Fhit Deficiency-Induced Global Genome Instability Promotes Mutation and Clonal Expansion

Loss of Fhit expression, encoded at chromosome fragile site FRA3B, leads to increased replication stress, genome instability and accumulation of genetic alterations. We have proposed that Fhit is a genome ‘caretaker’ whose loss initiates genome instability in preneoplastic lesions. We have characterized allele copy number alterations and expression changes observed in Fhit-deficient cells in conjunction with alterations in cellular proliferation and exome mutations, using cells from mouse embryo fibroblasts (MEFs), mouse kidney, early and late after establishment in culture, and in response to carcinogen treatment. Fhit -/- MEFs escape senescence to become immortal more rapidly than Fhit +/+ MEFs; -/- MEFs and kidney cultures show allele losses and gains, while +/+ derived cells show few genomic alterations. Striking alterations in expression of p53, p21, Mcl1 and active caspase 3 occurred in mouse kidney -/- cells during progressive tissue culture passage. To define genomic changes associated with preneoplastic changes in vivo, exome DNAs were sequenced for +/+ and -/- liver tissue after treatment of mice with the carcinogen, 7,12-dimethylbenz[a]anthracene, and for +/+ and -/- kidney cells treated in vitro with this carcinogen. The -/- exome DNAs, in comparison with +/+ DNA, showed small insertions, deletions and point mutations in more genes, some likely related to preneoplastic changes. Thus, Fhit loss provides a ‘mutator’ phenotype, a cellular environment in which mild genome instability permits clonal expansion, through proliferative advantage and escape from apoptosis, in response to pressures to survive.


Introduction
In hereditary cancers, genomic instability resulting from mutations in DNA repair genes, known as caretaker genes, drives cancer development, but sequencing of many nonfamilial cancers has not detected frequent mutations in DNA repair genes. Thus, for sporadic cancers the molecular basis of genomic instability is not known. A prevailing view for the development of genome instability in sporadic cancers is that it is due to oncogene activation at some point during cancer development. According to this view, supported mainly by oncogene overexpression experiments, the mutation patterns of specific tumor suppressor genes, such as TP53, support the oncogene-induced DNA replication stress model, which attributes genomic instability and suppressor gene mutations to oncogene-induced DNA damage [1].
It has been known for years that common chromosome fragile loci, that are expressed as chromatid gaps or breaks in blood cells of all individuals under conditions of mild replication stress [2], are involved in chromosome deletions in early stages of preneoplasia [3][4][5] before oncogene activation. These deletions, seen in developing cancers as the most frequent genetic alterations, have been considered passenger events rather than drivers [6] of clonal expansion. The human Fhit protein is encoded by the FHIT/FRA3B locus [7], the most active fragile locus in human lymphoblasts, though not the most fragile region in cells derived from other tissues [8]. Fhit expression is lost or reduced in many precancerous lesions and in >50% of cancers of humans [9]. Furthermore, Fhit knockout (Fhit -/-) mice are highly susceptible to carcinogeninduction of tumors and FHIT replacement in these tumors, by gene therapy, induced apoptosis and significantly reduced tumor burden [10][11][12]. Numerous reports have confirmed that the FHIT gene is a preferential target of allelic deletion and that Fhit inactivation has roles in initiation, development and progression of cancers ( [13] for review), and we have recently reported that Fhit protein deficiency causes reduced expression of thymidine kinase, subsequent dTTP imbalance, impaired DNA replication fork progression, and spontaneous DNA breaks that are transmitted to daughter cells, leading to genome instability [14]. Genomic instability is observed in human precancerous lesions, and concurrent loss of Fhit expression has been detected in precancerous lesions, suggesting that, due to fragile site susceptibility to replication fork stress, Fhit loss is among the earliest changes to occur in the preneoplastic process [4,5]. We have concluded that loss of Fhit, a genome 'caretaker', initiates the onset of genomic instability in precancerous lesions that drives tumorigenesis and links common fragile site instability to genomic instability and cancer development. Following our finding that loss of Fhit expression leads to accumulation of DNA damage in cells established from Fhit -/-tissues [14], the goal of the current research was to illustrate the consequences of loss of Fhit caretaker function by demonstrating the 'mutator' phenotype of Fhit-deficient cells and tissues.

Ethics Statement
The experiments involving isolation of mouse tissues for DNA analysis and for establishment of cell lines was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocols for these experiments were approved by The Ohio State University Institutional Animal Care and Use Committee (IACUC, Protocol Numbers: 2007A0174-R2 and 2008A0151-R1). All efforts were made to minimize suffering as outlined in the protocols and approved by the IACUC.

Cell lines, mouse tissues and reagents
We previously established mouse embryo fibroblast (MEF)derived cell lines from 3 Fhit +/+ and 3 Fhit -/-embryos for each genotype, showed that -/-cell lines became immortalized at early tissue culture passage and exhibited Copy Number Variations (CNVs) [14]. Since most human cancers derive from epithelial cells of major organs, we have also established epithelial cell lines from +/+ and -/-baby mouse kidney tissue, cloned lines from these cultures, compared proliferation characteristics and examined the effect of carcinogen treatment on +/+ and -/-cells. To define consequences of the Fhit loss-induced genome instability in vitro, we have examined CNVs and mRNA expression in these mouse cell models. To examine in vivo effects of genome instability in -/-mouse tissue, we have treated +/+ and -/-mice with 7,12-Dimethylbenz[a]anthracene (DMBA, Sigma Aldrich), a carcinogen known to induce mutations in mouse liver [15], and have compared genome-wide exome sequences of +/+ and -/liver DNAs at 1 and 4 weeks post DMBA injection.
Fhit +/+ and Fhit -/-mouse kidney cells from C57Bl/B6 background mice were cultured in MEM with 10% FBS and 100 μg/ml gentamicin. At passage (P)15, cells were plated at a low density (100 cells per 100 mm culture dish). After 10-12 days, 8 randomly chosen colonies were isolated and designated +/+ clones 1-8 and -/-clones 1-8. Fhit -/-cells that survived carcinogen treatment were established after exposure of +/+ and -/-mouse kidney cells to DMBA at P20; 1 x 10 5 cells from growing cultures were seeded in 60 mm dishes in medium including 20 µM DMBA; 3 days after seeding, the medium was replaced with normal medium and cells were cultivated. Two weeks later, no colonies were observed in +/+ mouse kidney dishes, whereas 13 colonies were observed among -/-mouse kidney dishes. The colonies were reseeded and cells expanded for further analyses (-/-DMBA-surviving cells).

Copy number variation analysis
Genomic DNA was isolated from +/+ and -/-mouse kidney cells at P4, +/+ and -/-cloned cells (n=3 lines), and +/+ and -/-MEFs at tissue culture P3 and P25 (n = 3 lines) using DNeasy Blood and Tissue isolation kit (Qiagen). Genomic DNA was also isolated from +/+ and -/-B6 weanling tail tissue. Genomic DNA samples were analyzed for CNVs at Jackson Labs using the Affymetrix Mouse Diversity Genotype Array. The B6 +/+ tail DNA served as reference DNA. This array simultaneously assays 623,124 single nucleotide polymorphisms of 12 inbred mouse strains and more than 900,000 invariant genomic regions in the mouse genome [16]. After analysis of the insertion/deletion data was completed, the single nucleotide polymorphism (SNP) data was used to confirm that deletions and gains were indeed due to B6 allele losses and gains rather than to known germline deletions or to retention of 129svJ alleles. In the results section, CNV loci are classified 'Loci not known to be fragile' or 'Loci known to be fragile' referring to the reports of Djalali et al [17] and Hosseini et al [18].

Colony formation assay
For colony formation assays, cells were plated at varying densities into 100 mm culture dishes, grown for 14-21 days. Cells were then fixed and stained with crystal violet and colony numbers counted. For clonogenicity assays, cells were treated with 5 or 20 µM DMBA 24 hr before plating. After DMBA exposure, cells were washed with PBS and growth medium; 3 days later, the surviving cells were replated (2 x 10 5 cells per 100 mm dish) in growth medium and cultured for 3 wks. Cells were then fixed, stained and colonies counted; mutation frequencies were calculated as average colony formation efficiency.

Statistical Analyses
Results of clonogenicity experiments were expressed as average values of three or four measurements. Error bars for all experiments represent one standard deviation around the mean. P values were determined by Student's two-sided t test in comparisons of results from experiments on Fhit positive and negative cells.

ß-galactosidase staining for detection of senescent cells
Senescent cells were detected by ß-Galactosidase staining [19] using the Senescence ß-Galactosidase Staining Kit (Cell Signaling) and recommended protocol. Stained cells were observed and enumerated by light and phase-contrast microscopy.

Gene expression analysis
Total RNA was isolated from +/+ and -/-mouse kidney cells at P4 and P15 (n=1 for the kidney cultures) and +/+ and -/-MEFs at P3 and P27 (n=3 for the MEF cultures [14]) using the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. cDNA was synthesized using RT 2 First Strand kit (Qiagen) and gene expression was examined by RT² Profiler PCR Array (Mouse Cellular Senescence PCR Array) according to the user manual instructions. Data was analyzed by Web-Based PCR Array Data Analysis (Qiagen). Genes differentially regulated in Fhit -/-cells vs Fhit +/+ cells were reported as relative fold-changes. Relative fold-changes, 2^(-ΔΔC t ), were calculated by dividing the normalized gene expression, 2^-ΔC t , in the test sample by the normalized gene expression, 2^-ΔC t , in the control sample. Specifically for the samples tested, relative fold-changes in the genes were calculated by comparing 2 -ΔC t values in +/+ mouse kidney cells at P4 compared to +/+ and -/-mouse kidney cells at P15. Foldchanges were also reported for MEFs, comparing 2 -ΔC t values for +/+ tissue culture at P3 with and -/-MEFs at tissue culture at P27.

Trp53 RT-PCR and Sequencing
Total RNA was extracted and cDNA synthesis was performed for 15 min at 42°C using 1 µg of total RNA as template, RT primer and quantitect reverse transcriptase (Qiagen Quantitect Reverse Transcription Kit, USA). PCR was then performed using Trp53 gene specific primers: forward 5' TGCTCACCCTGGCTAAAGTTCTGT 3' and reverse 5' ATGCAGACAGGCTTTGCAGAATGG 3'. Per reaction, 1 U of Platinum PCR SuperMix High Fidelity Taq polymerase was used and the cycling conditions were: pre-incubation for 5 min at 95°C, 30 cycles of 15 s at 95°C, 30 s at 54°C and 2 min at 68°C and final incubation at 4°C. PCR product quality was analyzed on a 0.8% agarose gel and purified for sequencing analysis via Qiagen QIAquick PCR Purification Kit. Sequencing samples containing 6.4 pmol of primer and 45 ng of template DNA were brought to a total of 12 µl with DNAse free water and sent for sequencing to the OSUCCC Nucleic Acid Shared Resource. Sequencing primers for the Trp53 gene were: forward 5' TGCTCACCCTGGCTAAAGTTCTGT 3'; reverse 5' ATGCAGACAGGCTTTGCAGAATGG 3'; a second forward 5' TTACCAGGGCAACTATGGCTTCCA 3'; second reverse 5' TTCGAGATGTTCCGGGAGCTGAAT3'.

Transfection
The mammalian expression vectors, pRcCMV-FHIT, pRcCMV-empty vector, pRcCMV-WT TP53 or pRcCMV-mouse Trp53 A135P, were transfected into mouse kidney cells using Lipofectamine 2000 (Invitrogen) reagent. Mouse kidney cells were plated on 60 mm 2 dishes and cultured to 60-80% confluence. Transfections were carried out with 1 µg plasmid DNA and 2 µl Lipofectamine diluted in Opti-MEM (Gibco) and incubated for 30 min at room temperature. Cells were washed in Opti-MEM, overlaid with the plasmid DNA/Lipofectamine solution and incubated overnight at 37° C. To optimize FHIT plasmid transfections, 1 µg DNA was mixed with either 2 or 4 µl Lipofectamine. Cell lysates were collected 24 h posttransfection for western blot analysis.

Exome sequencing of genomic DNAs of mouse liver tissue and mouse kidney cell lines
Genome DNA from kidney cells or liver tissue of the following genotypes was prepared for exome sequencing: wild-type (Fhit +/+ ) C57BL/6 (B6) and Fhit −/− on B6 background. B6 mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Fhit −/− mice were generated as described [10,11]. Mice were treated at 10-14 days of age with a single intraperitoneal injection of corn oil (1 mouse of each strain, mock), or with a single intraperitoneal injection of corn oil containing 20 nmol/g body weight DMBA (2 mice of each strain). Mice were sacrificed 1 or 4 wks after treatment. Liver, lung and spleen were removed from each mouse and stored in All Protect Tissue Reagent (Qiagen). The mouse kidney cell lines (+/+ P22, -/-P22, -/-cl 6 P20, -/-DMBA-surviving culture P22) were treated with DMBA as described above. Total genomic DNAs were isolated using DNeasy Blood and Tissue (Qiagen) and total protein was prepared by RIPA buffer lysis. Genomic DNA samples were analyzed for exome sequences at EdgeBio using Illumina HiSeq 2000, using the Agilent SureSelect Mouse All Exon Kit. More information about the kit can be found at this link: http://www.genomics.agilent.com/ CollectionSubpage.aspx? PageType=Product&SubPageType=ProductData&PageID=247 8 and http://www.chem.agilent.com/Library/datasheets/Public/ 5990-7235en_lo.pdf. The kit results in a ~50 Mb capture region covering ~221,784 exons and 24,306 genes. C57BL/6 (B6) DNA served as reference genome sequence.

Exome sequence data analysis
The small insertions and deletions (indels) identified from exome sequencing data using the GATK package were further filtered to remove any indel with low mapping quality (MQ<10), indel calling score (QUAL<10) or reads depth (DP<10). For all the exome data analysis, only indels passing the above filtering step were used. The indel frequency was calculated using the number of reads containing indels (reads mapped to alternative allele 1 and 2) divided by the total number of reads mapped to a specific indel location. Matlab "hist" function was used to generate histograms of indel frequency. To plot the frequency histogram, moving average of the frequency counts by Matlab "smooth" function was used to generate more smooth lines. To determine and remove common indels shared by two samples, the exact indel locations were used to cross-check between two indel summary files. To retrieve genes with indels in the coding region, RefSeq transcripts of mouse genome mm9 was used.
SNP sequences for each kidney and liver exome data set was analyzed to determine the number of genes containing clustered mutations. A gene containing a minimum of three single nucleotide variations was defined as a gene containing clustered mutations. The total number of genes containing clustered mutations was determined following the elimination of all genes that did not exhibit clustered exome mutations.

Allele copy number alterations in Fhit -/--derived cell lines and tissue
For examination of the genome instability induced by loss of Fhit expression, we assessed somatic CNVs, defined as DNA gains or losses spanning >10 kb, using genomic DNA isolated from +/+ and -/-mouse kidney cells (at P4) and cloned cells established from mouse kidney cells (n=3), with wild type (WT) +/+ C57BL/6 (B6) tail DNA as reference. We have included the CNV results reported previously for MEF cells at P3 and P25 (n=3) and -/-tail DNA [14] for comparison. Figure 1A, B summarizes CNVs of >10 kb detected across the cell types tested; the genomic changes in Figure 1A represent alterations observed at genomic sites not currently known to be fragile and Figure 1B summarizes alterations observed at chromosome fragile sites, based on the reports by Djalali et al [17] for mouse fibroblasts and Hosseini et al [18] for epithelial cells. Blue in columns indicate loss of alleles, and red gain of alleles. The mean fold-change of lost alleles ranged from 0.26 to 0.83 and of gained alleles from 1.51 to 2.77 (see Table S1 for details of individual CNVs). As reported [14] previously, CNVs, mostly allele losses, occurred more frequently in -/-MEFs than in the comparable +/+ cells. CNVs were also detected in the genome of -/-weanling tail tissue, illustrating the existence of genome instability in normal tissue not related to pre-cancerous lesions and suggesting that weanling tail tissue developed from few clones. The alleles involved in most CNVs are unique to mouse kidney or MEF-derived cells. Interestingly, the established kidney cell lines and subcloned lines of both +/+ and -/genotypes exhibit few CNVs in comparison to the -/-MEF cell lines. Most CNV loci in the tail tissue differ from those detected in the cell lines and do not correspond to known CNVs in the different mouse strains. This result suggests that specific CNVs may be selected under in vitro conditions and that different genome alterations contributed to survival of individual cell lines.

The mouse kidney cell model: establishment and comparison to the MEF model
MEF cell lines were previously established from three +/+ and three -/-embryos and examined at early and late tissue subcultures, before and after senescence and immortalization [14]. As previously reported, CNVs were not observed in +/+ MEF cell lines while several were detected in all -/-MEF cell lines (shown in Figure 1 for comparison with the CNV analysis of the kidney cell lines and clones). As noted previously, two of three post-senescence -/-MEF cell lines acquired amplifications within chromosome band 10D2 encompassing the murine Mdm2 gene ( Figure 1B), an oncogene frequently involved in cell transformation [20], and Mdm2 gene amplification correlated with ~4-fold increase in Mdm2 mRNA expression (see below for expression data).
Mouse kidney cells from P4 and each of eight cell lines subcloned from them at P17 were observed during establishment and proliferation at progressive tissue culture passages; Figure 2A illustrates changes in expression of survival-associated proteins from P4 to P22 of the +/+ and -/kidney cells. The +/+ kidney cells lose expression of Fhit gradually with passage, show increasing Survivin expression and steady Mcl1 to P22 [21,22], two modulators of apoptosis. This alteration of Fhit expression levels in +/+ cells may be due to epigenetic modifications, since we have not detected allelic losses at the fragile Fra14a2 locus within the murine Fhit gene ( Figure 1B). The most striking changes occurred in the -/kidney cells: the p53 level increased dramatically between P7 and P12 and the Cdkn1a gene product, p21 [23], an important downstream p53 target, disappeared abruptly in parallel, strongly suggesting that a mutant Trp53 gene and mutant p53 protein appeared post P7 and -/-cells carrying this mutation expanded to take over the culture by P12; though Survivin may increase slightly in parallel, Mcl1, an anti-apoptotic protein, increased in expression abruptly at P15 through P22. The absence of p21 expression is due to mutation in the p53 protein in Fhit -/-cells; when tested at P24 these cells, unlike the +/+ cells at early and late passage and -/-cells at early passage, displayed a single missense mutation, a G to C substitution, changing the amino acid at position 135 from alanine to proline; this mutation is in the DNA-binding domain of p53 and likely affects its ability to transactivate the Cdkn1a gene. Late passage -/-kidney cells expressing the mutant p53 protein were transfected with a WT p53 expression vector and p21 was re-expressed in the transfected cells ( Figure 3). Similar over-expression of FHIT in the same -/-kidney cells did not result in re-expression of p21 protein ( Figure 3) though it has been reported that Fhit overexpression induces expression of Cdkn1a mRNA [24] in H460 lung cancer cells that are negative for endogenous Fhit protein expression. This difference in results may be due to differences in cell lines used. Perhaps there is an accessory protein involved in the transcriptional activation, a protein that is missing or altered in the -/-kidney cells. Since Fhit is a cytoplasmic protein that is unlikely to act as a transcriptional activator, it is likely that its activation of Cdkn1a mRNA is indirect, and involves nuclear activating proteins.  (A) Summary of CNVs detected at chromosome loci not known to be fragile, in mouse kidney-derived cells, kidney-derived cloned cells, and pre-and post-senescence MEF cell lines derived from +/+ and -/-mice, and -/-tail tissue, using the Affymetrix Mouse Diversity Genotype Array; CNV data for MEF cultures was previously reported [14] and is included here for comparison to the kidney culture DNAs. Blue columns indicate allele loss (foldchange range, 0.26 to 0.61), and red columns indicate gain (fold-change range, 1.51 to 1.91). For more detail, see Table S1. (B) CNVs at chromosome fragile loci. CNVs at fragile loci detected in the same mouse cell lines and tissues. Blue columns indicate allele loss (fold-change range, -0.5 to -0.83) and red columns indicate gain at specific fragile chromosome loci (fold-change range, 1.50 to 2.51). During cultivation, we observed morphological heterogeneity in the Fhit -/-kidney-derived cells and isolated clonal cell populations to further examine heterogeneity in these cells; 8 cloned cell lines were established from +/+ and -/-cells, stained with crystal biored and observed by microscopy ( Figure 2B and Figure S1). The Fhit -/-clones appeared more morphologically varied than Fhit +/+ clones ( Figure 2B, Figure S1), though there were only a few variations in CNVs. CNVs were not observed in Fhit +/+ mouse kidney cells at P4, though there were some allelic gains in one of the +/+ clones (Figure 1).  [19], a marker of senescence. The results imply that the clonogenic late passage Fhit-deficient cells had acquired increased resistance to DMBA-induced apoptosis and cell cycle arrest.
To examine differences in expression of apoptosis related proteins, immunoblots were performed using cells before, and 6 and 24 h after exposure to 20 µM DMBA. The expression of cleaved-Caspase 3 and cleaved PARP were detected after DMBA treatment, indicating that DMBA induced apoptosis in kidney ( Figure 4C) and MEF cells ( Figure 5C). Apoptosisassociated proteins were similarly expressed in -/-kidney cells at P22 ( Figure 4C) and MEF-/-5 cells ( Figure 5C). Although the -/-P22 kidney cells had shown few DMBA-induced morphological changes ( Figure S2C), copious cleaved-Caspase 3 was detected in the -/-P22 cells ( Figure 4C). For investigation of cells that developed resistance to DMBA, we established DMBA-surviving -/-kidney cells from colonies that formed 2 wks after DMBA treatment. After expansion of these colonies, and treatment with 20 µM DMBA for 24 h, morphology and protein expression were again examined. The -/-kidney cell DMBA survivors showed neither apoptosis nor morphological changes after DMBA treatment ( Figure S4). Next we examined expression of apoptosis-related protein expression after DMBA treatment. Neither the increase in expression of cleaved-Caspase 3 nor cleaved-PARP was detected in DMBA surviving -/-kidney cells on further DMBA treatment ( Figure 4C, right panel), in agreement with absence of ß-gal staining of the cells.

Mechanisms of DMBA resistance differ in mouse kidney and MEF cells
We also assessed expression of apoptosis-related proteins Mcl1, Survivin and p21 in the DMBA-treated kidney cells ( Figure 4C). Mcl1 is an anti-apoptotic protein that confers resistance to Caspase-3 mediated apoptosis [22]. The level of expression of Mcl1 in Fhit +/+ kidney cells at P22 and P10 was   Figure 5D). The change in Survivin expression level in MEF cells after DMBA treatment was also interesting, because MEF +/+3 cells showed decreased Survivin expression, while MEF -/-5 cells showed an increase ( Figure  5D). Markedly decreased expression of p21 was seen in -/kidney cells at P22 without treatment. This expression was upregulated after DMBA treatment in +/+ kidney cells at P10 and P22 and -/-kidney cells at P10, whereas -/-kidney cells at P22 exhibited negligible expression of p21 after DMBA treatment ( Figure 4C). In MEF cells, induction of p21 is observed in both MEF+/+3 and MEF-/-5 cells after DMBA treatment ( Figure 5D). Thus, although -/-kidney cells and MEF cells resistant to DMBA treatment were selected, the resistance mechanisms differed. The DMBA-surviving -/-kidney cells showed highly down-regulated cleaved Caspase 3 and reduced cleaved PARP in the complete absence of p21 expression due to mutated p53; the MEF-/-DMBA survivors instead expressed WT p53 and highly activated p21 but expressed a steady level of Mcl1 and many-fold increased Survivin during DMBA treatment. Thus, the kidney -/-cells became DMBA resistant through silencing apoptotic signals and the MEF-/-cells became resistant through activation of anti-apoptotic signals.

mRNA expression profile changes in Fhit +/+ and Fhit -/cells
To examine gene expression changes in early and late passage kidney and MEF-derived cells, total RNA was isolated from +/+ and -/-mouse kidney cells at P4 and P15 and +/+ and -/-MEFs at P3 and P27 and cDNA synthesized. Senescenceassociated gene expression was assessed using the RT² Profiler PCR Array (Mouse Cellular Senescence PCR Array). Data was analyzed using the Web-Based PCR Array Data Analysis program (Qiagen). Figure 6 summarizes the results for those mRNAs that showed altered expression in the kidney cells where shades of blue illustrate varying levels of reduced mRNA and shades of red varying levels of increased mRNA (complete results for the expression studies are shown in Table  S2 and S3). Among down-regulated genes, Cd44, Cdkn1a (p21) are each down more than 10-fold in -/-kidney cells, a greater change than seen in the +/+ kidney cells; Twist1 is also down 10-fold in -/-kidney cells vs 2-fold down in +/+ kidney cells. Creg1 is down 5-fold and it has been observed that Creg1 is decreased during immortalization [26]. Nox4 [27,28] and Egr1 [29] are down 10-fold and Fn 1 5-fold. On the other hand, Cdkn1c, Tert, Ets2, Ing1, Nfkb1, Sod1, E2f1 [30], Cdkn2d and Ccnd1 are up ~2-fold in -/-kidney cells and Chek2, Cdkn2c, Rbl1 [31,32] and Chek1 are up ~5-fold and Cdk6 up 10-fold in -/-kidney cells. In Figure 6, first column, gene names marked in bold indicate possibly interconnected signal pathways that are altered in -/-kidney cells as the basis for the enhanced proliferation, reduced apoptosis/senescence, increased clonogenicity and DMBA resistance observed. Note that Trp53 is bold in Figure 6, to reflect the large increase in p53 protein observed in the -/-kidney cells. Figure 7 illustrates the varying expression patterns among both the +/+ and -/-MEF cultures. Note the increased expression of Mdm2 in two of the -/-MEF lines, along with increased expression of Atm, E2F1 and Bmi1 in two MEF lines. We did not detect common mRNA expression changes in kidney and MEF cells at late vs early passage cells in accord with the finding that MEF and kidney cells in vitro selected different signal pathways for survival in culture. Exceptions are increases in expression of E2f1 and Ccnd1 in -/-cells of both cell types, perhaps illustrating the centrality of these proteins in control of cellular proliferation.

DMBA-induced mutations are increased in Fhit -/-liver tissue and kidney cells
To examine the effect of Fhit-deficiency on in vivo carcinogen-induced mutations, the exomes of +/+ and -/mouse liver tissues (see Figure 8A for expression of Fhit in the tissues) were sequenced 1 wk and 4 wks following DMBA treatment; exomes of untreated +/+ and -/-liver tissues were also sequenced. The total number of reads for each sample ranged from 77,650,121 to 91,060,272, and at least 98.6% of reads were aligned. The mean coverage ranged from 82X to 98X, and at least 98.1% of base pairs were covered at least 10X. After the initial filtering for small insertions and deletions (indels) based on quality scores and depth of reads (see Materials and Methods), we then subtracted all indels that were identical in both the untreated +/+ and untreated -/-samples for all downstream analyses. There was a 4-fold increase in the number of indels in the -/-untreated liver tissues compared to the +/+ untreated, and a 3-fold increase in the number of indels in the DMBA-treated -/-liver tissues compared to matching +/+ liver tissues ( Figure 8D). Many of the indels in the +/+ and more in the -/-samples were present at an allele frequency of greater than 0.9. Because the previous CNV analysis [14] revealed the presence of germline variants in the -/-line, we presume that the indels present at an allele frequency approaching 0.5 or 1.0 may be germline variants. Thus, to focus on somatic mutations caused by DMBA treatment or a combination of Fhit absence and DMBA exposure, only indels present in fewer than 50% of sequencing reads were considered somatic ( Figure 8B and C). Furthermore, we excluded indels that were identical between untreated and DMBA-treated samples. In the untreated groups, the -/-tissues had 1.5-fold increase in the number of somatic indels compared to the +/+ tissues, whereas there was no significant difference in the number of somatic indels in the DMBA-treated +/+ vs -/-samples ( Figure 8D). Our analysis may be an underestimate of the fold-increase in the number of somatic indels present in the -/-DNA compared to the +/+ DNA, as the majority of indels in the -/-tissues were present at an allele frequency >0.3, while the majority of indels in the +/+ tissues were present at an allele frequency <0.3. Nevertheless, the 1.5-fold increase in somatic indels in the untreated -/confirmed in vivo the occurrence of mild genome instability in Fhit-negative cells.
We next considered possible progressive changes 1 wk and 4 wks following DMBA treatment. In the +/+ tissues, there was an increase in the number of somatic indels (allele freq < 0.3), but no other major changes. In the -/-tissues, there was an initial decrease in the number of somatic indels with allele frequency <0.3 one week after DMBA treatment. However, at both 1 and 4 wks post-DMBA treatment, there was an increase in the number of indels with an allele frequency ranging between 0.3 and 0.6 ( Figure 8B). This increase in frequency may indicate selective expansion of some liver cell populations with DMBA-induced beneficial mutations. We have also obtained exome sequences for DNA of +/+ and -/-kidney cell lines at late passage, with and without DMBA exposure, and a figure illustrating indel frequencies in the -/-cells is shown in the histogram in Figure 9A and a summary Table in Figure 9B. A striking feature of the data in Figure 9B is that each of the -/cell lines has accumulated ~3-fold more indels than the +/+ cells, though all were cultured for the same number of passages.
We have analyzed the exome sequencing data in several ways, by examining overlap of detected mutations, identification of indels unique to each individual DNA sample as in Figure 8B, and for the -/-control liver DNA vs the 1 wk and 4 wk DMBA treated liver DNAs ( Figure 8C). Many of the candidate mutations involve expanded or contracted minisatellite sequences and many involve deletions or insertions of a few nucleotides. See Table S4 for lists of indels in the +/+ and -/-liver DNAs with and without DMBA treatment, after removal of common indels and Table S5 for the list of indels in the mouse kidney cells that survived DMBA treatment.   The final column represents numbers of unique indels that occurred at low frequency in liver DNAs of individual -/-and +/+ mice, possibly due to the carcinogen treatment. Indels that were common to all the mice or common to the +/+ or -/-DNAs at these low frequencies were subtracted. Note that the largest difference is between the control, untreated mice, with 201 indels in the -/-liver vs 137 in the +/+ liver. The control livers were from B6 mice from Jackson laboratory while the -/-livers were from -/-mice backcrossed ~9x to B6 (also Jackson lab derived many years ago) and maintained by brother-sister matings for several years. Thus the B6 control DNA would exhibit many differences relative to the -/-background genes.

Discussion
In hereditary cancers, genomic instability arises as a result of germline mutations in DNA repair genes; this genome instability drives cancer development in the familial cases, as predicted by the mutator hypothesis. For sporadic cancers the molecular basis of genomic instability is not known, but sequencing of the genomes of many cancers have suggested that mutations in known DNA repair genes are infrequent, possibly ruling out the mutator hypothesis for these cancers. According to a prevailing proposal, the mutation patterns observed by cancer cell large scale sequencing, documenting TP53, ataxia telangiectasia mutated (ATM) and cyclindependent kinase inhibitor 2A (CDKN2A; which encodes p16INK4A and p14ARF) mutations, support the oncogeneinduced DNA replication stress model, which attributes genomic instability and TP53 and ATM mutations to oncogeneinduced DNA damage [1]. According to this proposal, genome instability in sporadic cancers results from the oncogene induced collapse of DNA replication forks, which in turn leads to DNA double strand breaks and genomic instability.
The mutator hypothesis predicts that in sporadic cancers, too, mutations affecting caretaker genes should occur early in cancer development and since targeted sequencing studies to look for mutations in known DNA repair and mitotic checkpoint genes failed to identify genes that were frequently mutated in sporadic human cancers, the mutator hypothesis has fallen out of favor. Our current results suggest that Fhit might fit the prediction that 'mutations affecting caretaker genes should occur early in cancer development'; certainly deletion of Fhit has been shown in many sequencing studies of multiple cancer types. Figure 10 presents a model that summarizes the results of this study showing that Fhit deficiency provides the 'soil' for mutation and precancerous changes during cancer initiation, in proliferative, senescence, apoptosis and survival pathways.

In vitro growth and genome instability in Fhit -/-vs Fhit +/+ cell lines
As the model in Figure 10 indicates, we propose that early loss of the Fhit caretaker function sets the stage for subsequent mutation accompanied by survival pressures for expansion of clones with heterogeneous genomic, likely epigenetic, and expression changes. It is interesting that the MEF-/-cell lines show more frequent copy number alterations Figure 10. Fhit loss provides a cellular environment for selective clonal expansion. Drawing of our model summarizing the results of this study. Fhit-deficiency causes dNTP pool imbalance [14] and leads to genome instability and accumulation of mutations, under selective pressures such as tissue culture conditions or carcinogen exposure. Thus, under selection for growth, Fhit-deficient cells become impaired in the ability to eliminate cells through apoptosis and senescence. Although many cells are eliminated, some survive due to precancerous changes. Thus, the genome instability induced by Fhit loss serves as the "soil" for selective growth and progression of precancerous clones. doi: 10.1371/journal.pone.0080730.g010 than the kidney cell line and derived clones. This may be deceptive since we studied only one kidney cell line from each strain, though each MEF-/-cell line shows more CNVs than the kidney -/-cell and its clones, with only a few CNVs that differ in the clones relative to the parental cell line 1 of the kidney +/+ clones has gained CNVs and the other two have just a few losses, similarly to the -/-kidney clones. This is in stark contrast to the +/+ MEFs, early and late passage, with no CNVs. These differences may be because the epithelial cells are more likely to show smaller genomic alterations and mutations and respond to survival pressures through genome and chromatin methylation, a subject for future investigation. It will be necessary to examine CNVs in more cell types from the -/-mice, including embryonic stem cells and tissues of fetal and neonatal organs as well as established cell lines to understand the patterns of CNVs in primary tissues and established cells after various environmental exposures and selective pressures.

Protein expression
Through protein and mRNA expression studies, we have investigated the consequences of loss of the Fhit caretaker function as summarized in the Figure 10 model. For the immortalization and escape from senescence of the -/-kidney cell culture, it seems likely that mutation of the Trp53 gene was the initial event that led to up-regulation of anti-apoptotic proteins and down-regulation of the executioners of apoptosis. These and associated regulatory events led to increased clonogenicity following tissue culture P12, though cleaved caspase 3 and cleaved PARP were expressed at high levels after exposure to DMBA in both the +/+ and -/-kidney cells. Clearly DMBA exposure kills many of the two cell types. A striking difference between the +/+ and -/-kidney cells was the increased frequency at which the -/-kidney cells survived DMBA exposure ( Figure 4B) and were then resistant to further killing by DMBA ( Figure 4C right panel), suggesting that the -/kidney cells exhibit a higher mutation frequency following DMBA exposure, a conclusion consistent with a 'mutator' phenotype. It will be important to establish additional epithelial -/-cell lines to examine additional mechanisms of selection for immortality and escape from apoptosis/senescence. +/+ MEFs, in contrast to the +/+ kidney cells, showed very poor clonogenicity at P30 with clonogenicity of untreated -/-MEFs up to 50-fold higher and 20-fold higher after DMBA treatment ( Figure 5A,B), again consistent with a mutator phenotype for the -/-MEFs.

mRNA Expression changes
The Mouse Cellular Senescence RT² Profiler™ PCR Array (http://www.sabiosciences.com/rt_pcr_product/HTML/ PAMM-050A.html) profiles the expression of 84 key genes involved in loss of ability of cells to divide. The cellular senescence program leads to activation of p53 and pRb signaling and to withdrawal from the cell cycle during very early tissue culture passages. The kidney cell RNAs were prepared and assessed at P4 and P15, before and after mutation of the Trp53 gene and protein. Since the RT-PCR assay did not show a large increase in the Trp53 mRNA level in the -/-kidney cells, though the +/+ Trp53 level was down relative to -/-, it is possible that the mutant p53 protein was stabilized in the -/kidney cells. Since the p21 protein was missing in -/-kidney cells ( Figure 2) and its mRNA down (Cdkn1a, Figure 6), it is likely that the other changes such as up-modulation of Rbl1, E2f1, possibly Cdk6 and Ccnd1 followed the initial changes in the p53-p21 axis. We have studied only one +/+ and -/-kidney cell pair thus far; additional -/-kidney cell lines may follow different pathways to immortality and resistance to carcinogeninduced apoptosis. Interestingly, the E2f1 mRNA level is also up in two of the MEF late passage -/-cultures. Although p53 is not mutated in the MEFs, MDM2 is up-regulated and amplified (Figure 1) in two MEFs at late passage, thus taking a post transcriptional route to p53 inactivation, and Atm mRNA level is up in the same two MEF late passage cell lines, suggesting ongoing DNA damage. In all MEF cell lines at late passage, transcription of Cdkn2a/p16 is down-regulated, as expected since this step is likely involved in immortalization.

The in vivo model and the mutator phenotype
To show that the mutator phenotype induced by Fhit loss (modeled in Figure 10) could also underlie preneoplastic changes in vivo, we examined carcinogen-induced alterations in -/-vs +/+ mouse livers. DMBA has been shown to cause preneoplastic changes in the liver of exposed mice as early as 4 wks after treatment: the mutation frequency after DMBA treatment was threefold higher in Gadd45a-null liver compared with wild-type liver [33]; authors concluded that lack of basal and DMBA-induced Gadd45a may result in enhanced tumorigenesis because of decreased DNA repair and increased mutation frequency, contributed by the genomic instability and loss of normal growth control in cells from Gadd45a-null mice.
We believe that our exome sequencing data for the kidney cell DNAs and DNA of livers of DMBA exposed mice can be similarly interpreted: the livers of the -/-mice with and without DMBA showed increased numbers of exome mutations relative to the WT counterpart mouse livers. Interestingly, many of the -/-exome sequences showed clustered mutations, ie several mutations clustered in numerous individual genes as illustrated in Figure S5. This type of clustered mutations and mutation 'showers' have recently been shown to be functionally linked with cancer development through the action of APOBEC (apolipoprotein B mRNA-editing enzyme, catalytic polypeptidelike) cytidine deaminases [34,35], especially APOBEC3B. The mouse genome includes only one APOBEC family member gene, APOBEC3 and we do not yet know if it is involved in development of these mutation clusters.
The main purpose of this study was to support the hypothesis that the Fhit loss-induced genome instability promotes, under selective conditions, the expansion of Fhitdeficient clones and the attendant mutations. The in vitro studies of the MEF and kidney cell lines provide evidence that Fhit-deficient cells have selective proliferative and survival advantages and the analysis of exome sequence alterations of the kidney cell and DMBA exposed liver tissue DNAs of Fhitdeficient mice showed enhanced mutation frequencies.
Negrini et al [1] suggested a temporal order by which the sporadic cancer hallmarks are acquired with deregulation of growth-regulating genes as the initiating event. In their model this would lead to DNA damage and DNA replication stress, which, in turn, would lead to genomic instability and selective pressure for TP53 inactivation. Then loss of p53 function would allow evasion of cell death, while the genomic instability would provide a fertile ground for additional mutations.
The results of our current study suggest that, as in hereditary cancers and as modeled in Figure 10, genome instability in sporadic cancers often occurs first due to replication stress at fragile regions, followed by inactivation of Fhit, that provides continued mild genome instability, followed by the other events as in hereditary cancers.
When the Fhit knockout mouse model was first developed and characterized, there were two noteworthy but unexplained features early in the study and another that was noted while tallying spontaneous tumors in mice between 1 and 2 years of age: first it was clear that carcinogen treatment induced manyfold more upper gastrointestinal tumors in Fhit -/-than in WT mice; secondly, some Fhit -/-mice on the mixed 129XB6 background spontaneously developed Muir-Torre-like sebaceous tumors [11], that had formerly been known to occur in mismatch repair-deficient cancers. The sebaceous tumors can now be explained by the absence of the Fhit caretaker function [36] in the -/-mice and it is likely that the exquisite sensitivity to carcinogen induction of tumors is also due, in part, to loss of this function. The third feature, a relatively high rate of systemic, lethal infections among the -/-mice might [10] also be due to global genome instability if it causes mistakes in immunoglobulin or T-cell receptor rearrangements, an hypothesis not yet explored.  Table S2.

Supporting Information
Detailed changes in expression of the Senescence Array genes in mouse kidney cells at early and late passages (related to Figure 6). Changes in gene expression between +/+ and -/-mouse kidney cells were determined via relative fold-changes in the mRNA levels assayed by RT-qPCR amplification. Fold-changes were calculated by comparing the 2 -ΔC t values in +/+ mouse kidney mRNAs at P4 to +/+ and -/-mouse kidney mRNAs at P15.  Table S3.
Detailed changes in expression of the Senescence Array genes in MEF cultures at early and late passages (related to Figure 7). Changes in gene expression between +/+ and -/-MEF cultures were determined via relative fold-changes in the mRNAs detected by RT-qPCR assays. Fold-changes represent a comparison of the 2 -ΔC t values for +/+ MEF mRNA levels at passage 3 with +/+ and -/-MEF mRNAs at passage 27. As in Table S2