Targeted and Nontargeted Effects of Ionizing Radiation That Impact Genomic Instability

Radiation-induced genomic instability, in which the progeny of irradiated cells display a high frequency of nonclonal genomic damage, occurs at a frequency inconsistent with mutation. We investigated the mechanism of this nontargeted effect in human mammary epithelial cells (HMEC) exposed to low doses of radiation. We identified a centrosome-associated expression signature in irradiated HMEC and show here that centrosome deregulation occurs in the first cell cycle after irradiation, is dose dependent, and that viable daughters of these cells are genomically unstable as evidenced by spontaneous DNA damage, tetraploidy, and aneuploidy. Clonal analysis of genomic instability showed a threshold of >10 cGy. Treatment with transforming growth factor beta1 (TGFbeta), which is implicated in regulation of genomic stability and is activated by radiation, reduced both the centrosome expression signature and centrosome aberrations in irradiated HMEC. Furthermore, TGFbeta inhibition significantly increased centrosome aberration frequency, tetraploidy, and aneuploidy in nonirradiated HMEC. Rather than preventing radiation-induced or spontaneous centrosome aberrations, TGFbeta selectively deleted unstable cells via p53-dependent apoptosis. Together, these studies show that radiation deregulates centrosome stability, which underlies genomic instability in normal human epithelial cells, and that this can be opposed by radiation-induced TGFbeta signaling.


Introduction
Radiation-induced genomic instability (RIGI) is a nontargeted radiation effect, i.e., occurring in the progeny of irradiated cells.RIGI is evidenced by increased chromosomal instability, apoptosis, and other deleterious events many generations after irradiation (1).The high frequency of genomic instability in irradiated bone marrow cells (2)(3)(4) and epithelial cells (5-7) is not consistent with a mutational mechanism.Although it is thought that RIGI may be an important determinant of radiation-induced carcinogenesis, the underlying mechanisms are poorly understood.Radiation-induced signaling clearly contributes to nontargeted effects (reviewed in ref. 8).Wright and colleagues (9) have shown in vivo that radiation induces activation of inflammatory cells, which generates additional DNA damage in bone marrow stem cells long after exposure.Bauer and colleagues (10) have shown that radiationinduced signals can eliminate transformed cells through selective apoptosis.These studies are evidence of a multicellular program of tissue response to damage that includes surveillance for abnormal cells (11).When normal signaling is altered by radiation, then surveillance of abnormal cells is compromised, and potentially genomically unstable cells accumulate and proliferate.
We have recently shown RIGI in clonally expanded, finite life span, normal human mammary epithelial cells (HMEC) as measured by aberrant karyotypes and supernumery centrosomes (7).Centrosome aberrations (CA) can induce variable mitotic outcomes including aneuploidy, genomic instability, and cell death (12).Radiation-induced CA has been reported in cancer cell lines exposed to high doses associated with prolonged cell cycle arrest (13,14), but CA that occur in tumor cells during the first cell cycle after radiation exposure have been shown to precede mitotic catastrophe and cell death, thus contributing to therapeutic cytotoxicity (14,15).To be relevant to RIGI, and ultimately to radiation carcinogenesis, CA must occur in nonmalignant, reproductively viable, epithelial cells.
In the current study, we examined CA in irradiated HMEC to determine whether they accompany, or are the source of, RIGI.We identified a centrosome gene expression signature based on published expression microarrays.We then determined that CA occur in the first cell cycle after radiation exposure, accumulate in a dose-dependent fashion in the progeny of irradiated HMEC, and are accompanied by tetraploidy and aneuploidy.CA and spontaneous DNA damage were significantly increased in cells cloned after irradiation with 50 or 100 cGy.We postulated that extracellular signaling from transforming growth factor h (TGFh), whose activity is induced by radiation in vivo and in vitro, mediated the cell fate of unstable cells.TGFh addition reduced the centrosome gene expression signature and decreased the frequency of CA, whereas TGFh inhibition increased genomic instability in irradiated and control HMEC.Rather than preventing CA, TGFh selectively deleted genomically unstable cells via apoptosis, resulting in an overall increase in population stability.Thus, endogenous TGFh suppresses radiation-induced and spontaneous genomic instability, whereas attenuation of TGFh signaling permits survival of cells with CA, resulting in a dramatic increase in aneuploid and tetraploid metaphases.These studies support the hypothesis that CA gives rise to genomic instability in the daughters of irradiated human epithelial cells but that TGFh surveillance restricts their persistence and expansion.Thus, interactions between targeted and nontargeted radiation effects in normal human epithelial cells determine the prevalence of genomic instability.

Materials and Methods
Cell culture.HMT-3522-S1, 184 A1, and MCF10A (American Type Culture Collection) were grown in serum-free medium as described (16).Cells were irradiated with the indicated doses within 6 h of seeding using 60 Co or 160 KvP X-ray; controls were processed identically.The concentrations used were 500 pg/mL TGFh (R&D Systems) and 400 nmol/L TGFh type I receptor kinase small molecule inhibitor (SMI; Calbiochem) or treated with 10 Ag/mL TGFh pan-specific neutralizing antibodies (R&D Systems) or mouse IgG as control.For colony assay, cells were seeded at clonal density (1,200/0.8cm 2 well) and grown to at least 50 cells per clone.For clonal analysis, irradiated MCF10A were grown at clonal density, and 6 to 10 colonies per dose were isolated with cloning rings before expansion in a larger vessel.Clones were then passaged as two populations for separate metaphase and centrosome analysis.In some cultures, demecolcine (Sigma) was added for 3 to 5 h to harvest mitotic cells and standard metaphase preparations were visually counted using Â40 magnification.Viability counts were performed using the Vi-Cell counting system (Beckman Coulter).For rescue experiments, cultures were washed at day 5 and supplemented for 2 d with TGFh or nutlin-3 (2 Amol/L; Calbiochem 444143).
Animals.All experiments were conducted with Lawrence Berkeley National Laboratory institutional animal research review and approval.Mammary glands were collected from Balb/C Tgfh1 +/À and +/+ mice, originally obtained from Adam Glick (National Cancer Institute), were bred and reared inhouse.The inguinal mammary glands were dissected from animals ages 6 or 18 mo and frozen tissue blocks were stored at À70jC until sectioning at 3-Am thickness.
For in vivo analysis of centrosomes, five sections from three mammary glands per genotype were examined with indirect immunofluorescence targeting pericentrin and counterstained with DAPI.Proliferation rates were determined using Ki67.Images were collected, processed, and analyzed as described (17).
Flow cytometry.Cells were fixed with methanol and stained with 50 Ag/mL propidium iodide according to the manufacturer's protocol for fluorescence-activated cell sorting analysis.Apoptosis was analyzed with a monoclonal M30-fluorescein conjugate (Roche) to detect a caspasedependent cleavage fragment of cytokeratin 18. M30-FITC positivity was defined as intensity >98th percentile in control cells.
Image acquisition and processing.Immunofluorescence images of cultured cells were obtained using a 40Â, 1.25 numerical aperture Zeiss Neofluar objective on a Zeiss Axiovert equipped with epifluorescence.Images were acquired by locating nuclei using the DAPI image without reference to the fluorochrome-labeled antibody.Centrosome number and structure were manually analyzed or analyzed using BIOQUANT image analysis software (18).Treatments were coded to avoid experimenter bias.Cells with 3 or more centrosomes, or large, misshapen centrosomes, were scored as CA.Irradiated populations were normalized to nonirradiated populations at corresponding time points.For each experimental point, a minimum of eight fields was acquired from each of three duplicate cultures.Experiments were replicated thrice unless noted otherwise.
Expression centrosome index.The expression centrosome index (exCI) was determined as reported (19,20).An important refinement was that expression values for centrin1(207209_at), tubg1(201714_at), and pericen-trin1(218014_at) were normalized to the mean expression of proliferation marker Ki67 (212020_s_at, 212021_s_at, 212022_s_at, 212023_s_at) for each data set, which improved the correlation for the centrosome genes by controlling indirectly for relative proliferation.Expression values for the three centrosome genes were obtained from published data sets from reference (16), in which microarrays were conducted in quadruplicate from sham-irradiated, irradiated, and/or TGFh-treated MCF10A cells.These values were then normalized to mean expression from the sham-treated data set.For coexpression analyses, we determined and ranked Pearson correlation coefficient (PCC) values between the exCI and all genes tested on the array over the eight relevant data sets ( four each of irradiated and irradiated treated with TGFh).As the exCI was normalized to Ki67 expression, Ki67 probes were removed from the coexpression analysis.
Statistical analysis.Differences between samples and/or treatment groups were compared with two-sided t test using Prism GraphPad software.ANOVA between groups was used for dose response experiments.K-S tests were used to compare population differences.

Results
Ionizing radiation induces CA in a rapid and dosedependent manner.A gene exCI has been shown to correlate with immunofluorescence analysis of CA and serve as a surrogate for centrosome amplification (19,20).We analyzed a previously reported gene expression microarray profiles of irradiated HMEC (16) and found that the exCI was significantly increased in three of four replicates after radiation relative to sham.We then determined the frequency of CA in nonmalignant HMEC during the first 72 hours after exposure to 50 cGy X-radiation, a dose that did not cause cell cycle arrest (data not shown).Centrosomes were monitored by either of two pericentriolar material markers, pericentrin or g-tubulin, which colocalize.The frequency of CA in the population of irradiated HMEC increased in a time-dependent manner (Fig. 1A), suggesting that such cells were viable and that CA were maintained through each generation of daughter cells.
We next exposed HMEC to graded radiation doses shortly after plating and measured CA upon reaching confluence 6 to 8 days later.CA frequency was increased linearly as a function of dose in two different HMEC cell lines (Fig. 1B).Replating a population arising from irradiated cells (2 Gy) showed that irradiated cells had nearly 2-fold more CA after subculture than controls (13.8%F 1.0% versus 7.0% F 0.7%; P < 0.001).Tetraploid (8N) metaphases indicate an ineffective postmitotic checkpoint and correlate with aneuploidy, karyotypic abnormalities, and tumorigenesis (21,22).Tetraploid metaphases were significantly increased (2.2% F 1.2% versus 0.3% F 0.2%) in the progeny of HMEC irradiated with 2 Gy compared with controls.Thus, in contrast to tumor cells (14,15), radiation-induced CA in nonmalignant HMEC does not result in mitotic catastrophe or immediate apoptosis, is not residual (i.e., cells with CA are reproductively viable), and is not a consequence of radiation-induced senescence.These data suggest that radiationinduced CAs are compatible with proliferation in normal human epithelial cells.
Genomic instability is defined as nonclonal genomic damage, is often highly variable within a given population, and is evidenced by multiple end points (23).Centrosome amplification has been postulated to drive genomic instability in breast cancer (24).Spontaneous DNA damage occurs in early malignant progression in human tissues concomitant with genomic instability (25,26).To confirm that CA and instability were contemporaneous features of the progeny of irradiated human epithelial cells, we measured CA (Fig. 1C) and spontaneous DNA damage response (DDR) foci measured using 53BP1 (Fig. 1D) in clonal outgrowths from cells irradiated with different doses.The frequency of CA in clones from irradiated cells was significantly increased (ANOVA, P < 0.05), as were DDR foci (ANOVA, P < 0.05).Furthermore, both CA and DDR foci were significantly increased in the clones from cells irradiated with 50 and 100 cGy compared with clones from sham irradiated and 10 cGy-irradiated cells (K-S test of sham and 10 cGy versus 50 and 100 cGy, P < 0.05).These data suggest a threshold of >10 cGy for RIGI as evidenced by increased CA and DDR in the viable progeny of irradiated cells.As expected, the frequency of CA and aneuploidy were correlated in HMEC clones such that clones with higher CA produced more aneuploid metaphases.Seventy-five percent (8 of 12) of clones from irradiated cells had high levels of aneuploidy compared with 20% (1 of 5) of control clones.Aneuploidy is a hallmark of epithelial tumors, and was proposed as a root cause of cancer more than a century ago (reviewed in ref. 27).
TGFB modulates centrosome amplification.One of the confounding issues in studying RIGI is that the level of response varies between cell types.We hypothesized that this variability was in part due to radiation-induced paracrine signaling that was unaccounted for.Others and we have shown that TGFh is rapidly activated in response to radiation in vivo and autocrine TGFh regulates mammary epithelial proliferation and DDRs in vivo.TGFh is activated by low doses of radiation even in the serum-free cultured conditions used in these studies (16), but the stroma is a major source of TGFh activity in vivo (28).Furthermore, TGFh is implicated in genomic instability in cultured Tgfh1 null keratinocytes (29), microsatellite instability in colon carcinoma cells (30), and intercellular induction of apoptosis, whereby nontransformed cells selectively remove transformed cells (10).As we have shown that chronic TGFh also promotes epithelial-mesenchymal transition in irradiated HMEC (16), its role in HMEC and RIGI is of interest.
We hypothesized that TGFh might be limiting in HMEC culture and that addition of TGFh would prevent CA in irradiated HMEC.However, TGFh treatment (500 pg/mL) did not affect radiationinduced CA in the first 3 days postradiation (data not shown), which indicated that TGFh did not block radiation-induced CA from occurring.Nonetheless, TGFh treatment for 6 to 8 days after radiation significantly reduced radiation-induced CA (Fig. 2A).TGFh treatment also reduced both radiation-induced and spontaneous CA in HMEC grown at clonal density (Fig. 2B).Furthermore, TGFh neutralizing antibodies, which block ligandreceptor interactions, increased CA after radiation compared with cells treated with control IgG (Fig. 2C).Because CA destabilize chromosome segregation, then TGFh should also suppress aberrant metaphases in irradiated HMEC.Radiation-induced tetraploidy was dramatically decreased by the addition of TGFh as a function of time (Fig. 2D).To further test whether TGFh mediates the cell fate decisions of irradiated cells, we used a SMI of TGFh type I receptor kinase, which impedes the signaling cascade after ligand binding (31).SMI treatment significantly increased, whereas TGFh treatment unexpectedly decreased, spontaneous CA (Fig. 3A), tetraploid metaphases (Fig. 3B), and aneuploidy (Fig. 3C) relative to control cultures.These data suggests that TGFh is crucial in mediating RIGI, but that levels under these serum-free culture conditions are suboptimal for strict control of aberrant cells.Together, these data indicate that both autocrine and paracrine TGFh affect the survival of genomically unstable epithelial cells that occur spontaneously or after radiation exposure.
Based on these data, we hypothesized that Tgfh1 heterozygote mammary epithelia would be at greater risk for centrosome deregulation.Centrosome analysis in situ requires the use of a centrosome/nucleus index (17).The centrosome index of the mammary epithelium of 6-month-old Tgfh1 heterozygote mice was significantly increased compared with wild-type mice (1.5 F 0.1 versus 1.1 F 0.1), and further increased at age 18 months (2.3 F 0.2 versus 1.3 F 0.1).To test whether there was a functional consequence of CA, we measured the frequency of tetraploid cells in mammary epithelial cells isolated from Tgfh1 heterozygote and wild-type mice.The frequency of tetraploid cells was almost doubled in Tgfh1 heterozygote compared with wild-type (4.1% F 0.4% versus 2.4% F 0.04%; P < 0.02).Thus, TGFh levels mediate the frequency of spontaneous CAs and tetraploidy in normal mammary epithelium in vivo.TGFB induces selective p53-dependent apoptosis of unstable cells.Because TGFh did not inhibit the production of CA over the first 72 hours, but effectively reduced CA frequency in established cultures, we hypothesized that cells with CA were eliminated after they were generated.We noted that the exCI signature was decreased in TGFh-treated HMEC; therefore, we examined the expression data sets for genes that were highly correlated, both positively and negatively, to exCI, after irradiation in the presence or absence of TGFh.PCC values were determined between the exCI and f22,300 probes for the microarray data  sets in each treatment.Of the genes most positively correlated with exCI in irradiated cells were three genes associated with prosurvival/antiapoptosis (DAXX, EMP1, and SPN), whereas two apoptotic-related genes (BRCA1 and APOE) were negatively correlated (Table 1).BRCA1 regulates caspase-3 cleavage, but more interestingly, decreased BRCA1 has been linked to genomic instability (32).We confirmed that both BRCA1 mRNA and protein levels were suppressed by TGFh inhibition (Supplementary Fig. S1).
We then determined that TGFh treatment significantly increased (P < 0.03) apoptosis in irradiated HMEC 48 to 72 hours postradiation compared with unirradiated controls using flow cytometry to measure M-30, an apoptotic marker (Fig. 4A).The delayed induction of apoptosis suggested that TGFh was acting selectively.To test whether TGFh selectively induced apoptosis of genomically unstable cells, untreated cultures and SMI-treated cultures were washed at day 5 and then treated with TGFh.TGFh addition increased apoptosis (Fig. 4B) concomitant with decreased frequency of CA (Fig. 4C).More importantly, TGFh rescue specifically increased apoptosis in the DNA fractions that represent tetraploid cells (Fig. 4D).Together, these data suggested that TGFh selectively induces apoptosis in genomically unstable human epithelial cells.

Discussion
Many studies over the last decade have shown that the progeny of irradiated cells exhibit increased nonclonal chromosomal rearrangements, gene amplification, and aneuploidy (reviewed in ref. 40).Given that genomic instability is highly correlated with early events in breast cancer progression (24,41), RIGI is likely an important determinant of radiation risk.However, the National Academy report on the biological effects of ionizing radiation (BEIR VII, 2006) concluded that most reports of RIGI lack mechanistic information and were thus judged premature to incorporate into the modeling of radiation epidemiologic data.We provide substantial evidence that centrosome deregulation is a mechanism to generate RIGI in human epithelial cells.RIGI measured by aberrant karyotypes occurs in clonally expanded, finite life span, normal HMEC and is accompanied by CA (7).We show here that CA occur in the first cell cycle postirradiation of nonmalignant HMEC, thus preceding RIGI.CA accumulated in a dose-dependent fashion in bulk population and were correlated with tetraploidy and anueploidy.Furthermore, HMEC cloned from cells irradiated with 50 or 100 cGy exhibit increased CA and aneuploidy, and also spontaneous DDR foci, which are highly correlated with malignant progression (25,26).Together, these data are consistent with a targeted radiation effect on centrosome regulation that initiates an unstable state in daughter cells.We then show that the cytokine TGFh plays a critical role in monitoring RIGI.TGFh addition did not inhibit the generation of CA but rather eliminated cells with established CA through a p53-dependent apoptotic mechanism.
A centrosome regulatory microarray signature that had been identified in myeloid leukemia (20,21) was evident in irradiated HMEC.This signature was positively correlated to the expression of prosurvival, antiapoptotic genes, and negatively correlated to the expression of proapoptotic genes and was reciprocally affected by TGFh.It is not known whether these genes are markers or instrumental in the responses we observed.The selective deletion of genomically unstable human epithelial cells by TGFh shown here parallels that shown by Bauer and colleagues (10) in which TGFh generated by nontransformed neighboring cells triggers selective ablation of transformed rodent fibroblasts in culture, which is augmented by low-dose radiation.Together, these studies support the existence of a surveillance network that operates between cells and tissues whose action reduces the burden of genomically unstable cells in the epithelium, and thus the risk of cancer (11).Recent mathematical models of such effects suggest that very low dose radiation exposures could stimulate surveillance and thereby reduce cancer incidence (42).Experimental studies by Evan and colleagues (43) have shown that restoration of p53 activity in p53 null mice 6 days after radiation exposure results in substantial tumor delay.In light of our prior work showing that radiation induces persistent TGFh activation in vivo and the current studies, it is reasonable to speculate that TGFh could be the endogenous trigger of this type of response.
TGFh effects are complex and paradoxical in carcinogenesis as it acts as a potent tumor suppressor early in carcinogenesis but later converts to a mediator of tumor progression and invasion (44).Part of the latter action is thought to be due to the ability of TGFh to drive epithelial to mesenchymal phenotypic switching in cancer cells with activated mitogen-activated protein kinase (45).We have shown that TGFh treatment, as might be derived from the stroma, of irradiated nonmalignant HMEC also induces epithelial to mesenchymal transition (EMT) and aberrant epithelial morphogenesis (16,46).Thus, our cell culture model consisting of HMEC exposed to low doses of radiation and cultured in TGFh is further evidence of the two-edged sword of TGFh.TGFh signaling in irradiated tissues could either both promote neoplastic progression (e.g., EMT) or be directed toward re-establishment of homeostasis and the elimination of abnormal cells.Our interpretation of the relationship between these phenomena observed in cell culture is that TGFh can delete genomically unstable cells that arise as a consequence of radiation, but that continued, chronic exposure induces the remaining population to undergo EMT.The complexity of radiation effects mediated by TGFh will require further study  using mouse models to determine whether, and under what circumstances, TGFh reduces the risk of cancer by suppressing RIGI, or if it accelerates radiogenic carcinogenesis by promoting EMT.CA seem to be a targeted radiation effect (i.e., occurring directly in the irradiated cells) because they are proportionally induced with doses of 10 cGy and above (Fig. 1B).TGFh-mediated aberrant HMEC morphogenesis is also acutely sensitive to radiation exposures of as little as 10 cGy, but the dose responses are quite distinct.IR acts more like a switch in priming cells to undergo EMT in that irradiation with either 2 or 200 cGy seem to be equally effective. 1Likewise, there does not seem to be dose-dependent induction of RIGI (7).Perhaps more importantly, although cells irradiated with 10 cGy exhibited CA, CA and DDR foci in cells cloned after 10 cGy were comparable with controls, suggesting that there is a threshold.We speculate that low level RIGI was efficiently suppressed by autocrine TGFh.Surveillance by endogenous TGFh was evident in that various indices of genomic instability increased when autocrine TGFh signaling was inhibited by SMI or neutralizing antibodies in both control and irradiated (2 Gy) HMEC.
In summary, our data indicate that RIGI results from a targeted effect that generates an unstable state in nontargeted daughters but is offset by selective deletion mediated by extracellular signaling.

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.

Figure 1 .
Figure 1.Centrosome deregulation precedes RIGI.A, the frequency of CA increased as a function of time after irradiation with 50 cGy in MCF10A.These data were normalized to the frequency of CA in sham-treated cells at corresponding time points.Points, mean is plotted from 3 to 7 experiments per time point and condition; bars, SE.B, dose dependent CA induction in the progeny of HMT3522 S1 or MCF10A exposed to graded radiation doses.Points, mean, n = 3 (MCF10A ) or 6 (S1 ) experiments; bars, SE.C, MCF10A cells exposed to the indicated doses were cloned, expanded, and assayed independently 4 to 6 wk later.The frequency of CA in the clonal progeny of cells irradiated to the indicated dose was significantly increased (ANOVA, P < 0.05).D, clonal analysis of the frequency of cells with >5 spontaneous DDR foci was significantly increased (ANOVA, P < 0.05) as a function of radiation dose.The daughters of HMEC irradiated with 50 cGy or more exhibited genomic instability.

Figure 2 .
Figure 2. TGFh mediates radiation-induced CA.A, TGFh treatment of irradiated HMT3522 S1 or MCF10A exhibit fewer CA 6 to 8 d postirradiation (2 Gy).B, TGFh reduced both spontaneous and radiation-induced (2 Gy) CA in MC10A grown at clonal density.Points, mean from >40 colonies; bars, SE.C, Pan-specific TGFh neutralizing antibodies (a-TGFb ) treatment increased both spontaneous and radiation induced CA compared with IgG-treated controls.D, TGFh treatment decreased the frequency of tetraploidy in irradiated HMEC.Frequency of tetraploidy within TGFh-treated cultures is represented as a fraction of sham-treated cultures at corresponding time points.The mean FSD are shown.Student's t compared with controls are indicated as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 3 .
Figure 3. Inhibition of TGFh signaling increases spontaneous genomic instability.A, SMI inhibition of TGFh signaling in MCF10A HMEC significantly increased CA relative to control at each time point.B, SMI treatment significantly increased the frequency of tetraploid metaphases.The mean F SD and t test compared with untreated control are shown as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001.C, chromosome number per metaphase from TGFh treated, untreated control, or SMI-treated MCF10A cells was determined in at least 45 DAPI-stained metaphases per sample.SMI significantly increased aneuploidy compared with TGFh-treated cells (t test, P < 0.01).

Figure 5 .
Figure 5. Roles of TGFh in the response to IR. Cartoon depicting the proposed relationship between centrosome deregulation and the development of genomic instability.Radiation-induced genomic instability is initiated with dose-dependent centrosome deregulation in cells that subsequently escape postmitotic checkpoints.Gray arrows, TGFh-mediated processes.Solid boxes, end states (e.g., apoptosis); dashed boxes, intermediate states that have multiple cellular outcomes.

Figure 4 .
Figure 4. TGFh acts in a p53-dependent fashion to eliminate genomically unstable cells.A, TGFh treatment increased apoptosis 48 to 72 h after irradiation.B, TGFh increased apoptosis in cells that were untreated or SMI-treated for 5 d.Nutlin treatment also increased apoptosis in untreated or SMI-treated cells but did affect apoptosis in cells treated with TGFh.The mean F SD and t test compared with untreated control are shown as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001.C, either TGFh or Nutlin treatment decreased CA frequency in SMI-treated cultures.D, either TGFh or Nutlin significantly increased apoptosis in 4N cells induced by SMI treatment.