Signature of Progression and Negative Outcome in Colorectal Cancer

Home , MCM2, MCM7, ORC2, ORC4, RAD17

Oncogene (2010) 29, 876–887 & 2010 Macmillan Publishers Limited All rights reserved 0950-9232/10 $32.00 www.nature.com/onc ORIGINAL ARTICLE A ‘DNA replication’ signature of progression and negative outcome in colorectal cancer

M-J Pillaire1,7, J Selves2,7, K Gordien2, P-A Gouraud3, C Gentil3, M Danjoux2,CDo3, V Negre4, A Bieth1, R Guimbaud2, D Trouche5, P Pasero6,MMe´chali6, J-S Hoffmann1 and C Cazaux1

1Genetic Instability and Cancer Group, Department Biology of Cancer, Institute of Pharmacology and Structural Biology, UMR5089 CNRS, University of Toulouse, University Paul Sabatier, Toulouse, France; 2INSERM U563, Federation of Digestive Cancerology and Department of Anatomo-pathology, University of Toulouse, University Paul Sabatier, Toulouse, France; 3Service of Epidemiology, INSERM U558, Faculty of Medicine, University of Toulouse, University Paul Sabatier, Alle´es Jules Guesde, Toulouse, France; 4aCGH GSO Canceropole Platform, INSERM U868, Val d’Aurelle, Montpellier, France; 5Laboratory of Cellular and Molecular Biology of Cell Proliferation Control, UMR 5099 CNRS, University of Toulouse, University Paul Sabatier, Toulouse, France and 6Institute of Human Genetics UPR1142 CNRS, Montpellier, France

Colorectal cancer is one of the most frequent cancers Introduction worldwide. As the tumor-node-metastasis (TNM) staging classification does not allow to predict the survival of DNA replication in normal cells is regulated by an patients in many cases, additional prognostic factors are ‘origin licensing’ mechanism that ensures that it occurs needed to better forecast their outcome. Genes involved in just once per cycle. Once cells enter the S-phase, the DNA replication may represent an underexplored source stability of DNA replication forks must be preserved to of such prognostic markers. Indeed, accidents during avoid susceptibility to DNA lesions or non-B DNA DNA replication can trigger ‘replicative stress’, one of the conformation. The requirement of faithful genome main features of cancer from earlier stages onward. In this duplication in dividing cells makes DNA replication study, we assessed the expression of 47 ‘DNA replication’ an important factor in cancer by limiting cancer risk genes in primary tumors and adjacent normal tissues from through preservation of genome integrity (Kunkel, 2003; a homogeneous series of 74 patients. We found that genes Hanawalt, 2007). Owing to this cardinal importance of coding for translesional (TLS) DNA polymerases, initia- DNA replication in cancer, many anticancer drugs tion of DNA replication, S-phase signaling and protection target various aspects of DNA replication. of replication forks were significantly deregulated in The key role of DNA replication in tumor prolifera- tumors. We also observed that the overexpression of tion is illustrated by many observations. Indeed, here- either the MCM7 helicase or the TLS DNA polymerase ditary forms of colon, breast, ovary and skin cancers can POLQ (if also associated with a concomitant over- be caused by mutations in DNA replication genes, such expression of firing genes) was significantly related to as translesion synthesis (for example, POLH), intra- poor patient survival. Our data suggest the existence of a S-phase signaling (for example, BRCA1/2) or mismatch ‘DNA replication signature’ that might represent a source repair genes (for example, hMLH1) (Sancar, 1994; of new prognostic markers. Such a signature could help in Marra and Boland, 1995; Bertwistle and Ashworth, understanding the molecular mechanisms underlying 1998; Masutani et al., 1999). In somatic cancers, such tumor progression in colorectal cancer patients. early mutations probably become ‘diluted’ during the Oncogene (2010) 29, 876–887; doi:10.1038/onc.2009.378; progression of the disease, making the relationship less published online 9 November 2009 obvious. However, many studies show that an alteration of genes involved in genome replication or supervision Keywords: DNA replication; S-phase checkpoint; color- promotes or favors ‘acquired’ cancers (for reviews, see ectal cancer; genetic instability; prognosis markers Kunkel, 2003; Mitchell et al., 2003). In mice, perturba- tion of the catalytic activity of the replicative DNA polymerase d increases genomic instability and accelerates tumorigenesis (Venkatesan et al., 2007). In addition, mouse fibroblasts expressing variable levels of MCM proteins, which are involved in DNA replication Correspondence: Professor C Cazaux, Institute of Pharmacology and firing, or of the MCM loader Cdt1 show numerous Structural Biology, IPBS UMR5089, CNRS University of Toulouse, University Paul Sabatier, 205 route de Narbonne, Toulouse, cedex 4, chromosomal abnormalities and form tumors in nude Midi Pyrenees 31300, France. mice (Arentson et al., 2002; Honeycutt et al., 2006). E-mail: [email protected] Colorectal cancer is one of the most frequent cancers C Cazaux and JS Hoffmann share the leadership of the Genetic worldwide, with genetic instability having a driving role Instability and Cancer Group. in this neoplasia. To clarify the importance of DNA 7These authors have equally contributed. Received 20 August 2008; revised 15 July 2009; accepted 5 October 2009; replication in colorectal adenocarcinomas, we analysed published online 9 November 2009 the variation of expression of 47 DNA replication genes DNA replication signature in colorectal cancer M-J Pillaire et al 877 in tumor versus adjacent normal colorectal tissues and MCM2, MCM7, MCM8 and SLD5) were overexpressed compared these data with disease progression and in tumors. In contrast, GEMININ, the S-phase-dependent clinical features. Our cohort included 74 patients with inhibitor of CDT1, was significantly underexpressed. We microsatellite instability-negative tumors who were not next assessed whether these modifications in gene expres- treated with neo-adjuvant therapies. sion were linked to a specific stage of disease progression. We observed that the genes involved in either the We determined that DBF4, CDC7 and CYCE1 were firing of the replication process or the maintenance of significantly overexpressed mostly in early-stage tumors, the replication fork structure were mostly overexpressed whereas GEMININ inhibition was significant in later in tumors compared with adjacent control tissues. In stages (Supplementary Table S2). contrast, genes involved in translesional (TLS) replica- Regarding the DNA polymerase family, although the tion, S-phase checkpoint were mostly repressed. Statis- expression of the replicative POLD and POLE was not tical analysis showed that some genes (POLQ, SLD5, significantly affected in tumors, we found a defective BRCA1, CYCE1, CDC7 and RUVBL1) were mainly expression of genes encoding the mitochondrial POLg deregulated in early-stage tumors, whereas others and the ‘non conventional’ TLS DNA polymerases (GEMININ, RAD9) were predominantly deregulated POLZ,i,k,l and Rev3/z (Table 1 and Figure 2), with the in late-stage carcinomas. exception of POLQ, which was significantly overex- We also observed that MCM7 overexpression was pressed mainly at early stages (Supplementary Table S2). correlated with negative outcome in colorectal cancer. We also found a lower expression of genes that sense This was also the case of a cluster of genes characterized DNA damage occurring in the S-phase, such as ATM, by concomitant overexpression of at least three ‘firing’ 53BP1, RAD17 and RAD9 (Table 1 and Figure 3a) and, genes and the translesional DNA polymerase POLQ. specifically in stage 4 tumors, RAD9 (Supplementary Interestingly, the levels of POLQ and MCM7 expression Table S2). The new MCM2-8 family member, MCM9, were not associated with the level of PCNA immuno- which is likely to have a role in an S-phase checkpoint staining, indicating that such an upregulation did not pathway, was also downregulated. In contrast, the depend on the proliferation status. S-phase checkpoint gene products that are involved in Taken together, these data suggest that, in colorectal the downstream protection of stalled DNA replication cancer, deregulation of DNA replication genes may forks RAD51, RUVBL1, BLM, BRCA1 and BRCA2, have a significant prognostic value and may be more with BRCA1 being significantly upregulated at early useful for clinicians than standard biomarkers or stages (Supplementary Table S2), as well as the the anatomo-pathological classifications. S-phase signaling mediators, CHK1 and CHK2, were overexpressed (Table 1). Finally, genes involved in the repair of replication- Results induced DNA breaks (TIP 60), cohesion of chromatids (ECO1) and resolution of replication intermediates Subsets of DNA replication genes are deregulated in (ERCC1, MUS81) were all repressed (Figure 3c). colorectal cancers TIP60 and MUS81 were downregulated in almost all Microarray studies have shown that DNA replication tumors at all stages (Supplementary Table S2). genes are often poorly expressed in human tissues To investigate whether the misregulation of replica- (NCBI GEO data sets). Therefore, we used a real-time tion genes was related to the proliferating status of PCR-based approach to generate gene expression pro- cancer tissues, we used a nonparametric Spearman’s files of 74 coupled primary colorectal carcinomas at dif- correlation test to compare the expression of 34 ferent stages of progression (Supplementary Table S1). representative replication genes with the part of We selected 47 genes known to have a role in DNA proliferating cells in tissues. The proliferation status synthesis, intra-S-phase checkpoint signaling or proces- was assessed in tumors by a parallel staining on tissue sing of DNA damage in the course of DNA elongation. microarrays of PCNA protein, a marker of S cells. We then identified those genes that were up- or Interestingly, Figure 4 indicates that the misregulation downregulated in tumors (T) compared with adjacent in tumors of SLD5, MCM9, DBF4, CDC6, ERCC1, control tissues (N). Deregulated genes were stratified POLA, POLK, RAD9, 53BP1 and BRCA2, as well as of into two groups on the basis of the number of tumors in TLS DNA polymerase POLQ genes, was not dependent which each gene was either over- or underexpressed on the proliferation status. This parallel estimation of (Table 1). As expected, both the ‘replicative’ DNA the fraction of cycling cells also showed that MCM7 polymerase/primase POLA, which is the only DNA overexpression was only moderately associated with polymerase involved in DNA replication and not in proliferation. DNA repair synthesis, and the proliferative marker As an additional control, we also microdissected in a PCNA were overexpressed. Conversely, the DNA subset of cases the proliferating bottom third of colon replication-independent p57 gene that is known to be crypts and used this normal and proliferating epithelium downregulated in colon cancer (Li et al., 2003), was to calculate T/Nm expression ratios, in which Nm is the mostly underexpressed (Table 1). gene expression in the microdissected normal adjacent Among the genes involved in ‘firing’ of DNA replica- tissue. We then tested whether T/N and T/Nm were in tion (Figure 1 and Table 1), we observed that most of agreement (Supplementary Table S3). We found that them (that is, CDC45, CDC6, CDC7, CDT1, DBF4, among genes for which expression was not associated

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 878 Table 1 Differential gene expression in tumors and adjacent normal tissues 3R gene 3R families n Over-expression Under-expression P

PCNA Replication/repair 73 52 21 0.0004 POLA (POLa) Replication 74 49 25 0.0070 POLQ (POLy) Replication /repair 72 60 12 0.0000 POLD1 (POLd) Replication /repair 73 40 33 0.4827 POLE (POLe) Replication /repair 74 35 39 0.7275 POLB (POLb) Replication /repair 67 38 29 0.3284 POLG (POLg) Replication /repair 74 16 58 0.0000 POLH (POLZ) Replication /repair 74 21 53 0.0003 POLI (POLi) Replication /repair 74 9 65 0.0000 POLK (POLk) Replication /repair 74 27 47 0.0265 POLL (POLl) Replication /repair 74 17 57 0.0000 POLM (POLm) Replication /repair 63 27 36 0.3140 REV3 (POLz) (Rev3) Replication /repair 74 8 66 0.0000 REV1 Replication /repair 74 30 44 0.1301 CDC45 Initiation 74 57 17 0.0000 CDC6 Initiation 74 68 6 0.0000 CDC7 Initiation 74 52 22 0.0006 CDT1 Initiation 73 60 13 0.0000 GEMININ Initiation 74 18 56 0.0000 DBF4 Initiation 74 52 22 0.0006 MCM2 Initiation 74 63 11 0.0000 MCM7 Initiation 74 60 14 0.0000 MCM8 Initiation/elongation 74 50 24 0.0034 SLD5 Initiation 74 61 13 0.0000 CYCE1 Initiation 73 41 32 0.3491 ORC2 Initiation 74 42 32 0.2954 ORC4 Initiation 74 21 53 0.0003 CHK1 S-checkpoint 74 61 13 0.0000 CHK2 S-checkpoint 74 52 22 0.0006 RAD17 S-checkpoint 74 12 62 0.0000 RAD9 S-checkpoint 74 20 54 0.0000 53BP1 S-checkpoint 61 1 60 0.0000 MCM9 S-checkpoint 62 14 48 0.0000 ATM S-checkpoint 74 12 62 0.0000 ATR S-checkpoint 74 40 34 0.5613 ATRIP S-checkpoint 74 34 40 0.5613 RAD51 Fork protection 73 45 28 0.0604 RUVBL1 Fork protection 74 54 20 0.0001 BLM Fork protection 74 66 8 0.0000 BRCA1 Fork protection 74 54 20 0.0001 BRCA2 Fork protection 74 70 4 0.0000 LIG1 DNA break repair 74 47 27 0.0265 TIP60 DNA break repair 74 2 72 0.0000 ERCC1 Fork resolution 74 23 51 0.0015 MUS81 Fork resolution 74 4 70 0.0000 ECO1 Cohesion 74 25 49 0.0070 P57 CCR control 74 17 57 0.0000

Abbreviation: FDR, false discovery rate. The exact number of samples available for each analysis is shown as well as the number and the proportion of over- and underexpressing samples. P-values from bilateral exact binomial tests are given uncorrected; the signiﬁcance level is evaluated using the Benjamini–Krieger–Yekutieli 2001 procedure for an overall FDR of 0.05.

with PCNA staining, most of them, that is, SLD5, the MCM loader CDT1 was correlated with that of MCM9, DBF4, CDC6, RAD9, BRCA2 and POLQ,can MCM2 (r ¼ 0.787) and MCM7 (r ¼ 0.778) (Figure 5a). probably be still classifiable as misregulated, as the over- In contrast, expression of the elongating MCM8 helicase or underexpression observed in tumors by comparing was independent of that of the two firing genes, MCM2 with the whole normal mucosa (T/N) was also found by (r ¼ 0.444) and MCM7 (r ¼ 0.581), as indicated by normalizing with corresponding microdissected prolif- the green area at the intersection between the corres- erating epithelium (T/Nm). ponding axes. We then found that expression defects of genes Clusters of DNA replication genes are concomitantly encoding the TLS polymerases POLb, POLk, POLi, deregulated POLZ, POLl and REV3/POLz, as well as the TLS We then investigated whether deregulation of a given polymerase-associated scaffold REV1, were significantly DNA replication gene in a tumor could be con- correlated (Figure 5b). In other terms, when one of these comitant to that of other DNA replication genes by seven ‘repair’ DNA polymerases was underexpressed in using a nonparametric Spearman’s test. Expression of a given colorectal tumor, the probability of observing a

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 879

Figure 1 The mRNA expression ratios (T(tumor)/N(normal)) of initiation genes in 74 tumors and adjacent control tissues. T/N mRNA expression ratios between normalized values were calculated showing that ﬁring/licensing genes are upregulated in colon cancers. T/N>1 indicates a higher expression in the tumor than in the adjacent normal tissue. T/No1 indicates a lower expression in cancers than in control tissues. For graphical representation, T/N values lower than 1 were transformed into inverse N/T values. The sample number is indicated in black boxes underneath the x-axis. P-values from the bilateral exact binomial test (change from normal) are given uncorrected; the signiﬁcance level is evaluated using the Benjamini–Krieger 2001 procedure for an overall false discovery rate (FDR) of 0.05.

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 880

Figure 2 The mRNA expression ratios (T(tumor)/N(normal)) of DNA polymerases in the 74 tumors and adjacent control tissues. T/N mRNA expression ratios between normalized values were calculated. T/N41 indicates a higher expression in the tumor sample compared with that in the adjacent normal tissue. T/No1 indicates a lower expression in cancers than in control tissues. For graphical representation, T/N values lower than 1 were transformed into inverse N/T values. The sample number is indicated in black boxes underneath the x-axis. P-values from the bilateral exact binomial test (change from normal) are given uncorrected; the signiﬁcance level is evaluated using the Benjamini–Krieger 2001 procedure for an overall false discovery rate (FDR) of 0.05. With the aim of presenting an exhaustive overview of all DNA polymerases, we have included (*) already published histograms (Lemee et al., 2006; Betous et al., 2008; Brondello et al., 2008).

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 881 concomitant downregulation of the others was high (all Deregulated expression of POLQ and MCM7 is r values >0.700). In contrast, genes encoding POLy associated with poor prognosis or the replicative DNA polymerases POLd, POLe and The ultimate goal of our study was to identify DNA POLa did not belong to this cluster (Figure 5b). replication genes, the expression level of which in Interestingly the expression of POLQ was concomitant tumors could provide information regarding patients’ to that of MCM2 or MCM7 (Figure 5a, r ¼ 0.739). survival. A log-rank test for equality of survivor Finally, we observed a concomitant deregulation of functions indicated that when an important (T/N>5) some DNA-damage sensors. Indeed, expressions of overexpression of at least three genes involved in ATM, RAD9 and RAD17 were significantly correlated licensing/firing of DNA replication (taken from among (Figure 5c, all r values >0.600). Conversely, expression CDC45, CDC6, CDT1, SLD5, MCM2 and MCM7) of kinases CHK1 and CHK2 was independent of that of was associated with a similar excess (T/N>5) of the other S-phase checkpoint genes (blue zones, ro0.400). translesional DNA polymerase POLQ, which was generally overexpressed in our cohort (Table 1 and Figure 2), higher morbidity and higher incidence could be expected (from 0.00027 in patients with underexpression of these genes to 0.00073 in patients with strong overexpression; Figure 6, P ¼ 0.050). In an attempt to analyse the deregulation of DNA replication genes at the protein level, we constructed a tissue microarray containing the 74 cancer samples and neighboring normal mucosal tissues (Figure 7a), and tested the immunoreactivity of anti-PCNA and anti- MCM7 antibodies (Figure 7b). We then evaluated the percentage of stained cells in both cancer and normal tissues (data not shown) and observed a correlation between protein expression and RNA level when PCNA (P ¼ 0.013) and MCM7 (P ¼ 0.032) were overexpressed (data not shown). The tissue microarray data indicated (Figure 7c) that MCM7 was significantly overexpressed in patients with a negative outcome (P ¼ 0.0003). Indeed, 2 years after surgery, only about 40% of patients with MCM7-overexpressing tumors (that is, including more than 72% MCM7 immunostained cells) were alive in comparison with about 90% of those with tumors characterized by lower levels of MCM7 (that is, with less than 40% marked cells).

Discussion

Besides the standard tumor-node-metastasis (TNM) classiﬁcation used by pathologists, there is a clear lack of tools to accurately predict the clinical outcome of a colorectal tumor. For example, it is unknown why 20–40% of patients with a stage II colorectal cancer (that is, early cancer without metastasis at diagnosis)

Figure 3 The mRNA expression ratios (T(tumor)/N(normal)) of genes involved in S-phase checkpoint signaling, protection of DNA replication forks and other DNA transactions in the 74 tumors and adjacent control tissues. T/N mRNA expression ratios between normalized values were calculated. T/N41 indicates a higher expression in the tumor sample compared with that in the adjacent normal tissue. T/No1 indicates a lower expression in cancers than in control tissues. For graphical representation, T/N values lower than 1 were transformed into inverse N/T values. The sample number is indicated in black boxes underneath the x-axis. P-values from the bilateral exact binomial test (change from normal) are given uncorrected; the signiﬁcance level is evaluated using the Benjamini–Krieger 2001 procedure for an overall false discovery rate (FDR) of 0.05. (a) Genes involved in S-phase checkpoint signaling and (b) in the maintenance of DNA replication forks; (c) DNA repair and cohesion genes. *Tip60 histogram has already been published (Mattera et al., 2008).

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 882

Figure 4 Correlation between the expression in tumors of representative replication genes ratios (T(tumor)/N(normal)) and parallel PCNA immunostaining on tissue microarray (TMA), assessed using a nonparametric Spearman’s correlation test. (a) Relative Spearman’s coefﬁcient was calculated by normalizing to PCNA. H, M and L indicate high (r>0.5), moderate (0.5oro0.25) and low (ro0.25) correlation between gene deregulation and proliferation status (that is, percentage of PCNA-immunostained cells in tumors), respectively. The numbers of expression ratios analysed for each gene are indicated by n in Table 1. (b). Protein or mRNA expression T/N ratios of PCNA and representative replication genes showing high (CDC7), moderate (MCM7) and low (POLQ and POLK) correlation with PCNA staining.

rapidly worsen and die. Our study focuses on the specific and a negative outcome could help clinicians to define expression in colorectal cancer of ‘DNA replication’ which colorectal cancers may have a bad prognosis. The genes involved in the initiation of DNA replication, MCM7 protein is also overexpressed in glioblastomas DNA-damage signaling, stabilization or resolution of (Facoetti et al., 2006a), astrocytomas (Facoetti et al., stalled DNA replication intermediates. 2006b) and prostate cancers (Padmanabhan et al., 2004), Our main finding related to the significant relation- and induces cervix tumors in transgenic mice (Brake ship between the overexpression of the MCM7 protein, et al., 2003). It is the only MCM family member

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 883

Figure 6 Incidence of the coexpression of POLQ and at least three licensing/initiating replication genes on overall survival. Kaplan– Meier’s analysis of the overall survival among the 74 patients according to whether POLQ and three DNA replication genes taken from among CDC45, CDC6, CDT1, SLD5, MCM2 and MCM7 are overexpressed ((T(tumor)/N(normal))>5, black lines) or not (T/No5, gray lines) in primary tumors. 25%S: time of 75% survival. i: incidence.

the promoter of which contains a MYC-binding site (Eisenman, 2001) and the upregulation of which favors tumorigenesis in mice (Honeycutt et al., 2006). Interest- ingly, MCM7-positive tumor cells are correlated with tumor metastases in colorectal cancer (Nishihara et al., 2008). J Blow and colleagues have shown in Xenopus that excess MCM7 could activate dormant origins, favoring re-replication and genetic instability (Wood- ward et al., 2006). Several studies have reported that MCM proteins are more reliable for labeling proliferative cells than are Ki67 cells, the routine histopatholo- gical marker of cell proliferation. Hence, a noninvasive assay has been already developed by R Laskey and colleagues to identify colorectal cancer by detection of MCM2 in colonocytes retrieved from the fecal surface (Davies et al., 2002), this study being further developed and currently in clinical trials. Besides MCM7, we also show in colorectal tumors a concomitant elevated expression of other genes involved in firing of DNA replication origins, such as CDT1, CDC6, CDC45, CDC7, DBF4, MCM2, MCM8 or SLD5. In contrast, the CDT1 inhibitor, GEMININ, was inhibited. Excess CDT1 in mice promotes cancer (Arentson et al., 2002), and CDT1 (and CDC6) overexpression is correlated with poor prognosis in non small-cell lung carcinoma (Karakaidos et al., 2004). Figure 5 Concomitant expression of DNA replication genes in tumors. We used a nonparametric Spearman’s correlation graphi- GEMININ is also considered as a prognostic factor in cal display to analyse the 2209 (47 Â 47) gene/gene correlations of breast and brain tumors (Gonzalez et al., 2004; Shrestha our cohort. The closer Spearman’s coefficient (r) was to 1, the more et al., 2007), probably because of its importance in the associated was the expression of the two genes. Each horizontal or regulation of replication origins (Lutzmann et al., 2006). vertical axis corresponds to a DNA replication gene. A yellow or red zone at the intersection of two axes indicates that the levels of In Xenopus, MCM8 was identified as an initiating expression of two given genes are correlated (r coefficient close to helicase also involved in DNA elongation (Maiorano one). Conversely, a blue zone indicates that they are independent. et al., 2005). We show here that human MCM8 is (a) Firing/licensing genes; (b) DNA polymerases; (c) DNA-damage overexpressed in colorectal cancer, together with the sensors. PCNA processive factor, and that this upregulation is independent of that of MCM2–7. Taken together, these results suggest that overexpression of licensing/firing

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 884 genes contributes to colorectal tumorigenesis. This could be linked to an increase in the number of active origins of replication (and therefore re-replication and aneuploidy), as already shown in yeast (Green et al., 2006). Conversely, TLS DNA polymerases were mainly concomitantly downregulated in our cohort of tumors. Aside from their role in translesion, TLS polymerases are also ‘specialized’ in key DNA transaction repairs, such as nucleotide excision repair for POLk (Ogi and Lehman, 2006), base excision repair for POLb and POLl (Sobol et al., 1996; Braithwaite et al., 2005), homologous recombination repair for POLZ (McIl- wraith et al., 2005) and nonhomologous recombination repair for POLl and POLm (Nick McElhinny et al., 2005; Capp et al., 2006). We have also shown that naturally occurring noncanonical DNA structures, which induce genomic instability in vivo (Wang et al., 2008) and are postulated to be responsible for break- point hotspots in cancer, are physiological substrates of both PolZ and Polk (Betous et al., 2008; Rey et al., 2009). Deregulation of POLk and l has already been observed in colon cancer (Albertella et al., 2005; Pan et al., 2005). In addition, loss of REV3/POLz expression promotes gross genetic instability (Wittschieben et al., 2006). The repression of TLS polymerases could favor a ‘de-specialization’ of both repair and TLS processes mediated by these enzymes, inducing genetic instability. The observed inhibition of genes involved in S-phase checkpoint could favor these genetic changes by inhibiting the rate-limiting recognition of DNA damage. Unlike all other TLS polymerases, our data confirm (Kawamura et al., 2004) that the newly identified POLQ is significantly overexpressed in colorectal tumors. POLy is the only DNA polymerase with a helicase domain and it can efficiently bypass endogenous lesions, such as AP sites and thymidine glycols (Seki et al., 2004); therefore, it could have a role in conventional DNA replication. Our study reveals that, when associated to overexpression of at least three firing genes (from among MCM2, MCM7, SLD5, CDT1, CDC6 and CDC45), upregulation of POLQ is independent of the proliferation status of tumors and is a negative prognostic marker. Interestingly we showed that con- trary to other representative A-, X- or Y-TLS, and to a smaller extent replicative DNA polymerases, excess POLQ is concomitant with the upregulation of these ‘firing’ genes in tumors (Supplementary Table S4). Figure 7 Incidence of the MCM7 expression level on overall survival. Protein expression was analysed using tissue microarray Finally, we found that S-phase checkpoint genes, such (TMAs) consisting of 74 pairs of cancer and adjacent normal as the DNA-damage sensors ATM, 53BP1, RAD17 or mucosa samples. Two different cores (1 mm in diameter) from both RAD9, are inhibited, whereas S-phase checkpoint tumor (right) and normal (left) tissue were removed from paraffin- mediators, such as BLM, CHK1, CHK2, BRCA1 or embedded tissue blocks that were prepared at the time of resection (a). They were then immunostained with anti-PCNA and anti- BRCA2, which are involved in the protection of stalled MCM7antibodies (b). (c) Kaplan–Meier survival graphs relating to replication forks, are mainly overexpressed. It was 69 patients who were divided into two groups: the lowest tercile recently suggested that the number of active origins (group 1, individuals whose percentage of immunostained cells was may be increased by modification of the breast cancer- lower than 40%, red line) and the highest tercile (group 2, associated protein 1 status (Cortez et al., 2004). As individuals whose percentage of immunostained cells was higher than 72%, blue line). To compare these survival curves, we used the the development of a tumor is probably facilitated by log-rank test. This test calculates a P-value testing the null an ‘at all costs’ DNA replication without blockage of hypothesis that survival curves are identical in the two populations. the cell cycle, we suggest that downregulation of in trans pathways (ATM, 53BP1, RAD17 and RAD9),

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 885 together with overexpression of in cis pathways and total RNA was extracted according to the manufacturer’s (BRCA1, BRCA2), may be a new sensitive signature of instructions (Qiagen, Courtaboeuf, France). The quality of aberrant proliferation. This hypothesis is reinforced by total RNA was assessed using the Agilent 2100 bio-analyser the observation that MCM9, a new MCM family using the RNA Nano Lab chip and RNA 6000 Nano Assay kit member that regulates licensing activity (Lutzmann (Agilent Technologies, Santa Clara, CA, USA). Total RNA and Mechali, 2008), is also downregulated. (1–2 mg) was reverse transcribed using the High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA, To summarize, our data show that, in matched USA). All studied genes and four control genes (18S, GAPDH, colorectal tumors, most ‘replication’ genes are misregu- HPRT and YWHAZ) were amplified in triplicate from tumor lated. We present evidence that misexpression of many and normal samples using the TaqMan Universal PCR Master of these genes does not just reflect the proliferating stage Mix and TaqMan Low Density Array technology (Applied of tumor cells and is not just a signature of dividing Biosystems). PCR amplifications were performed using the cells. Indeed, we found that it was not associated with TaqMan Low Density Array technology and the 7900HT fast the level of PCNA and that compared with dividing real-time PCR system. To normalize gene expression in control tissues, it was not dependent on the proliferation matched tumor and normal samples, QBase software (http:// status of normal mucosa. Replication genes involved in medgen.ugent.be/qbase/) was used to test the expression DNA synthesis are found to be upregulated, whereas stability of control genes. For each sample pair, the two most stable control genes were used by QBase to normalize the genes important for the cellular response to DNA expression of all transcripts. damage are mainly downregulated. These findings reinforce the recent hypothesis regarding the existence of a relationship between replication stress and tumor Tissue microarrays progression (Bartkova et al., 2006). Indeed, hyperrepli- Using a tissue microarray tool (Beecher Instruments, Alphelys, cation (more origins, more stabilized replication forks) Plaisir, France), we removed from each patient two tumors could induce more DNA damage that might lead to and two normal tissue cores (1 mm in diameter) from paraffin- genetic instability in the absence of functional S-phase embedded tissue blocks that were prepared at the time of checkpoint or DNA repair pathways, as observed in resection. Immunostaining was performed on paraffin sections our cohort. Alongside the identification of individual from tissue microarray. Antibodies were anti-PCNA (Dako, markers, our study thus suggests that subsets of ‘DNA Glostrup, Denmark) and anti-MCM7 (Abcam, Cambridge, UK). After antigen retrieval and quenching of endogenous replication’ genes could be considered as useful cancer biotin, sections were incubated at room temperature for 60 min ‘metamarkers’. with primary antibodies. Antibody binding was detected using the iVIEW detection system (Ventana, Illkirch, France). Slides were analysed using a workstation (Spot Browser, Alphelys) Materials and methods by two pathologists (JS and MD). Staining was exclusively nuclear with all antibodies. The percentage of stained cells was Patients and methods evaluated in both cancer and normal tissues and the ratios between tumor and normal tissues were calculated for each We selected 74 patients who underwent surgery for primary patient. colorectal adenocarcinoma at the Toulouse Hospital between 1999 and 2002. The eligibility criteria for inclusion were availability of frozen normal and malignant tissues for each patient and possibility of obtaining high-quality RNA and Statistical analysis DNA. Exclusion criteria were treatment with adjuvant therapy The major criteria of analysis were the individual ratios and presence of tumor microsatellite instability. Histopatho- between standardized tumor and adjacent normal tissue gene logical and clinical findings were scored according to the expressions. To take into account the non-Gaussian distribu- TNM international staging system. Microsatellite instability tion of normalized malignant versus normal gene expression status was assessed by PCR amplification and anti-hMLH1, ratios, nonparametric statistics analyses were used. Binomial -hMSH2 and -hMSH6 immunostaining. Before RNA and exact tests evaluated the significance of gene over- and DNA extraction, a frozen section was cut and stained with underexpression. Correlation between genes was assessed hematoxylin and eosin to evaluate the percentage of cancer using a nonparametric Spearman’s correlation. Comparison cells. Malignant cells accounted for at least 50% of the tumor of expression levels in samples with different tumor stages was area. Adjacent normal tissues were only mucosa. Observation performed using a nonparametric Wilcoxon’s test. Log-rank time was defined as the interval between surgical resection and tests were used to compare survival curves (including all last contact. Data were censored at the last follow-up for living censored observations with a minimal follow-up period of 3 patients and for those who had died. The minimal duration of years) between groups defined according to the gene expres- follow-up was 36 months and the mean duration was 43 sion’s modifications. The significance level was set for an months. This study was approved by the National Institute overall table-wide P-value of 0.05, uncorrected P-value of Cancer (INCa) following the recommendations of the was shown systematically and multiple testing was taken National Agency of Agreement and Evaluation for Health into account according to Benjamini–Krieger’s procedure. (ANAES). Tumor samples were collected in agreement with All computations were performed using Stata 9.0 SE (Stata the 2004 French Bioethics law. Corp LP, College Station, TX, USA).

Quantification of gene expression For RNA extraction, 40–80 thick frozen sections of 5–10 mm Conflict of interest thickness from tumor and adjacent normal tissues were at once homogenized in the lysis buffer of the RNeasy extraction kit The authors declare no conflict of interest.

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 886 Acknowledgements platform, IFR31, Toulouse), as well as F Viala and S Mazeres (IPBS, Toulouse) for iconography and technical assistance, We thank the ‘GSO/3R’ Consortium for helpful discussions. We respectively. This study was supported by INCa (Canceropole thank B Orsetti for CGH experiments (GSO aCGH platform, GSO, grant ACI ‘Genetic instability as a negative outcome’ CRLCC Montpellier). We also acknowledge JJ Maoret (Q-PCR 2004/07 and 2008/09 to CC; grants to JS, DT and PP).

References

Albertella MR, Lau AA, O’Connor MJ. (2005). The overexpression of Karakaidos P, Taraviras S, Vassiliou LV, Zacharatos P, Kastrinakis specialized DNA polymerases in cancer. DNA repair 4: 583–593. NG, Kougiou D et al. (2004). Overexpression of the replica- Arentson E, Faloon P, Seo J, Moon E, Studts JM, Fremont DH et al. tion licensing regulators hCdt1 and hCdc6 characterizes a subset (2002). Oncogenic potential of the DNA replication licensing of non-small-cell lung carcinomas: synergistic effect with mutant protein Cdt1. Oncogene 21: 1150–1158. p53 on tumor growth and chromosomal instability—evidence Bartkova J, Rezaei N, Liontos M, Karakaidos P, Kletsas D, Issaeva N of E2F-1 transcriptional control over hCdt1. Am J Pathol 165: et al. (2006). Oncogene-induced senescence is part of the tumori- 1351–1365. genesis barrier imposed by DNA damage checkpoints. Nature 444: Kawamura K, Bahar R, Seimiya M, Chiyo M, Wada A, Okada S et al. 633–637. (2004). DNA polymerase theta is preferentially expressed in lymphoid Bertwistle D, Ashworth A. (1998). Functions of the BRCA1 and tissues and up-regulated in human cancers. Int J Cancer 109: 9–16. BCRA2 genes. Curr Opin Genet Dev 8: 14–20. Kunkel TA. (2003). Considering the cancer consequences of altered Betous R, Rey L, Wang G, Pillaire MJ, Puget N, Selves J et al. (2008). DNA polymerase function. Cancer Cell 3: 105–110. Role of TLS DNA polymerases eta and kappa in processing Lemee F, Bavoux C, Pillaire MJ, Bieth A, Machado CR, Pena SD naturally occurring structured DNA in human cells. Mol Carcino- et al. (2006). Characterization of promoter regulatory elements genesis 48: 369–378. involved in downexpression of the DNA polymerase kappa in Braithwaite EK, Prasad R, Shock DD, Hou EW, Beard WA, colorectal cancer. Oncogene 26: 3387–3394. Wilson SH. (2005). J Biol Chem 280: 18469–18475. Li JQ, Wu F, Usuki H, Kubo A, Masaki T, Fujita J et al. (2003). Loss Brake T, Connor JP, Petereit DG, Lambert PF. (2003). Comparative of p57KIP2 is associated with colorectal carcinogenesis. Int J Oncol analysis of cervical cancer in women and in a human papilloma- 23: 1537–1543. virus-transgenic mouse model: identification of minichromosome Lutzmann M, Maiorano D, Mechali M. (2006). A Cdt1–geminin maintenance protein 7 as an informative biomarker for human complex licenses chromatin for DNA replication and prevents cervical cancer. Cancer Res 63: 8173–8180. rereplication during S phase in Xenopus. EMBO J 25: 5764–5774. Brondello JM, Pillaire MJ, Rodriguez C, Gourraud PA, Selves J, Lutzmann M, Mechali M. (2008). MCM9 binds Cdt1 and is required Cazaux C et al. (2008). Novel evidences for a tumor suppressor for the assembly of pre-replication complexes. Mol Cell 31: 190–200. role of Rev3, the catalytic subunit of Pol zeta. Oncogene 27: Maiorano D, Cuvier O, Danis E, Mechali M. (2005). MCM8 is an 6093–6101. MCM2–7-related protein that functions as a DNA helicase during Capp JP, Boudsocq F, Bertrand P, Laroche-Clary A, Pourquier P, replication elongation and not initiation. Cell 120: 315–328. Lopez B et al. (2006). The DNA polymerase lambda is required for Marra G, Boland CR. (1995). Hereditary nonpoloposis colorectal the repair of non-compatible DNA double strand breaks by NHEJ cancer (HNPCC): the syndrome, the genes and historical perspec- in mammalian cells. Nucleic Acids Res 34: 2998–3007. tives. J Nat Cancer Inst 87: 1114–1125. Cortez D, Glick G, Elledge SJ. (2004). Minichromosome maintenance Masutani C, Kusumoto R, Yamada A, Dohmae N, Yokoi M, proteins are direct targets of the ATM and ATR checkpoint kinases. Yuasa M et al. (1999). The XPV (xeroderma pigmentosum Proc Natl Acad Sci USA 101: 10078–10083. variant) gene encodes human DNA polymerase eta. Nature 399: Davies RJ, Freeman A, Morris LS, Bingham S, Dilworth S, Scott I 700–704. et al. (2002). Analysis of minichromosome maintenance proteins as Mattera L, Escaffit F, Pillaire MJ, Selves J, Tyteca S, Hoffmann JS a novel method for detection of colorectal cancer in stool. Lancet et al. (2008). The p400/Tip60 ratio is critical for colorectal cancer 359: 1917–1919. cells proliferation through DNA damage response pathways. Eisenman RN. (2001). Deconstructing myc. Genes Dev 15: 2023–2030. Oncogene 28: 1506–1517. Facoetti A, Ranza E, Benericetti E, Ceroni M, Tedeschi F, Nano R. McIlwraith M, Vaisman A, Liu Y, Fanning E, Woodgate R, West S. (2006a). Minichromosome maintenance protein 7: a reliable tool for (2005). Human DNA polymerase eta promotes DNA synthesis from glioblastoma proliferation index. Anticancer Res 26: 1071–1075. strand invasion intermediates of homologous recombination. Mol Facoetti A, Ranza E, Grecchi I, Benericetti E, Ceroni M, Morbini P Cell 20: 783–792. et al. (2006b). Immunohistochemical evaluation of minichromo- Mitchell JR, Hoeijmakers JH, Niedernhofer LJ. (2003). Divide and some maintenance protein 7 in astrocytoma grading. Anticancer Res conquer: nucleotide excision repair battles cancer and ageing. Curr 26: 3513–3516. Opin Cell Biol 15: 232–240. Gonzalez MA, Tachibana KE, Chin SF, Callagy G, Madine MA, Nick McElhinny S, Havener J, Garcia-Diaz M, Juarez R, Bebenek K, Vowler SL et al. (2004). Geminin predicts adverse clinical outcome Kee BL et al. (2005). A gradient of template dependence defines in breast cancer by reflecting cell-cycle progression. J Pathol 204: distinct biological roles for family X polymerases in non homo- 121–130. logous end joining. Mol Cell 19: 357–366. Green BM, Morreale RJ, Ozaydin B, Derisi JL, Li JJ. (2006). Genome- Nishihara K, Shomori K, Fujioka S, Tokuyasu N, Inaba A, Osaki M wide mapping of DNA synthesis in Saccharomyces cerevisiae reveals et al. (2008). Minichromosome maintenance protein 7 in colo- that mechanisms preventing reinitiation of DNA replication are not rectal cancer: implication of prognostic significance. Int J Oncol 33: redundant. Mol Biol Cell 17: 2401–2414. 245–251. Hanawalt PC. (2007). Paradigms for the three Rs: DNA replication, Ogi T, Lehman AR. (2006). The Y-family DNA polymerase kappa recombination, and repair. Mol Cell 28: 702–707. functions in mammalian nucleotide-excision repair. Nature Cell Biol Honeycutt KA, Chen Z, Koster MI, Miers M, Nuchtern J, Hicks J 8: 640–642. et al. (2006). Deregulated minichromosomal maintenance protein Padmanabhan V, Callas P, Philips G, Trainer TD, Beatty BG. (2004). MCM7 contributes to oncogene driven tumorigenesis. Oncogene 25: DNA replication regulation protein Mcm7 as a marker of proli- 4027–4032. feration in prostate cancer. J Clin Pathol 57: 1057–1062.

Oncogene DNA replication signature in colorectal cancer M-J Pillaire et al 887 Pan Q, Fang Y, Xu Y, Zhang K, Hu X. (2005). Down-regulation of polymerase-beta in base-excision repair. Nature 379: 183–186. DNA polymerases kappa, eta, iota, and zeta in human lung, Erratum in: Nature 1996; 379: 848. Nature 1996 383: 457. stomach, and colorectal cancers. Cancer Lett 217: 139–147. Venkatesan RN, Treuting PM, Fuller ED, Goldsby RE, Norwood TH, Rey L, Sidorova J, Puget N, Boudsocq F, Biard D, Monnat RJ et al. Gooley TA et al. (2007). Mutation at the polymerase active (2009). Human DNA polymerase eta is required for common fragile site of mouse DNA polymerase delta increases genomic site stability during unperturbed DNA replication. Mol Cell Biol 29: instability and accelerates tumorigenesis. Mol Cell Biol 27: 7669– 3344–3354. 7682. Sancar A. (1994). Mechanisms of DNA excision repair. Science 266: Wang G, Carbajal S, Vijg J, DiGiovanni J, Vasquez KM. (2008). 1954–1956. DNA structure-induced genomic instability in vivo. J Nat Cancer Seki M, Masutani C, Yang LW, Schuffert A, Iwai S, Bahar I et al. Inst 100: 1815–1817. (2004). High-efﬁciency bypass of DNA damage by human DNA Wittschieben JP, Reshmi SC, Gollin SM, Wood RD. (2006). Loss of polymerase Q. EMBO J 23: 4484–4494. DNA polymerase zeta causes chromosomal instability in mamma- Shrestha P, Saito T, Hama S, Ariﬁn M, Kajiwara Y, Yamasaki F et al. lian cells. Cancer Res 66: 134–142. (2007). Geminin: a good prognostic factor in high-grade astrocytic Woodward AM, Go¨hler T, Luciani MG, Oehlmann M, Ge X, Gartner brain tumors. Cancer 109: 949–956. A et al. (2006). Excess Mcm2–7 license dormant origins of Sobol RW, Horton JK, Ku¨hn R, Gu H, Singhal RK, replication that can be used under conditions of replicative stress. Prasad R et al. (1996). Requirement of mammalian DNA J Cell Biol 173: 673–683.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene