Cells were treated with indicated doses of doxorubicin for three hours and micronuclei were counted. (B) Centrosome abnormalities in SWING cells. Control and NeuT-expressing cells were fixed on day 6 postinfection and immunostained with γ-tubulin to visualize centrosomes. Fractions of cells with defective centrosomes or increased number of centrosomes were counted. Centrosomes are marked by white arrowheads. Blow up of a section of cells with abnormal centrosomes is shown in two small panels. Nuclei with normal centrosome pair is shown on the right panel, in the middle. Of note, overexpression of H2AX reverses the effect of NeuT expression. (C) Levels of H2AX in cells infected with H2AX-expressing retrovirus. (D) Overexpression of H2AX reverses effects of NeuT on p21 induction by doxorubicin. Conditions were the same as in Fig. 4. (E) SWING cells are more sensitive to genotoxic stress than control cells. Graph shows fractions of Ki67-positive cells in control and NeuT-expressing populations following 3-day treatment with indicated doses of doxorubicin. (F) Effects of H2AX expression on propagation of SWING cells following doxorubicin treatment (50 nM) as measured in PKH67 loading experiment, same as in Fig. 3B.

Increased DNA instability in cancer is often manifested in aneuploidy. There are numerous reports demonstrating that aneuploidy can be triggered by the centrosome dysfunction (Katsura et al., 2009). To address whether SWING state associates with centrosome dysfunction, we compared the number of centrosomes detected by immunofluorescence with anti-γ-tubulin antibody in control MCF10A and MCF10A/NeuT cells. NeuT expression significantly potentiated centrosome abnormalities. More than 30% of MCF10A/NeuT cells had either more than two centrosomes per cell or broken centrosomes, as compared to 13% in control MCF10A cells (Fig. 6B).

To test whether H2AX downregulation potentiates the centrosome abnormalities in SWING cells, we expressed H2AX protein using the retroviral expression system. (Expression levels of H2AX are shown on Fig. 6C.) Specifically, cells were simultaneously infected with NeuT and H2AX viruses or corresponding control viruses and centrosomes were scored at day 6 postinfection. H2AX expression significantly alleviated centrosome abnormalities seen in SWING cells (Fig. 6B). Taken together, these data indicate that SWING state is characterized by enhanced DNA instability apparently because of suppression of certain branches of the DDR.

SWING cells are sensitive to genotoxic stresses

As SWING cells have reduced activation of Chk1 and H2AX and higher DNA instability, we hypothesized that low-level genotoxic stresses may cause stronger DNA damage in these cells compared with control cells, thus leading to enhanced induction of p21. To test this possibility, we treated control and MCF10A/NeuT cells with doxorubicin and assayed p21 levels by immunoblotting. Following doxorubicin treatment, the levels of p21 were higher in MCF10A/NeuT cells than in control cells (Fig. 6D). To test whether stronger p21 induction was caused by the reduced expression of H2AX, we expressed H2AX in control and MCF10A/NeuT cells using a retroviral expression system, and monitored p21 levels. A buildup of p21 upon exposure to low doses of doxorubicin was suppressed in H2AX-MCF10A/NeuT cells, as compared to MCF10A/NeuT cells (Fig. 6D). Therefore, the H2AX deficiency observed in SWING cells is critical for over induction of p21.

Increased accumulation of p21 suggests that SWING cells exposed to DNA-damaging stresses may be vulnerable to senescence and permanent growth arrest, and thereby may have increased sensitivity to genotoxic stress. In line with this possibility, there was only minor reduction in the fraction of Ki67 positive cells when control MCF10A cells were treated with 50 nm doxorubicin (from 78% to 70%), while under similar conditions the fraction of Ki67-positive MCF10A/NeuT cells dropped from 84% to 30% (Fig. 6E). Incubation with 100 nm doxorubicin led to about 40% of Ki67-positive control cells and 14% of Ki67-positive MCF10A/NeuT cells (Fig. 6E).

Further, we monitored growth of control and MCF10A/NeuT cells challenged with low doses of doxorubicin (40 nm) using PKH67. As mentioned earlier, under these conditions, MCF10A/NeuT cells almost completely stopped dividing as manifested by accumulation of significant proportion (about 80%) of brightly stained cells (Fig. 2B). On the other hand, this treatment led to only minor reduction in the rate of dye dilution in control MCF10A cells, indicating that the low dose of doxorubicin did not provoke significant growth inhibitory response in these cells (Fig. 2B right panels and graph). Similarly, PKH67 labeling of MCF10A/Ras cells showed that these cells become highly prone to senescence-inducing stresses and readily stopped dividing upon treatment with 40 nm doxorubicin (Fig. 2B right panels). These experiments indicate that expression of NeuT or Ras oncogenes sensitized MCF10A cells to DNA damage by promoting growth arrest following genotoxic insult.

A plausible explanation for this effect is that SWING cells have a defect in H2AX, which is critical for DNA repair, and thus experience stronger DNA damage when exposed to genotoxic treatments. Hence, we tested whether expression of H2AX can reverse growth inhibition of SWING cells challenged by low doses of genotoxic stress. Growth rates of cells infected simultaneously or separately with NeuT and H2AX viruses as well as control viruses were evaluated using the PKH67 loading experiment, as described earlier (see Fig. 2B). Expression of H2AX significantly potentiated growth of MCF10A/NeuT cells treated with low doses of doxorubicin (Fig. 6F). Of note, growth of untreated or doxorubicin-treated control MCF10A cells was not affected by H2AX overexpression. Altogether these experiments indicate that (i) downregulation of H2AX is an important property of the SWING cells, (ii) downregulation of H2AX leads to DNA instability and subsequent over induction of p21, which in turn makes SWING cells sensitive to genotoxic stresses.

Discussion

Oncogene-induced senescence represents one of the major protective responses that guard multicellular organisms from tumor development. However, oncogenes are reported to have opposing effects on cell proliferation and may trigger both senescence and transformation (Braig et al., 2005; Bartkova et al., 2006). These observations make it critical to investigate what specific parameters determine the outcome of oncogene activation on cellular proliferation. Here, we report that expression of either NeuT or Ras oncogene in immortalized HMEC lines MCF10Aand 184B triggers the development of the SWING state where cells acquire major characteristics of senescent cells, including enlargement, flattening, vacuolization and SA-β-gal activity, but do not exit the cell cycle. In contrast, expression of NeuT oncogene in finite HMEC triggers all the same morphological changes accompanied with a similar extent of SA-β-gal accumulation, while completely preventing cell proliferation. These unexpected findings indicate that appearance of SA-β-gal, which commonly serves as a major criterion of senescence, may be disconnected from the onset of growth arrest. In line with this idea, there are many reports demonstrating that proliferative hyperplastic tissues and early tumors stain positive for SA-β-gal (Collado et al., 2005; Di Micco et al., 2006; Reddy et al., 2010). Importantly, this work implies that OIS could be viewed as a complex of distinct subprograms, which can be expressed independently of each other, and therefore, are under separate molecular control.

Next, we investigated weather failure of MCF10A cells to achieve the major senescence-associated end point of growth arrest, in response to Ras or NeuT activation, is because of deficiencies in cell cycle regulation that occurred during the process of immortalization. For example, MCF10A cells do not express the CDK inhibitor p16 and have re-activated telomerase. Noteworthy, we have found that reconstitution of p16 levels in MCF10A cells did not restore OIS following expression of NeuT, and did not affect the development of the SWING state. In contrast, expression of hTERT in finite lifespan HMEC was sufficient to switch cellular response to NeuT from OIS to the SWING state. Indeed, these cells, while expressing SA-β-gal at about the same level as control HMEC, continue proliferating and were Ki67-positive. Therefore, telomerase appears to play the major role in development of the SWING state, while p16 seems to be irrelevant. These data indicate that NeuT- or Ras-induced expression of p21 in cells that have reactivated telomerase does not precipitate OIS, but rather triggers transition to the SWING state. Noteworthy, further switching off of cell cycle regulators prevented the development of all signs of senescence. In fact, we have shown that SWING phenotype critically depends upon p21 and did not develop in either MCF10A p21KO cells or p53 knockdown cells.

Interestingly, the SWING state is associated with the suppression of DDR, i.e., is characterized by impaired activation of H2AX and Chk1. Of note, DDR suppression apparently is also mediated by the p21 pathway as it was observed neither in MCF10A cells depleted of p53 nor in MCF10A p21 knockout cells.

Suppression of H2AX and Chk1 causes a number of defects in SWING cells, including, increased sensitivity to genotoxic stress, increased growth arrest in response to low doses of genotoxic stress, and increased chromosome instability. Importantly, all these deficiencies can be reversed upon expression of H2AX. We therefore suggest that reduced H2AX activation leads to DNA damage, especially upon exposure to minor genotoxic stress, which in turn promotes instability. An interesting possibility is that increased DNA instability in cancer, at least in early stages, could be associated with the SWING phenotype of cells and consequently suppression of SWING could reduce DNA instability in cancer. In line with this possibility, we observed that expression of NeuT promotes upregulation of p21 and downregulation of H2AX in mammary tissue of transgenic mice obtained from 3-month-old animals. These effects take place very soon after the beginning of NeuT expression and 9–10 month prior to cancer development. We propose that p21-dependent downregulation of H2AX may be an important factor in DNA instability in early Her2-driven precancerous lesions (Chen et al., 2009).

An important observation is that cells in SWING state remain vulnerable to senescence when challenged by a minor genotoxic stress that did not affect proliferation of non-SWING cells (Fig. 2B). These results raise the possibility that the enhanced sensitivity of SWING cells to genotoxic stresses could be used as an approach toward cancer therapy. In fact, these cells may be especially sensitive to additional suppression of H2AX.

Noteworthy, despite that SWING cells may easily progress to either senescence arrest or malignant transformation, the SWING state is relatively stable and persists while passing cells through xenografts and re-establishing cell culture. Furthermore, we were able to detect a population of SWING cells, expressing both SA-β-gal and Ki67 markers, in mammary tumors from the MMTVneu mouse model. These findings suggest that SWING cells may account for at least some of the previously reported SA-β-gal-positive cells found in tumors.

These data further support the suggestion that cell immortalization is a critical step in tumorigenesis as immortalization favors transition into the SWING upon oncogene activation, which can further evolve to a malignant state. In contrast, as reported previously, activation of oncogenes in normal epithelial and fibroblast cells leads to the onset of OIS (Zhu et al., 1998). The reactivation of telomerase associated with immortalization may be part of the process responsible for this shift in response to oncogene activation from OIS to SWING (Olsen et al., 2002; Stampfer & Yaswen, 2003).

Overall, this work describes new phenomenon associated with oncogene activation and OIS, i.e., the existence of the SWING state, associated with decreased DDR capacity and increased vulnerability to genotoxic stresses. These results suggest a novel contributor to DNA instability in cancer, and the potential for novel clinical interventions in cancer by engaging the intact capacity of cells in the SWING state to growth arrest in response to genotoxic stresses.

Experimental procedures

Cell cultures, treatments, and reagents

HEK293 were from American Type Culture Collection. There were two sources of MCF10A cells, from ATCC and a kind gift of Dr. B. Park. MCF10A p21KO cells were from Dr. B. Park. HEK293 cells were grown in DMEM supplemented with 10% fetal bovine serum; MCF10A and MCF10A p21KO cells were grown in DMEM/F12 (1:1, v:v) containing 5% horse serum and supplemented with 10 mg L⁻¹ insulin, 20 μg L⁻¹ epidermal growth factor, 50 μg L⁻¹ cholera toxin, 50 mg L⁻¹ hydrocortisone, 100 units mL⁻¹ penicillin, and 0.1 mg mL⁻¹ streptomycin in a humidified environment at 37 °C with 5% CO₂. Finite prestasis HMEC from specimen 184, batch D, and their immortalized derivative lines, 184BTERT, 184DTERT, and 184B5, were derived and grown as described (Stampfer & Bartley, 1985; Garbe et al., 2009). UVC irradiation was performed with UV Stratalinker 1800 from Stratagene (Cedar Creek, TX, USA). Doxorubicin was from BioMol (Plymouth Meeting, PA, USA) (Cat. # GR-319); PKH67 was from Sigma (Sigma-Aldrich Corporation, MO, USA) (Cat. # 0099K0787).

Cell cycle reporters

Lentiviruses expressing G1/S and S/G2 cell-cycle reporters (mKO2-hCdt1(30/120) [red] or mAG-hGem(1/110)[green]) were from RIKEN Bioresource Center (Tsukuba, Ibaraki Prefecture, Japan) (Sakaue-Sawano et al., 2008). Cells were infected with corresponding viruses and observed on day 2 postinfection. Cells in S-phase, expressing both markers can be seen as yellow.

Recombinant retroviral vectors

H2AX-expressing lentivirus and control ‘empty’ lentivirus were a kind gift of Dr. E. Brown (University of Pennsylvania Medical School); NeuT and control (pBABE) retroviral vectors were a kind gift of Dr. C. Spangenberg (Trost et al., 2005). This version of NeuT carries the activating V664E mutation (NeuT). H-RAS V12 and control (Babe) retroviral vectors were kindly provided by Dr. S. Lowe (Gabai et al., 2009). Retro- and lenti-viruses were produced as reported before (Yaglom et al., 2007; Gabai et al., 2009). Briefly, HEK293T cells were co-transfected with plasmids expressing retroviral proteins Gag-Pol, vesicular stomatitis virus glycoprotein pseudotype and enhanced green fluorescent protein or our constructs using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). At 48 h after transfection, supernatants containing the retroviral particles were collected and frozen at −80 °C until use. Cells were infected with diluted supernatant in the presence of 10 μg mL⁻¹ Polybrene overnight, and were selected with puromycin (0.75 μg mL⁻¹) 48 h after infection. Retroviral vectors expressing enhanced green fluorescent protein was used as infection efficiency indicator: usually approximately 90% of cells were fluorescent 2 days after infection. Lentiviruses were produced the same except for HEK293T transfections lentivirus-specific packaging plasmids psPAX2 and PMD2.G from Addgene were used.

Immunoblotting

Cells lysates were prepared as described (Yaglom et al., 2007). Antibodies used in the study were β-actin from Sigma; phospho-Chk1 (Ser345), p21 from BD Pharmingen (San Diego, CA, USA); NeuT (Ser 1981) were the kind gift from Dr. T Kowalik (University of Massachusetts Medical Center, Worchester); γH2AX (Ser139) were from Millipore (Billerica, MA, USA) (Cat.# 05-636). Rabbit Polyclonal Ki 67 Antibody was from Thermo Scientific (Barrington, IL, USA) (Cat. #:RB-9043-P), and γ-tubulin was from Santa Cruz (sc-17787; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA). Quantification of blots was performed using Quantity One software (Bio-Rad, Waltham, MA, USA).

Cell growth analysis

Growth rate

Cell growth curve was performed by seeding 5000 cells per well in 12-well plate; on three consecutive days, cells were harvested by trypsinization and cell numbers were counted with Scepter Automated Cell Counter (Millipore).

Measurement of cellular diameters was performed using AuxioVision software.

FACS analysis was performed on BD Biosciences (San Jose, CA, USA) FACSCalibur, and data were analyzed with BD Cellquest Pro v5.2 software.

PKH67 assay was carried out according to manufacture protocol. Briefly, PKH67 dye was added to trypsinized single-cell suspension, incubated for 5 min; staining was stopped by adding equal volume of horse serum for 1 min. Cells were washed twice and plated. First images were taken 16 h after plating.

H&E staining was carried out using Mayer’s hematoxylin solution (Sigma).

Immunohistochemistry

One-centimeter tumors were excised from animals and fixed in formalin at 4 °C overnight. Next day, tumors were trimmed into thin pieces and stained with β-gal solution for 48 h. Tumor pieces were embedded into paraffin and processed into slides (10 μm in thickness). Ki-67 staining was performed on these slides via standard ABC method (Vector Laboratory, Burlingame, CA, USA). H2AX and p21 staining was performed on paraffin-embedded sections of mammary glands.

Immunofluorescence

After treatments, cells were fixed in 100% ice-cold methanol at −20 °C for 30 min, permeabilized in 0.2% Triton X-100/PBS for 15 min at room temperature (RT), blocked in 3% BSA/PBS for 1 h, and incubated with corresponding primary Ab in the cold room (γH2AX at 1:200; Ki 67 at 1:100; γ-tubulin – 1:200; all dilutions were carried out with 3% BSA/PBS) followed by incubation with secondary Ab for 1 h at RT. Images were analyzed using Zeuss fluorescent microscope (Carl Zeiss, Oberkochen, Germany).

Mouse xenografts and cell-culture re-established from tumor

Animal maintenance and experiments were conducted in compliance with the guidelines of the Institutional Animal Care and Use Committee. Briefly, NeuT-infected MCF-10A cells were trypsinized, mixed at 1:1 ratio with matrigel (BD Scientific, San Jose, CA, USA) and 0.5 million cells were injected subcutaneously into 6-week-old female NCR nude mice (Taconic, Hudson, NY, USA). Tumor growth was monitored weekly.

When developed tumors reached 0.5 cm, they were excised and immediately placed in PBS. Tumor was cut into small fragments that were placed in MCF10A media w/out serum supplemented with 100 u mL⁻¹ of collagenase (Sigma) and incubated ON at 37 °C. Next morning cell suspension was passed through cell strainer with 75 μm mesh (BD scientific). Cells were washed twice with media with serum and plated.

β-Galactosidase assay

The β-galactosidase assay was performed using X-Gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside) at pH 6.0, as described earlier (Yaglom et al., 2007). The number of stained cells was counted under a microscope from five different fields, and the proportion of stained cells was calculated. The results were expressed as mean value ± SD on three independent experiments.

MN assay

Cells were incubated with cytochalasin B (3.5 μg mL⁻¹) for 48 h, fixed in Carnoy’s fixative (3:1 methanol:acetic acid) three times for 15 min each, and stored at −20 °C. Micronucleus scoring was carried out after cells were dropped on slides, air dried, and stained with acridine orange (30 μg mL⁻¹). Per each point at least 300 bi-nucleated cells were evaluated. Three independent experiments were performed, and MN frequency was expressed as the number of MN per binucleated cell.

Statistical analysis

The data shown are means ± SE of three independent experiments.